Commit 83897e8789acc07c7bc72f24ab1aff7ababaca66
1 parent
9f0cc086
Split QPDF.cc into QPDF.cc and QPDF_objects.cc
Move methods responsible for loading or keeping track of objects to QPDF_objects.cc.
Showing
3 changed files
with
13 additions
and
2924 deletions
libqpdf/CMakeLists.txt
libqpdf/QPDF.cc
| ... | ... | @@ -2,10 +2,8 @@ |
| 2 | 2 | |
| 3 | 3 | #include <qpdf/QPDF_private.hh> |
| 4 | 4 | |
| 5 | -#include <array> | |
| 6 | 5 | #include <atomic> |
| 7 | 6 | #include <cstring> |
| 8 | -#include <limits> | |
| 9 | 7 | #include <map> |
| 10 | 8 | #include <regex> |
| 11 | 9 | #include <sstream> |
| ... | ... | @@ -409,17 +407,6 @@ QPDF::findHeader() |
| 409 | 407 | return valid; |
| 410 | 408 | } |
| 411 | 409 | |
| 412 | -bool | |
| 413 | -QPDF::findStartxref() | |
| 414 | -{ | |
| 415 | - if (readToken(*m->file).isWord("startxref") && readToken(*m->file).isInteger()) { | |
| 416 | - // Position in front of offset token | |
| 417 | - m->file->seek(m->file->getLastOffset(), SEEK_SET); | |
| 418 | - return true; | |
| 419 | - } | |
| 420 | - return false; | |
| 421 | -} | |
| 422 | - | |
| 423 | 410 | void |
| 424 | 411 | QPDF::parse(char const* password) |
| 425 | 412 | { |
| ... | ... | @@ -464,1019 +451,18 @@ QPDF::warn(QPDFExc const& e) |
| 464 | 451 | } |
| 465 | 452 | m->warnings.push_back(e); |
| 466 | 453 | if (!m->suppress_warnings) { |
| 467 | - *m->log->getWarn() << "WARNING: " << m->warnings.back().what() << "\n"; | |
| 468 | - } | |
| 469 | -} | |
| 470 | - | |
| 471 | -void | |
| 472 | -QPDF::warn( | |
| 473 | - qpdf_error_code_e error_code, | |
| 474 | - std::string const& object, | |
| 475 | - qpdf_offset_t offset, | |
| 476 | - std::string const& message) | |
| 477 | -{ | |
| 478 | - warn(QPDFExc(error_code, getFilename(), object, offset, message)); | |
| 479 | -} | |
| 480 | - | |
| 481 | -void | |
| 482 | -QPDF::Xref_table::initialize_empty() | |
| 483 | -{ | |
| 484 | - initialized_ = true; | |
| 485 | - trailer_ = QPDFObjectHandle::newDictionary(); | |
| 486 | - auto rt = qpdf.makeIndirectObject(QPDFObjectHandle::newDictionary()); | |
| 487 | - auto pgs = qpdf.makeIndirectObject(QPDFObjectHandle::newDictionary()); | |
| 488 | - pgs.replaceKey("/Type", QPDFObjectHandle::newName("/Pages")); | |
| 489 | - pgs.replaceKey("/Kids", QPDFObjectHandle::newArray()); | |
| 490 | - pgs.replaceKey("/Count", QPDFObjectHandle::newInteger(0)); | |
| 491 | - rt.replaceKey("/Type", QPDFObjectHandle::newName("/Catalog")); | |
| 492 | - rt.replaceKey("/Pages", pgs); | |
| 493 | - trailer_.replaceKey("/Root", rt); | |
| 494 | - trailer_.replaceKey("/Size", QPDFObjectHandle::newInteger(3)); | |
| 495 | -} | |
| 496 | - | |
| 497 | -void | |
| 498 | -QPDF::Xref_table::initialize_json() | |
| 499 | -{ | |
| 500 | - initialized_ = true; | |
| 501 | - table.resize(1); | |
| 502 | - trailer_ = QPDFObjectHandle::newDictionary(); | |
| 503 | - trailer_.replaceKey("/Size", QPDFObjectHandle::newInteger(1)); | |
| 504 | -} | |
| 505 | - | |
| 506 | -void | |
| 507 | -QPDF::Xref_table::initialize() | |
| 508 | -{ | |
| 509 | - // PDF spec says %%EOF must be found within the last 1024 bytes of/ the file. We add an extra | |
| 510 | - // 30 characters to leave room for the startxref stuff. | |
| 511 | - file->seek(0, SEEK_END); | |
| 512 | - qpdf_offset_t end_offset = file->tell(); | |
| 513 | - // Sanity check on object ids. All objects must appear in xref table / stream. In all realistic | |
| 514 | - // scenarios at least 3 bytes are required. | |
| 515 | - if (max_id_ > end_offset / 3) { | |
| 516 | - max_id_ = static_cast<int>(end_offset / 3); | |
| 517 | - } | |
| 518 | - qpdf_offset_t start_offset = (end_offset > 1054 ? end_offset - 1054 : 0); | |
| 519 | - PatternFinder sf(qpdf, &QPDF::findStartxref); | |
| 520 | - qpdf_offset_t xref_offset = 0; | |
| 521 | - if (file->findLast("startxref", start_offset, 0, sf)) { | |
| 522 | - xref_offset = QUtil::string_to_ll(read_token().getValue().c_str()); | |
| 523 | - } | |
| 524 | - | |
| 525 | - try { | |
| 526 | - if (xref_offset == 0) { | |
| 527 | - QTC::TC("qpdf", "QPDF can't find startxref"); | |
| 528 | - throw damaged_pdf("can't find startxref"); | |
| 529 | - } | |
| 530 | - try { | |
| 531 | - read(xref_offset); | |
| 532 | - } catch (QPDFExc&) { | |
| 533 | - throw; | |
| 534 | - } catch (std::exception& e) { | |
| 535 | - throw damaged_pdf(std::string("error reading xref: ") + e.what()); | |
| 536 | - } | |
| 537 | - } catch (QPDFExc& e) { | |
| 538 | - if (attempt_recovery_) { | |
| 539 | - reconstruct(e); | |
| 540 | - QTC::TC("qpdf", "QPDF reconstructed xref table"); | |
| 541 | - } else { | |
| 542 | - throw; | |
| 543 | - } | |
| 544 | - } | |
| 545 | - | |
| 546 | - initialized_ = true; | |
| 547 | -} | |
| 548 | - | |
| 549 | -void | |
| 550 | -QPDF::Xref_table::reconstruct(QPDFExc& e) | |
| 551 | -{ | |
| 552 | - if (reconstructed_) { | |
| 553 | - // Avoid xref reconstruction infinite loops. This is getting very hard to reproduce because | |
| 554 | - // qpdf is throwing many fewer exceptions while parsing. Most situations are warnings now. | |
| 555 | - throw e; | |
| 556 | - } | |
| 557 | - | |
| 558 | - // If recovery generates more than 1000 warnings, the file is so severely damaged that there | |
| 559 | - // probably is no point trying to continue. | |
| 560 | - const auto max_warnings = qpdf.m->warnings.size() + 1000U; | |
| 561 | - auto check_warnings = [this, max_warnings]() { | |
| 562 | - if (qpdf.m->warnings.size() > max_warnings) { | |
| 563 | - throw damaged_pdf("too many errors while reconstructing cross-reference table"); | |
| 564 | - } | |
| 565 | - }; | |
| 566 | - | |
| 567 | - reconstructed_ = true; | |
| 568 | - // We may find more objects, which may contain dangling references. | |
| 569 | - qpdf.m->fixed_dangling_refs = false; | |
| 570 | - | |
| 571 | - warn_damaged("file is damaged"); | |
| 572 | - qpdf.warn(e); | |
| 573 | - warn_damaged("Attempting to reconstruct cross-reference table"); | |
| 574 | - | |
| 575 | - // Delete all references to type 1 (uncompressed) objects | |
| 576 | - for (auto& iter: table) { | |
| 577 | - if (iter.type() == 1) { | |
| 578 | - iter = {}; | |
| 579 | - } | |
| 580 | - } | |
| 581 | - | |
| 582 | - std::vector<std::tuple<int, int, qpdf_offset_t>> objects; | |
| 583 | - std::vector<qpdf_offset_t> trailers; | |
| 584 | - int max_found = 0; | |
| 585 | - | |
| 586 | - file->seek(0, SEEK_END); | |
| 587 | - qpdf_offset_t eof = file->tell(); | |
| 588 | - file->seek(0, SEEK_SET); | |
| 589 | - // Don't allow very long tokens here during recovery. All the interesting tokens are covered. | |
| 590 | - static size_t const MAX_LEN = 10; | |
| 591 | - while (file->tell() < eof) { | |
| 592 | - QPDFTokenizer::Token t1 = read_token(MAX_LEN); | |
| 593 | - qpdf_offset_t token_start = file->tell() - toO(t1.getValue().length()); | |
| 594 | - if (t1.isInteger()) { | |
| 595 | - auto pos = file->tell(); | |
| 596 | - QPDFTokenizer::Token t2 = read_token(MAX_LEN); | |
| 597 | - if (t2.isInteger() && read_token(MAX_LEN).isWord("obj")) { | |
| 598 | - int obj = QUtil::string_to_int(t1.getValue().c_str()); | |
| 599 | - int gen = QUtil::string_to_int(t2.getValue().c_str()); | |
| 600 | - if (obj <= max_id_) { | |
| 601 | - objects.emplace_back(obj, gen, token_start); | |
| 602 | - if (obj > max_found) { | |
| 603 | - max_found = obj; | |
| 604 | - } | |
| 605 | - } else { | |
| 606 | - warn_damaged("ignoring object with impossibly large id " + std::to_string(obj)); | |
| 607 | - } | |
| 608 | - } | |
| 609 | - file->seek(pos, SEEK_SET); | |
| 610 | - } else if (!trailer_ && t1.isWord("trailer")) { | |
| 611 | - trailers.emplace_back(file->tell()); | |
| 612 | - } | |
| 613 | - file->findAndSkipNextEOL(); | |
| 614 | - } | |
| 615 | - | |
| 616 | - table.resize(toS(max_found) + 1); | |
| 617 | - | |
| 618 | - for (auto tr: trailers) { | |
| 619 | - file->seek(tr, SEEK_SET); | |
| 620 | - auto t = read_trailer(); | |
| 621 | - if (!t.isDictionary()) { | |
| 622 | - // Oh well. It was worth a try. | |
| 623 | - } else { | |
| 624 | - trailer_ = t; | |
| 625 | - break; | |
| 626 | - } | |
| 627 | - check_warnings(); | |
| 628 | - } | |
| 629 | - | |
| 630 | - auto rend = objects.rend(); | |
| 631 | - for (auto it = objects.rbegin(); it != rend; it++) { | |
| 632 | - auto [obj, gen, token_start] = *it; | |
| 633 | - insert(obj, 1, token_start, gen); | |
| 634 | - check_warnings(); | |
| 635 | - } | |
| 636 | - | |
| 637 | - if (!trailer_) { | |
| 638 | - qpdf_offset_t max_offset{0}; | |
| 639 | - // If there are any xref streams, take the last one to appear. | |
| 640 | - int i = -1; | |
| 641 | - for (auto const& item: table) { | |
| 642 | - ++i; | |
| 643 | - if (item.type() != 1) { | |
| 644 | - continue; | |
| 645 | - } | |
| 646 | - auto oh = qpdf.getObject(i, item.gen()); | |
| 647 | - try { | |
| 648 | - if (!oh.isStreamOfType("/XRef")) { | |
| 649 | - continue; | |
| 650 | - } | |
| 651 | - } catch (std::exception&) { | |
| 652 | - continue; | |
| 653 | - } | |
| 654 | - auto offset = item.offset(); | |
| 655 | - if (offset > max_offset) { | |
| 656 | - max_offset = offset; | |
| 657 | - trailer_ = oh.getDict(); | |
| 658 | - } | |
| 659 | - check_warnings(); | |
| 660 | - } | |
| 661 | - if (max_offset > 0) { | |
| 662 | - try { | |
| 663 | - read(max_offset); | |
| 664 | - } catch (std::exception&) { | |
| 665 | - throw damaged_pdf( | |
| 666 | - "error decoding candidate xref stream while recovering damaged file"); | |
| 667 | - } | |
| 668 | - QTC::TC("qpdf", "QPDF recover xref stream"); | |
| 669 | - } | |
| 670 | - } | |
| 671 | - | |
| 672 | - if (!trailer_) { | |
| 673 | - // We could check the last encountered object to see if it was an xref stream. If so, we | |
| 674 | - // could try to get the trailer from there. This may make it possible to recover files with | |
| 675 | - // bad startxref pointers even when they have object streams. | |
| 676 | - | |
| 677 | - throw damaged_pdf("unable to find trailer dictionary while recovering damaged file"); | |
| 678 | - } | |
| 679 | - if (table.empty()) { | |
| 680 | - // We cannot check for an empty xref table in parse because empty tables are valid when | |
| 681 | - // creating QPDF objects from JSON. | |
| 682 | - throw damaged_pdf("unable to find objects while recovering damaged file"); | |
| 683 | - } | |
| 684 | - check_warnings(); | |
| 685 | - if (!initialized_) { | |
| 686 | - initialized_ = true; | |
| 687 | - qpdf.getAllPages(); | |
| 688 | - check_warnings(); | |
| 689 | - if (qpdf.m->all_pages.empty()) { | |
| 690 | - initialized_ = false; | |
| 691 | - throw damaged_pdf("unable to find any pages while recovering damaged file"); | |
| 692 | - } | |
| 693 | - } | |
| 694 | - // We could iterate through the objects looking for streams and try to find objects inside of | |
| 695 | - // them, but it's probably not worth the trouble. Acrobat can't recover files with any errors | |
| 696 | - // in an xref stream, and this would be a real long shot anyway. If we wanted to do anything | |
| 697 | - // that involved looking at stream contents, we'd also have to call initializeEncryption() here. | |
| 698 | - // It's safe to call it more than once. | |
| 699 | -} | |
| 700 | - | |
| 701 | -void | |
| 702 | -QPDF::Xref_table::read(qpdf_offset_t xref_offset) | |
| 703 | -{ | |
| 704 | - std::map<int, int> free_table; | |
| 705 | - std::set<qpdf_offset_t> visited; | |
| 706 | - while (xref_offset) { | |
| 707 | - visited.insert(xref_offset); | |
| 708 | - char buf[7]; | |
| 709 | - memset(buf, 0, sizeof(buf)); | |
| 710 | - file->seek(xref_offset, SEEK_SET); | |
| 711 | - // Some files miss the mark a little with startxref. We could do a better job of searching | |
| 712 | - // in the neighborhood for something that looks like either an xref table or stream, but the | |
| 713 | - // simple heuristic of skipping whitespace can help with the xref table case and is harmless | |
| 714 | - // with the stream case. | |
| 715 | - bool done = false; | |
| 716 | - bool skipped_space = false; | |
| 717 | - while (!done) { | |
| 718 | - char ch; | |
| 719 | - if (1 == file->read(&ch, 1)) { | |
| 720 | - if (QUtil::is_space(ch)) { | |
| 721 | - skipped_space = true; | |
| 722 | - } else { | |
| 723 | - file->unreadCh(ch); | |
| 724 | - done = true; | |
| 725 | - } | |
| 726 | - } else { | |
| 727 | - QTC::TC("qpdf", "QPDF eof skipping spaces before xref", skipped_space ? 0 : 1); | |
| 728 | - done = true; | |
| 729 | - } | |
| 730 | - } | |
| 731 | - | |
| 732 | - file->read(buf, sizeof(buf) - 1); | |
| 733 | - // The PDF spec says xref must be followed by a line terminator, but files exist in the wild | |
| 734 | - // where it is terminated by arbitrary whitespace. | |
| 735 | - if ((strncmp(buf, "xref", 4) == 0) && QUtil::is_space(buf[4])) { | |
| 736 | - if (skipped_space) { | |
| 737 | - QTC::TC("qpdf", "QPDF xref skipped space"); | |
| 738 | - warn_damaged("extraneous whitespace seen before xref"); | |
| 739 | - } | |
| 740 | - QTC::TC( | |
| 741 | - "qpdf", | |
| 742 | - "QPDF xref space", | |
| 743 | - ((buf[4] == '\n') ? 0 | |
| 744 | - : (buf[4] == '\r') ? 1 | |
| 745 | - : (buf[4] == ' ') ? 2 | |
| 746 | - : 9999)); | |
| 747 | - int skip = 4; | |
| 748 | - // buf is null-terminated, and QUtil::is_space('\0') is false, so this won't overrun. | |
| 749 | - while (QUtil::is_space(buf[skip])) { | |
| 750 | - ++skip; | |
| 751 | - } | |
| 752 | - xref_offset = process_section(xref_offset + skip); | |
| 753 | - } else { | |
| 754 | - xref_offset = read_stream(xref_offset); | |
| 755 | - } | |
| 756 | - if (visited.count(xref_offset) != 0) { | |
| 757 | - QTC::TC("qpdf", "QPDF xref loop"); | |
| 758 | - throw damaged_pdf("loop detected following xref tables"); | |
| 759 | - } | |
| 760 | - } | |
| 761 | - | |
| 762 | - if (!trailer_) { | |
| 763 | - throw damaged_pdf("unable to find trailer while reading xref"); | |
| 764 | - } | |
| 765 | - int size = trailer_.getKey("/Size").getIntValueAsInt(); | |
| 766 | - | |
| 767 | - if (size < 3) { | |
| 768 | - throw damaged_pdf("too few objects - file can't have a page tree"); | |
| 769 | - } | |
| 770 | - | |
| 771 | - // We are no longer reporting what the highest id in the xref table is. I don't think it adds | |
| 772 | - // anything. If we want to report more detail, we should report the total number of missing | |
| 773 | - // entries, including missing entries before the last actual entry. | |
| 774 | -} | |
| 775 | - | |
| 776 | -QPDF::Xref_table::Subsection | |
| 777 | -QPDF::Xref_table::subsection(std::string const& line) | |
| 778 | -{ | |
| 779 | - auto terminate = [this]() -> void { | |
| 780 | - QTC::TC("qpdf", "QPDF invalid xref"); | |
| 781 | - throw damaged_table("xref syntax invalid"); | |
| 782 | - }; | |
| 783 | - | |
| 784 | - // is_space and is_digit both return false on '\0', so this will not overrun the null-terminated | |
| 785 | - // buffer. | |
| 786 | - char const* p = line.c_str(); | |
| 787 | - char const* start = line.c_str(); | |
| 788 | - | |
| 789 | - // Skip zero or more spaces | |
| 790 | - while (QUtil::is_space(*p)) { | |
| 791 | - ++p; | |
| 792 | - } | |
| 793 | - // Require digit | |
| 794 | - if (!QUtil::is_digit(*p)) { | |
| 795 | - terminate(); | |
| 796 | - } | |
| 797 | - // Gather digits | |
| 798 | - std::string obj_str; | |
| 799 | - while (QUtil::is_digit(*p)) { | |
| 800 | - obj_str.append(1, *p++); | |
| 801 | - } | |
| 802 | - // Require space | |
| 803 | - if (!QUtil::is_space(*p)) { | |
| 804 | - terminate(); | |
| 805 | - } | |
| 806 | - // Skip spaces | |
| 807 | - while (QUtil::is_space(*p)) { | |
| 808 | - ++p; | |
| 809 | - } | |
| 810 | - // Require digit | |
| 811 | - if (!QUtil::is_digit(*p)) { | |
| 812 | - terminate(); | |
| 813 | - } | |
| 814 | - // Gather digits | |
| 815 | - std::string num_str; | |
| 816 | - while (QUtil::is_digit(*p)) { | |
| 817 | - num_str.append(1, *p++); | |
| 818 | - } | |
| 819 | - // Skip any space including line terminators | |
| 820 | - while (QUtil::is_space(*p)) { | |
| 821 | - ++p; | |
| 822 | - } | |
| 823 | - auto obj = QUtil::string_to_int(obj_str.c_str()); | |
| 824 | - auto count = QUtil::string_to_int(num_str.c_str()); | |
| 825 | - if (obj > max_id() || count > max_id() || (obj + count) > max_id()) { | |
| 826 | - throw damaged_table("xref table subsection header contains impossibly large entry"); | |
| 827 | - } | |
| 828 | - return {obj, count, file->getLastOffset() + toI(p - start)}; | |
| 829 | -} | |
| 830 | - | |
| 831 | -std::vector<QPDF::Xref_table::Subsection> | |
| 832 | -QPDF::Xref_table::bad_subsections(std::string& line, qpdf_offset_t start) | |
| 833 | -{ | |
| 834 | - std::vector<QPDF::Xref_table::Subsection> result; | |
| 835 | - file->seek(start, SEEK_SET); | |
| 836 | - | |
| 837 | - while (true) { | |
| 838 | - line.assign(50, '\0'); | |
| 839 | - file->read(line.data(), line.size()); | |
| 840 | - auto [obj, num, offset] = result.emplace_back(subsection(line)); | |
| 841 | - file->seek(offset, SEEK_SET); | |
| 842 | - for (qpdf_offset_t i = obj; i - num < obj; ++i) { | |
| 843 | - if (!std::get<0>(read_entry())) { | |
| 844 | - QTC::TC("qpdf", "QPDF invalid xref entry"); | |
| 845 | - throw damaged_table("invalid xref entry (obj=" + std::to_string(i) + ")"); | |
| 846 | - } | |
| 847 | - } | |
| 848 | - qpdf_offset_t pos = file->tell(); | |
| 849 | - if (read_token().isWord("trailer")) { | |
| 850 | - return result; | |
| 851 | - } else { | |
| 852 | - file->seek(pos, SEEK_SET); | |
| 853 | - } | |
| 854 | - } | |
| 855 | -} | |
| 856 | - | |
| 857 | -// Optimistically read and parse all subsection headers. If an error is encountered return the | |
| 858 | -// result of bad_subsections. | |
| 859 | -std::vector<QPDF::Xref_table::Subsection> | |
| 860 | -QPDF::Xref_table::subsections(std::string& line) | |
| 861 | -{ | |
| 862 | - auto recovery_offset = file->tell(); | |
| 863 | - try { | |
| 864 | - std::vector<QPDF::Xref_table::Subsection> result; | |
| 865 | - | |
| 866 | - while (true) { | |
| 867 | - line.assign(50, '\0'); | |
| 868 | - file->read(line.data(), line.size()); | |
| 869 | - auto& sub = result.emplace_back(subsection(line)); | |
| 870 | - auto count = std::get<1>(sub); | |
| 871 | - auto offset = std::get<2>(sub); | |
| 872 | - file->seek(offset + 20 * toO(count) - 1, SEEK_SET); | |
| 873 | - file->read(line.data(), 1); | |
| 874 | - if (!(line[0] == '\n' || line[0] == '\n')) { | |
| 875 | - return bad_subsections(line, recovery_offset); | |
| 876 | - } | |
| 877 | - qpdf_offset_t pos = file->tell(); | |
| 878 | - if (read_token().isWord("trailer")) { | |
| 879 | - return result; | |
| 880 | - } else { | |
| 881 | - file->seek(pos, SEEK_SET); | |
| 882 | - } | |
| 883 | - } | |
| 884 | - } catch (...) { | |
| 885 | - return bad_subsections(line, recovery_offset); | |
| 886 | - } | |
| 887 | -} | |
| 888 | - | |
| 889 | -// Returns (success, f1, f2, type). | |
| 890 | -std::tuple<bool, qpdf_offset_t, int, char> | |
| 891 | -QPDF::Xref_table::read_bad_entry() | |
| 892 | -{ | |
| 893 | - qpdf_offset_t f1{0}; | |
| 894 | - int f2{0}; | |
| 895 | - char type{'\0'}; | |
| 896 | - // Reposition after initial read attempt and reread. | |
| 897 | - file->seek(file->getLastOffset(), SEEK_SET); | |
| 898 | - auto line = file->readLine(30); | |
| 899 | - | |
| 900 | - // is_space and is_digit both return false on '\0', so this will not overrun the null-terminated | |
| 901 | - // buffer. | |
| 902 | - char const* p = line.data(); | |
| 903 | - | |
| 904 | - // Skip zero or more spaces. There aren't supposed to be any. | |
| 905 | - bool invalid = false; | |
| 906 | - while (QUtil::is_space(*p)) { | |
| 907 | - ++p; | |
| 908 | - QTC::TC("qpdf", "QPDF ignore first space in xref entry"); | |
| 909 | - invalid = true; | |
| 910 | - } | |
| 911 | - // Require digit | |
| 912 | - if (!QUtil::is_digit(*p)) { | |
| 913 | - return {false, 0, 0, '\0'}; | |
| 914 | - } | |
| 915 | - // Gather digits | |
| 916 | - std::string f1_str; | |
| 917 | - while (QUtil::is_digit(*p)) { | |
| 918 | - f1_str.append(1, *p++); | |
| 919 | - } | |
| 920 | - // Require space | |
| 921 | - if (!QUtil::is_space(*p)) { | |
| 922 | - return {false, 0, 0, '\0'}; | |
| 923 | - } | |
| 924 | - if (QUtil::is_space(*(p + 1))) { | |
| 925 | - QTC::TC("qpdf", "QPDF ignore first extra space in xref entry"); | |
| 926 | - invalid = true; | |
| 927 | - } | |
| 928 | - // Skip spaces | |
| 929 | - while (QUtil::is_space(*p)) { | |
| 930 | - ++p; | |
| 931 | - } | |
| 932 | - // Require digit | |
| 933 | - if (!QUtil::is_digit(*p)) { | |
| 934 | - return {false, 0, 0, '\0'}; | |
| 935 | - } | |
| 936 | - // Gather digits | |
| 937 | - std::string f2_str; | |
| 938 | - while (QUtil::is_digit(*p)) { | |
| 939 | - f2_str.append(1, *p++); | |
| 940 | - } | |
| 941 | - // Require space | |
| 942 | - if (!QUtil::is_space(*p)) { | |
| 943 | - return {false, 0, 0, '\0'}; | |
| 944 | - } | |
| 945 | - if (QUtil::is_space(*(p + 1))) { | |
| 946 | - QTC::TC("qpdf", "QPDF ignore second extra space in xref entry"); | |
| 947 | - invalid = true; | |
| 948 | - } | |
| 949 | - // Skip spaces | |
| 950 | - while (QUtil::is_space(*p)) { | |
| 951 | - ++p; | |
| 952 | - } | |
| 953 | - if ((*p == 'f') || (*p == 'n')) { | |
| 954 | - type = *p; | |
| 955 | - } else { | |
| 956 | - return {false, 0, 0, '\0'}; | |
| 957 | - } | |
| 958 | - if ((f1_str.length() != 10) || (f2_str.length() != 5)) { | |
| 959 | - QTC::TC("qpdf", "QPDF ignore length error xref entry"); | |
| 960 | - invalid = true; | |
| 961 | - } | |
| 962 | - | |
| 963 | - if (invalid) { | |
| 964 | - qpdf.warn(damaged_table("accepting invalid xref table entry")); | |
| 965 | - } | |
| 966 | - | |
| 967 | - f1 = QUtil::string_to_ll(f1_str.c_str()); | |
| 968 | - f2 = QUtil::string_to_int(f2_str.c_str()); | |
| 969 | - | |
| 970 | - return {true, f1, f2, type}; | |
| 971 | -} | |
| 972 | - | |
| 973 | -// Optimistically read and parse xref entry. If entry is bad, call read_bad_xrefEntry and return | |
| 974 | -// result. Returns (success, f1, f2, type). | |
| 975 | -std::tuple<bool, qpdf_offset_t, int, char> | |
| 976 | -QPDF::Xref_table::read_entry() | |
| 977 | -{ | |
| 978 | - qpdf_offset_t f1{0}; | |
| 979 | - int f2{0}; | |
| 980 | - char type{'\0'}; | |
| 981 | - std::array<char, 21> line; | |
| 982 | - f1 = 0; | |
| 983 | - f2 = 0; | |
| 984 | - if (file->read(line.data(), 20) != 20) { | |
| 985 | - // C++20: [[unlikely]] | |
| 986 | - return {false, 0, 0, '\0'}; | |
| 987 | - } | |
| 988 | - line[20] = '\0'; | |
| 989 | - char const* p = line.data(); | |
| 990 | - | |
| 991 | - int f1_len = 0; | |
| 992 | - int f2_len = 0; | |
| 993 | - | |
| 994 | - // is_space and is_digit both return false on '\0', so this will not overrun the null-terminated | |
| 995 | - // buffer. | |
| 996 | - | |
| 997 | - // Gather f1 digits. NB No risk of overflow as 9'999'999'999 < max long long. | |
| 998 | - while (*p == '0') { | |
| 999 | - ++f1_len; | |
| 1000 | - ++p; | |
| 1001 | - } | |
| 1002 | - while (QUtil::is_digit(*p) && f1_len++ < 10) { | |
| 1003 | - f1 *= 10; | |
| 1004 | - f1 += *p++ - '0'; | |
| 1005 | - } | |
| 1006 | - // Require space | |
| 1007 | - if (!QUtil::is_space(*p++)) { | |
| 1008 | - // Entry doesn't start with space or digit. | |
| 1009 | - // C++20: [[unlikely]] | |
| 1010 | - return {false, 0, 0, '\0'}; | |
| 1011 | - } | |
| 1012 | - // Gather digits. NB No risk of overflow as 99'999 < max int. | |
| 1013 | - while (*p == '0') { | |
| 1014 | - ++f2_len; | |
| 1015 | - ++p; | |
| 1016 | - } | |
| 1017 | - while (QUtil::is_digit(*p) && f2_len++ < 5) { | |
| 1018 | - f2 *= 10; | |
| 1019 | - f2 += static_cast<int>(*p++ - '0'); | |
| 1020 | - } | |
| 1021 | - if (QUtil::is_space(*p++) && (*p == 'f' || *p == 'n')) { | |
| 1022 | - // C++20: [[likely]] | |
| 1023 | - type = *p; | |
| 1024 | - // No test for valid line[19]. | |
| 1025 | - if (*(++p) && *(++p) && (*p == '\n' || *p == '\r') && f1_len == 10 && f2_len == 5) { | |
| 1026 | - // C++20: [[likely]] | |
| 1027 | - return {true, f1, f2, type}; | |
| 1028 | - } | |
| 1029 | - } | |
| 1030 | - return read_bad_entry(); | |
| 1031 | -} | |
| 1032 | - | |
| 1033 | -// Read a single cross-reference table section and associated trailer. | |
| 1034 | -qpdf_offset_t | |
| 1035 | -QPDF::Xref_table::process_section(qpdf_offset_t xref_offset) | |
| 1036 | -{ | |
| 1037 | - file->seek(xref_offset, SEEK_SET); | |
| 1038 | - std::string line; | |
| 1039 | - auto subs = subsections(line); | |
| 1040 | - | |
| 1041 | - auto cur_trailer_offset = file->tell(); | |
| 1042 | - auto cur_trailer = read_trailer(); | |
| 1043 | - if (!cur_trailer.isDictionary()) { | |
| 1044 | - QTC::TC("qpdf", "QPDF missing trailer"); | |
| 1045 | - throw qpdf.damagedPDF("", "expected trailer dictionary"); | |
| 1046 | - } | |
| 1047 | - | |
| 1048 | - if (!trailer_) { | |
| 1049 | - unsigned int sz; | |
| 1050 | - trailer_ = cur_trailer; | |
| 1051 | - | |
| 1052 | - if (!trailer_.hasKey("/Size")) { | |
| 1053 | - QTC::TC("qpdf", "QPDF trailer lacks size"); | |
| 1054 | - throw qpdf.damagedPDF("trailer", "trailer dictionary lacks /Size key"); | |
| 1055 | - } | |
| 1056 | - if (!trailer_.getKey("/Size").getValueAsUInt(sz)) { | |
| 1057 | - QTC::TC("qpdf", "QPDF trailer size not integer"); | |
| 1058 | - throw qpdf.damagedPDF("trailer", "/Size key in trailer dictionary is not an integer"); | |
| 1059 | - } | |
| 1060 | - if (sz >= static_cast<unsigned int>(max_id_)) { | |
| 1061 | - QTC::TC("qpdf", "QPDF trailer size impossibly large"); | |
| 1062 | - throw qpdf.damagedPDF("trailer", "/Size key in trailer dictionary is impossibly large"); | |
| 1063 | - } | |
| 1064 | - table.resize(sz); | |
| 1065 | - } | |
| 1066 | - | |
| 1067 | - for (auto [obj, num, offset]: subs) { | |
| 1068 | - file->seek(offset, SEEK_SET); | |
| 1069 | - for (qpdf_offset_t i = obj; i - num < obj; ++i) { | |
| 1070 | - if (i == 0) { | |
| 1071 | - // This is needed by checkLinearization() | |
| 1072 | - first_item_offset_ = file->tell(); | |
| 1073 | - } | |
| 1074 | - // For xref_table, these will always be small enough to be ints | |
| 1075 | - auto [success, f1, f2, type] = read_entry(); | |
| 1076 | - if (!success) { | |
| 1077 | - throw damaged_table("invalid xref entry (obj=" + std::to_string(i) + ")"); | |
| 1078 | - } | |
| 1079 | - if (type == 'f') { | |
| 1080 | - insert_free(QPDFObjGen(toI(i), f2)); | |
| 1081 | - } else { | |
| 1082 | - insert(toI(i), 1, f1, f2); | |
| 1083 | - } | |
| 1084 | - } | |
| 1085 | - qpdf_offset_t pos = file->tell(); | |
| 1086 | - if (read_token().isWord("trailer")) { | |
| 1087 | - break; | |
| 1088 | - } else { | |
| 1089 | - file->seek(pos, SEEK_SET); | |
| 1090 | - } | |
| 1091 | - } | |
| 1092 | - | |
| 1093 | - if (cur_trailer.hasKey("/XRefStm")) { | |
| 1094 | - if (ignore_streams_) { | |
| 1095 | - QTC::TC("qpdf", "QPDF ignoring XRefStm in trailer"); | |
| 1096 | - } else { | |
| 1097 | - if (cur_trailer.getKey("/XRefStm").isInteger()) { | |
| 1098 | - // Read the xref stream but disregard any return value -- we'll use our trailer's | |
| 1099 | - // /Prev key instead of the xref stream's. | |
| 1100 | - (void)read_stream(cur_trailer.getKey("/XRefStm").getIntValue()); | |
| 1101 | - } else { | |
| 1102 | - throw qpdf.damagedPDF("xref stream", cur_trailer_offset, "invalid /XRefStm"); | |
| 1103 | - } | |
| 1104 | - } | |
| 1105 | - } | |
| 1106 | - | |
| 1107 | - if (cur_trailer.hasKey("/Prev")) { | |
| 1108 | - if (!cur_trailer.getKey("/Prev").isInteger()) { | |
| 1109 | - QTC::TC("qpdf", "QPDF trailer prev not integer"); | |
| 1110 | - throw qpdf.damagedPDF( | |
| 1111 | - "trailer", cur_trailer_offset, "/Prev key in trailer dictionary is not an integer"); | |
| 1112 | - } | |
| 1113 | - QTC::TC("qpdf", "QPDF prev key in trailer dictionary"); | |
| 1114 | - return cur_trailer.getKey("/Prev").getIntValue(); | |
| 1115 | - } | |
| 1116 | - | |
| 1117 | - return 0; | |
| 1118 | -} | |
| 1119 | - | |
| 1120 | -// Read a single cross-reference stream. | |
| 1121 | -qpdf_offset_t | |
| 1122 | -QPDF::Xref_table::read_stream(qpdf_offset_t xref_offset) | |
| 1123 | -{ | |
| 1124 | - if (!ignore_streams_) { | |
| 1125 | - QPDFObjGen x_og; | |
| 1126 | - QPDFObjectHandle xref_obj; | |
| 1127 | - try { | |
| 1128 | - xref_obj = qpdf.readObjectAtOffset( | |
| 1129 | - false, xref_offset, "xref stream", QPDFObjGen(0, 0), x_og, true); | |
| 1130 | - } catch (QPDFExc&) { | |
| 1131 | - // ignore -- report error below | |
| 1132 | - } | |
| 1133 | - if (xref_obj.isStreamOfType("/XRef")) { | |
| 1134 | - QTC::TC("qpdf", "QPDF found xref stream"); | |
| 1135 | - return process_stream(xref_offset, xref_obj); | |
| 1136 | - } | |
| 1137 | - } | |
| 1138 | - | |
| 1139 | - QTC::TC("qpdf", "QPDF can't find xref"); | |
| 1140 | - throw qpdf.damagedPDF("", xref_offset, "xref not found"); | |
| 1141 | - return 0; // unreachable | |
| 1142 | -} | |
| 1143 | - | |
| 1144 | -// Return the entry size of the xref stream and the processed W array. | |
| 1145 | -std::pair<int, std::array<int, 3>> | |
| 1146 | -QPDF::Xref_table::process_W( | |
| 1147 | - QPDFObjectHandle& dict, std::function<QPDFExc(std::string_view)> damaged) | |
| 1148 | -{ | |
| 1149 | - auto W_obj = dict.getKey("/W"); | |
| 1150 | - if (!(W_obj.isArray() && W_obj.getArrayNItems() >= 3 && W_obj.getArrayItem(0).isInteger() && | |
| 1151 | - W_obj.getArrayItem(1).isInteger() && W_obj.getArrayItem(2).isInteger())) { | |
| 1152 | - throw damaged("Cross-reference stream does not have a proper /W key"); | |
| 1153 | - } | |
| 1154 | - | |
| 1155 | - std::array<int, 3> W; | |
| 1156 | - int entry_size = 0; | |
| 1157 | - auto w_vector = W_obj.getArrayAsVector(); | |
| 1158 | - int max_bytes = sizeof(qpdf_offset_t); | |
| 1159 | - for (size_t i = 0; i < 3; ++i) { | |
| 1160 | - W[i] = w_vector[i].getIntValueAsInt(); | |
| 1161 | - if (W[i] > max_bytes) { | |
| 1162 | - throw damaged("Cross-reference stream's /W contains impossibly large values"); | |
| 1163 | - } | |
| 1164 | - if (W[i] < 0) { | |
| 1165 | - throw damaged("Cross-reference stream's /W contains negative values"); | |
| 1166 | - } | |
| 1167 | - entry_size += W[i]; | |
| 1168 | - } | |
| 1169 | - if (entry_size == 0) { | |
| 1170 | - throw damaged("Cross-reference stream's /W indicates entry size of 0"); | |
| 1171 | - } | |
| 1172 | - return {entry_size, W}; | |
| 1173 | -} | |
| 1174 | - | |
| 1175 | -// Validate Size entry and return the maximum number of entries that the xref stream can contain and | |
| 1176 | -// the value of the Size entry. | |
| 1177 | -std::pair<int, size_t> | |
| 1178 | -QPDF::Xref_table::process_Size( | |
| 1179 | - QPDFObjectHandle& dict, int entry_size, std::function<QPDFExc(std::string_view)> damaged) | |
| 1180 | -{ | |
| 1181 | - // Number of entries is limited by the highest possible object id and stream size. | |
| 1182 | - auto max_num_entries = std::numeric_limits<int>::max(); | |
| 1183 | - if (max_num_entries > (std::numeric_limits<qpdf_offset_t>::max() / entry_size)) { | |
| 1184 | - max_num_entries = toI(std::numeric_limits<qpdf_offset_t>::max() / entry_size); | |
| 1185 | - } | |
| 1186 | - | |
| 1187 | - auto Size_obj = dict.getKey("/Size"); | |
| 1188 | - long long size; | |
| 1189 | - if (!dict.getKey("/Size").getValueAsInt(size)) { | |
| 1190 | - throw damaged("Cross-reference stream does not have a proper /Size key"); | |
| 1191 | - } else if (size < 0) { | |
| 1192 | - throw damaged("Cross-reference stream has a negative /Size key"); | |
| 1193 | - } else if (size >= max_num_entries) { | |
| 1194 | - throw damaged("Cross-reference stream has an impossibly large /Size key"); | |
| 1195 | - } | |
| 1196 | - // We are not validating that Size <= (Size key of parent xref / trailer). | |
| 1197 | - return {max_num_entries, toS(size)}; | |
| 1198 | -} | |
| 1199 | - | |
| 1200 | -// Return the number of entries of the xref stream and the processed Index array. | |
| 1201 | -std::pair<int, std::vector<std::pair<int, int>>> | |
| 1202 | -QPDF::Xref_table::process_Index( | |
| 1203 | - QPDFObjectHandle& dict, int max_num_entries, std::function<QPDFExc(std::string_view)> damaged) | |
| 1204 | -{ | |
| 1205 | - auto size = dict.getKey("/Size").getIntValueAsInt(); | |
| 1206 | - auto Index_obj = dict.getKey("/Index"); | |
| 1207 | - | |
| 1208 | - if (Index_obj.isArray()) { | |
| 1209 | - std::vector<std::pair<int, int>> indx; | |
| 1210 | - int num_entries = 0; | |
| 1211 | - auto index_vec = Index_obj.getArrayAsVector(); | |
| 1212 | - if ((index_vec.size() % 2) || index_vec.size() < 2) { | |
| 1213 | - throw damaged("Cross-reference stream's /Index has an invalid number of values"); | |
| 1214 | - } | |
| 1215 | - | |
| 1216 | - int i = 0; | |
| 1217 | - long long first = 0; | |
| 1218 | - for (auto& val: index_vec) { | |
| 1219 | - if (val.isInteger()) { | |
| 1220 | - if (i % 2) { | |
| 1221 | - auto count = val.getIntValue(); | |
| 1222 | - if (count <= 0) { | |
| 1223 | - throw damaged( | |
| 1224 | - "Cross-reference stream section claims to contain " + | |
| 1225 | - std::to_string(count) + " entries"); | |
| 1226 | - } | |
| 1227 | - // We are guarding against the possibility of num_entries * entry_size | |
| 1228 | - // overflowing. We are not checking that entries are in ascending order as | |
| 1229 | - // required by the spec, which probably should generate a warning. We are also | |
| 1230 | - // not checking that for each subsection first object number + number of entries | |
| 1231 | - // <= /Size. The spec requires us to ignore object number > /Size. | |
| 1232 | - if (first > (max_num_entries - count) || | |
| 1233 | - count > (max_num_entries - num_entries)) { | |
| 1234 | - throw damaged( | |
| 1235 | - "Cross-reference stream claims to contain too many entries: " + | |
| 1236 | - std::to_string(first) + " " + std::to_string(max_num_entries) + " " + | |
| 1237 | - std::to_string(num_entries)); | |
| 1238 | - } | |
| 1239 | - indx.emplace_back(static_cast<int>(first), static_cast<int>(count)); | |
| 1240 | - num_entries += static_cast<int>(count); | |
| 1241 | - } else { | |
| 1242 | - first = val.getIntValue(); | |
| 1243 | - if (first < 0) { | |
| 1244 | - throw damaged( | |
| 1245 | - "Cross-reference stream's /Index contains a negative object id"); | |
| 1246 | - } else if (first > max_num_entries) { | |
| 1247 | - throw damaged("Cross-reference stream's /Index contains an impossibly " | |
| 1248 | - "large object id"); | |
| 1249 | - } | |
| 1250 | - } | |
| 1251 | - } else { | |
| 1252 | - throw damaged( | |
| 1253 | - "Cross-reference stream's /Index's item " + std::to_string(i) + | |
| 1254 | - " is not an integer"); | |
| 1255 | - } | |
| 1256 | - i++; | |
| 1257 | - } | |
| 1258 | - QTC::TC("qpdf", "QPDF xref /Index is array", index_vec.size() == 2 ? 0 : 1); | |
| 1259 | - return {num_entries, indx}; | |
| 1260 | - } else if (Index_obj.isNull()) { | |
| 1261 | - QTC::TC("qpdf", "QPDF xref /Index is null"); | |
| 1262 | - return {size, {{0, size}}}; | |
| 1263 | - } else { | |
| 1264 | - throw damaged("Cross-reference stream does not have a proper /Index key"); | |
| 1265 | - } | |
| 1266 | -} | |
| 1267 | - | |
| 1268 | -qpdf_offset_t | |
| 1269 | -QPDF::Xref_table::process_stream(qpdf_offset_t xref_offset, QPDFObjectHandle& xref_obj) | |
| 1270 | -{ | |
| 1271 | - auto damaged = [this, xref_offset](std::string_view msg) -> QPDFExc { | |
| 1272 | - return qpdf.damagedPDF("xref stream", xref_offset, msg.data()); | |
| 1273 | - }; | |
| 1274 | - | |
| 1275 | - auto dict = xref_obj.getDict(); | |
| 1276 | - | |
| 1277 | - auto [entry_size, W] = process_W(dict, damaged); | |
| 1278 | - auto [max_num_entries, size] = process_Size(dict, entry_size, damaged); | |
| 1279 | - auto [num_entries, indx] = process_Index(dict, max_num_entries, damaged); | |
| 1280 | - | |
| 1281 | - std::shared_ptr<Buffer> bp = xref_obj.getStreamData(qpdf_dl_specialized); | |
| 1282 | - size_t actual_size = bp->getSize(); | |
| 1283 | - auto expected_size = toS(entry_size) * toS(num_entries); | |
| 1284 | - | |
| 1285 | - if (expected_size != actual_size) { | |
| 1286 | - QPDFExc x = damaged( | |
| 1287 | - "Cross-reference stream data has the wrong size; expected = " + | |
| 1288 | - std::to_string(expected_size) + "; actual = " + std::to_string(actual_size)); | |
| 1289 | - if (expected_size > actual_size) { | |
| 1290 | - throw x; | |
| 1291 | - } else { | |
| 1292 | - qpdf.warn(x); | |
| 1293 | - } | |
| 1294 | - } | |
| 1295 | - | |
| 1296 | - if (!trailer_) { | |
| 1297 | - trailer_ = dict; | |
| 1298 | - if (size > toS(max_id_)) { | |
| 1299 | - throw damaged("Cross-reference stream /Size entry is impossibly large"); | |
| 1300 | - } | |
| 1301 | - table.resize(size); | |
| 1302 | - } | |
| 1303 | - | |
| 1304 | - bool saw_first_compressed_object = false; | |
| 1305 | - | |
| 1306 | - // Actual size vs. expected size check above ensures that we will not overflow any buffers here. | |
| 1307 | - // We know that entry_size * num_entries is less or equal to the size of the buffer. | |
| 1308 | - auto p = bp->getBuffer(); | |
| 1309 | - for (auto [obj, sec_entries]: indx) { | |
| 1310 | - // Process a subsection. | |
| 1311 | - for (int i = 0; i < sec_entries; ++i) { | |
| 1312 | - // Read this entry | |
| 1313 | - std::array<qpdf_offset_t, 3> fields{}; | |
| 1314 | - if (W[0] == 0) { | |
| 1315 | - QTC::TC("qpdf", "QPDF default for xref stream field 0"); | |
| 1316 | - fields[0] = 1; | |
| 1317 | - } | |
| 1318 | - for (size_t j = 0; j < 3; ++j) { | |
| 1319 | - for (int k = 0; k < W[j]; ++k) { | |
| 1320 | - fields[j] <<= 8; | |
| 1321 | - fields[j] |= *p++; | |
| 1322 | - } | |
| 1323 | - } | |
| 1324 | - | |
| 1325 | - // Get the generation number. The generation number is 0 unless this is an uncompressed | |
| 1326 | - // object record, in which case the generation number appears as the third field. | |
| 1327 | - if (saw_first_compressed_object) { | |
| 1328 | - if (fields[0] != 2) { | |
| 1329 | - uncompressed_after_compressed_ = true; | |
| 1330 | - } | |
| 1331 | - } else if (fields[0] == 2) { | |
| 1332 | - saw_first_compressed_object = true; | |
| 1333 | - } | |
| 1334 | - if (obj == 0) { | |
| 1335 | - // This is needed by checkLinearization() | |
| 1336 | - first_item_offset_ = xref_offset; | |
| 1337 | - } else if (fields[0] == 0) { | |
| 1338 | - // Ignore fields[2], which we don't care about in this case. This works around the | |
| 1339 | - // issue of some PDF files that put invalid values, like -1, here for deleted | |
| 1340 | - // objects. | |
| 1341 | - insert_free(QPDFObjGen(obj, 0)); | |
| 1342 | - } else { | |
| 1343 | - insert(obj, toI(fields[0]), fields[1], toI(fields[2])); | |
| 1344 | - } | |
| 1345 | - ++obj; | |
| 1346 | - } | |
| 1347 | - } | |
| 1348 | - | |
| 1349 | - if (dict.hasKey("/Prev")) { | |
| 1350 | - if (!dict.getKey("/Prev").isInteger()) { | |
| 1351 | - throw qpdf.damagedPDF( | |
| 1352 | - "xref stream", "/Prev key in xref stream dictionary is not an integer"); | |
| 1353 | - } | |
| 1354 | - QTC::TC("qpdf", "QPDF prev key in xref stream dictionary"); | |
| 1355 | - return dict.getKey("/Prev").getIntValue(); | |
| 1356 | - } else { | |
| 1357 | - return 0; | |
| 1358 | - } | |
| 1359 | -} | |
| 1360 | - | |
| 1361 | -void | |
| 1362 | -QPDF::Xref_table::insert(int obj, int f0, qpdf_offset_t f1, int f2) | |
| 1363 | -{ | |
| 1364 | - // Populate the xref table in such a way that the first reference to an object that we see, | |
| 1365 | - // which is the one in the latest xref table in which it appears, is the one that gets stored. | |
| 1366 | - // This works because we are reading more recent appends before older ones. | |
| 1367 | - | |
| 1368 | - // If there is already an entry for this object and generation in the table, it means that a | |
| 1369 | - // later xref table has registered this object. Disregard this one. | |
| 1370 | - | |
| 1371 | - int new_gen = f0 == 2 ? 0 : f2; | |
| 1372 | - | |
| 1373 | - if (!(obj > 0 && static_cast<size_t>(obj) < table.size() && 0 <= f2 && new_gen < 65535)) { | |
| 1374 | - // We are ignoring invalid objgens. Most will arrive here from xref reconstruction. There | |
| 1375 | - // is probably no point having another warning but we could count invalid items in order to | |
| 1376 | - // decide when to give up. | |
| 1377 | - QTC::TC("qpdf", "QPDF xref overwrite invalid objgen"); | |
| 1378 | - return; | |
| 1379 | - } | |
| 1380 | - | |
| 1381 | - auto& entry = table[static_cast<size_t>(obj)]; | |
| 1382 | - auto old_type = entry.type(); | |
| 1383 | - | |
| 1384 | - if (!old_type && entry.gen() > 0) { | |
| 1385 | - // At the moment we are processing the updates last to first and therefore the gen doesn't | |
| 1386 | - // matter as long as it > 0 to distinguish it from an uninitialized entry. This will need | |
| 1387 | - // to be revisited when we want to support incremental updates or more comprhensive | |
| 1388 | - // checking. | |
| 1389 | - QTC::TC("qpdf", "QPDF xref deleted object"); | |
| 1390 | - return; | |
| 1391 | - } | |
| 1392 | - | |
| 1393 | - if (f0 == 2 && static_cast<int>(f1) == obj) { | |
| 1394 | - qpdf.warn(qpdf.damagedPDF( | |
| 1395 | - "xref stream", "self-referential object stream " + std::to_string(obj))); | |
| 1396 | - return; | |
| 1397 | - } | |
| 1398 | - | |
| 1399 | - if (old_type && entry.gen() >= new_gen) { | |
| 1400 | - QTC::TC("qpdf", "QPDF xref reused object"); | |
| 1401 | - return; | |
| 1402 | - } | |
| 1403 | - | |
| 1404 | - switch (f0) { | |
| 1405 | - case 1: | |
| 1406 | - // f2 is generation | |
| 1407 | - QTC::TC("qpdf", "QPDF xref gen > 0", (f2 > 0) ? 1 : 0); | |
| 1408 | - entry = {f2, Uncompressed(f1)}; | |
| 1409 | - break; | |
| 1410 | - | |
| 1411 | - case 2: | |
| 1412 | - entry = {0, Compressed(toI(f1), f2)}; | |
| 1413 | - object_streams_ = true; | |
| 1414 | - break; | |
| 1415 | - | |
| 1416 | - default: | |
| 1417 | - throw qpdf.damagedPDF( | |
| 1418 | - "xref stream", "unknown xref stream entry type " + std::to_string(f0)); | |
| 1419 | - break; | |
| 1420 | - } | |
| 1421 | -} | |
| 1422 | - | |
| 1423 | -void | |
| 1424 | -QPDF::Xref_table::insert_free(QPDFObjGen og) | |
| 1425 | -{ | |
| 1426 | - // At the moment we are processing the updates last to first and therefore the gen doesn't | |
| 1427 | - // matter as long as it > 0 to distinguish it from an uninitialized entry. This will need to be | |
| 1428 | - // revisited when we want to support incremental updates or more comprhensive checking. | |
| 1429 | - if (og.getObj() < 1) { | |
| 1430 | - return; | |
| 1431 | - } | |
| 1432 | - size_t id = static_cast<size_t>(og.getObj()); | |
| 1433 | - if (id < table.size() && !type(id)) { | |
| 1434 | - table[id] = {1, {}}; | |
| 1435 | - } | |
| 1436 | -} | |
| 1437 | - | |
| 1438 | -QPDFObjGen | |
| 1439 | -QPDF::Xref_table::at_offset(qpdf_offset_t offset) const noexcept | |
| 1440 | -{ | |
| 1441 | - int id = 0; | |
| 1442 | - int gen = 0; | |
| 1443 | - qpdf_offset_t start = 0; | |
| 1444 | - | |
| 1445 | - int i = 0; | |
| 1446 | - for (auto const& item: table) { | |
| 1447 | - auto o = item.offset(); | |
| 1448 | - if (start < o && o <= offset) { | |
| 1449 | - start = o; | |
| 1450 | - id = i; | |
| 1451 | - gen = item.gen(); | |
| 1452 | - } | |
| 1453 | - ++i; | |
| 454 | + *m->log->getWarn() << "WARNING: " << m->warnings.back().what() << "\n"; | |
| 1454 | 455 | } |
| 1455 | - return QPDFObjGen(id, gen); | |
| 1456 | 456 | } |
| 1457 | 457 | |
| 1458 | -std::map<QPDFObjGen, QPDFXRefEntry> | |
| 1459 | -QPDF::Xref_table::as_map() const | |
| 1460 | -{ | |
| 1461 | - std::map<QPDFObjGen, QPDFXRefEntry> result; | |
| 1462 | - int i{0}; | |
| 1463 | - for (auto const& item: table) { | |
| 1464 | - switch (item.type()) { | |
| 1465 | - case 0: | |
| 1466 | - break; | |
| 1467 | - case 1: | |
| 1468 | - result.emplace(QPDFObjGen(i, item.gen()), item.offset()); | |
| 1469 | - break; | |
| 1470 | - case 2: | |
| 1471 | - result.emplace( | |
| 1472 | - QPDFObjGen(i, 0), QPDFXRefEntry(item.stream_number(), item.stream_index())); | |
| 1473 | - break; | |
| 1474 | - default: | |
| 1475 | - throw std::logic_error("Xref_table: invalid entry type"); | |
| 1476 | - } | |
| 1477 | - ++i; | |
| 1478 | - } | |
| 1479 | - return result; | |
| 458 | +void | |
| 459 | +QPDF::warn( | |
| 460 | + qpdf_error_code_e error_code, | |
| 461 | + std::string const& object, | |
| 462 | + qpdf_offset_t offset, | |
| 463 | + std::string const& message) | |
| 464 | +{ | |
| 465 | + warn(QPDFExc(error_code, getFilename(), object, offset, message)); | |
| 1480 | 466 | } |
| 1481 | 467 | |
| 1482 | 468 | void |
| ... | ... | @@ -1485,54 +471,6 @@ QPDF::showXRefTable() |
| 1485 | 471 | m->xref_table.show(); |
| 1486 | 472 | } |
| 1487 | 473 | |
| 1488 | -void | |
| 1489 | -QPDF::Xref_table::show() | |
| 1490 | -{ | |
| 1491 | - auto& cout = *qpdf.m->log->getInfo(); | |
| 1492 | - int i = -1; | |
| 1493 | - for (auto const& item: table) { | |
| 1494 | - ++i; | |
| 1495 | - if (item.type()) { | |
| 1496 | - cout << std::to_string(i) << "/" << std::to_string(item.gen()) << ": "; | |
| 1497 | - switch (item.type()) { | |
| 1498 | - case 1: | |
| 1499 | - cout << "uncompressed; offset = " << item.offset() << "\n"; | |
| 1500 | - break; | |
| 1501 | - | |
| 1502 | - case 2: | |
| 1503 | - cout << "compressed; stream = " << item.stream_number() | |
| 1504 | - << ", index = " << item.stream_index() << "\n"; | |
| 1505 | - break; | |
| 1506 | - | |
| 1507 | - default: | |
| 1508 | - throw std::logic_error( | |
| 1509 | - "unknown cross-reference table type while showing xref_table"); | |
| 1510 | - } | |
| 1511 | - } | |
| 1512 | - } | |
| 1513 | -} | |
| 1514 | - | |
| 1515 | -// Resolve all objects in the xref table. If this triggers a xref table reconstruction abort and | |
| 1516 | -// return false. Otherwise return true. | |
| 1517 | -bool | |
| 1518 | -QPDF::Xref_table::resolve() | |
| 1519 | -{ | |
| 1520 | - bool may_change = !reconstructed_; | |
| 1521 | - int i = -1; | |
| 1522 | - for (auto& item: table) { | |
| 1523 | - ++i; | |
| 1524 | - if (item.type()) { | |
| 1525 | - if (qpdf.isUnresolved(QPDFObjGen(i, item.gen()))) { | |
| 1526 | - qpdf.resolve(QPDFObjGen(i, item.gen())); | |
| 1527 | - if (may_change && reconstructed_) { | |
| 1528 | - return false; | |
| 1529 | - } | |
| 1530 | - } | |
| 1531 | - } | |
| 1532 | - } | |
| 1533 | - return true; | |
| 1534 | -} | |
| 1535 | - | |
| 1536 | 474 | // Ensure all objects in the pdf file, including those in indirect references, appear in the object |
| 1537 | 475 | // cache. |
| 1538 | 476 | void |
| ... | ... | @@ -1562,18 +500,6 @@ QPDF::getObjectCount() |
| 1562 | 500 | return toS(og.getObj()); |
| 1563 | 501 | } |
| 1564 | 502 | |
| 1565 | -std::vector<QPDFObjectHandle> | |
| 1566 | -QPDF::getAllObjects() | |
| 1567 | -{ | |
| 1568 | - // After fixDanglingReferences is called, all objects are in the object cache. | |
| 1569 | - fixDanglingReferences(); | |
| 1570 | - std::vector<QPDFObjectHandle> result; | |
| 1571 | - for (auto const& iter: m->obj_cache) { | |
| 1572 | - result.push_back(newIndirect(iter.first, iter.second.object)); | |
| 1573 | - } | |
| 1574 | - return result; | |
| 1575 | -} | |
| 1576 | - | |
| 1577 | 503 | void |
| 1578 | 504 | QPDF::setLastObjectDescription(std::string const& description, QPDFObjGen const& og) |
| 1579 | 505 | { |
| ... | ... | @@ -1589,220 +515,6 @@ QPDF::setLastObjectDescription(std::string const& description, QPDFObjGen const& |
| 1589 | 515 | } |
| 1590 | 516 | } |
| 1591 | 517 | |
| 1592 | -QPDFObjectHandle | |
| 1593 | -QPDF::Xref_table::read_trailer() | |
| 1594 | -{ | |
| 1595 | - qpdf_offset_t offset = file->tell(); | |
| 1596 | - bool empty = false; | |
| 1597 | - auto object = QPDFParser(*file, "trailer", tokenizer, nullptr, &qpdf, true).parse(empty, false); | |
| 1598 | - if (empty) { | |
| 1599 | - // Nothing in the PDF spec appears to allow empty objects, but they have been encountered in | |
| 1600 | - // actual PDF files and Adobe Reader appears to ignore them. | |
| 1601 | - qpdf.warn(qpdf.damagedPDF("trailer", "empty object treated as null")); | |
| 1602 | - } else if (object.isDictionary() && read_token().isWord("stream")) { | |
| 1603 | - qpdf.warn(qpdf.damagedPDF("trailer", file->tell(), "stream keyword found in trailer")); | |
| 1604 | - } | |
| 1605 | - // Override last_offset so that it points to the beginning of the object we just read | |
| 1606 | - file->setLastOffset(offset); | |
| 1607 | - return object; | |
| 1608 | -} | |
| 1609 | - | |
| 1610 | -QPDFObjectHandle | |
| 1611 | -QPDF::readObject(std::string const& description, QPDFObjGen og) | |
| 1612 | -{ | |
| 1613 | - setLastObjectDescription(description, og); | |
| 1614 | - qpdf_offset_t offset = m->file->tell(); | |
| 1615 | - bool empty = false; | |
| 1616 | - | |
| 1617 | - StringDecrypter decrypter{this, og}; | |
| 1618 | - StringDecrypter* decrypter_ptr = m->encp->encrypted ? &decrypter : nullptr; | |
| 1619 | - auto object = | |
| 1620 | - QPDFParser(*m->file, m->last_object_description, m->tokenizer, decrypter_ptr, this, true) | |
| 1621 | - .parse(empty, false); | |
| 1622 | - if (empty) { | |
| 1623 | - // Nothing in the PDF spec appears to allow empty objects, but they have been encountered in | |
| 1624 | - // actual PDF files and Adobe Reader appears to ignore them. | |
| 1625 | - warn(damagedPDF(*m->file, m->file->getLastOffset(), "empty object treated as null")); | |
| 1626 | - return object; | |
| 1627 | - } | |
| 1628 | - auto token = readToken(*m->file); | |
| 1629 | - if (object.isDictionary() && token.isWord("stream")) { | |
| 1630 | - readStream(object, og, offset); | |
| 1631 | - token = readToken(*m->file); | |
| 1632 | - } | |
| 1633 | - if (!token.isWord("endobj")) { | |
| 1634 | - QTC::TC("qpdf", "QPDF err expected endobj"); | |
| 1635 | - warn(damagedPDF("expected endobj")); | |
| 1636 | - } | |
| 1637 | - return object; | |
| 1638 | -} | |
| 1639 | - | |
| 1640 | -// After reading stream dictionary and stream keyword, read rest of stream. | |
| 1641 | -void | |
| 1642 | -QPDF::readStream(QPDFObjectHandle& object, QPDFObjGen og, qpdf_offset_t offset) | |
| 1643 | -{ | |
| 1644 | - validateStreamLineEnd(object, og, offset); | |
| 1645 | - | |
| 1646 | - // Must get offset before accessing any additional objects since resolving a previously | |
| 1647 | - // unresolved indirect object will change file position. | |
| 1648 | - qpdf_offset_t stream_offset = m->file->tell(); | |
| 1649 | - size_t length = 0; | |
| 1650 | - | |
| 1651 | - try { | |
| 1652 | - auto length_obj = object.getKey("/Length"); | |
| 1653 | - | |
| 1654 | - if (!length_obj.isInteger()) { | |
| 1655 | - if (length_obj.isNull()) { | |
| 1656 | - QTC::TC("qpdf", "QPDF stream without length"); | |
| 1657 | - throw damagedPDF(offset, "stream dictionary lacks /Length key"); | |
| 1658 | - } | |
| 1659 | - QTC::TC("qpdf", "QPDF stream length not integer"); | |
| 1660 | - throw damagedPDF(offset, "/Length key in stream dictionary is not an integer"); | |
| 1661 | - } | |
| 1662 | - | |
| 1663 | - length = toS(length_obj.getUIntValue()); | |
| 1664 | - // Seek in two steps to avoid potential integer overflow | |
| 1665 | - m->file->seek(stream_offset, SEEK_SET); | |
| 1666 | - m->file->seek(toO(length), SEEK_CUR); | |
| 1667 | - if (!readToken(*m->file).isWord("endstream")) { | |
| 1668 | - QTC::TC("qpdf", "QPDF missing endstream"); | |
| 1669 | - throw damagedPDF("expected endstream"); | |
| 1670 | - } | |
| 1671 | - } catch (QPDFExc& e) { | |
| 1672 | - if (m->attempt_recovery) { | |
| 1673 | - warn(e); | |
| 1674 | - length = recoverStreamLength(m->file_sp, og, stream_offset); | |
| 1675 | - } else { | |
| 1676 | - throw; | |
| 1677 | - } | |
| 1678 | - } | |
| 1679 | - object = {QPDF_Stream::create(this, og, object, stream_offset, length)}; | |
| 1680 | -} | |
| 1681 | - | |
| 1682 | -void | |
| 1683 | -QPDF::validateStreamLineEnd(QPDFObjectHandle& object, QPDFObjGen og, qpdf_offset_t offset) | |
| 1684 | -{ | |
| 1685 | - // The PDF specification states that the word "stream" should be followed by either a carriage | |
| 1686 | - // return and a newline or by a newline alone. It specifically disallowed following it by a | |
| 1687 | - // carriage return alone since, in that case, there would be no way to tell whether the NL in a | |
| 1688 | - // CR NL sequence was part of the stream data. However, some readers, including Adobe reader, | |
| 1689 | - // accept a carriage return by itself when followed by a non-newline character, so that's what | |
| 1690 | - // we do here. We have also seen files that have extraneous whitespace between the stream | |
| 1691 | - // keyword and the newline. | |
| 1692 | - while (true) { | |
| 1693 | - char ch; | |
| 1694 | - if (m->file->read(&ch, 1) == 0) { | |
| 1695 | - // A premature EOF here will result in some other problem that will get reported at | |
| 1696 | - // another time. | |
| 1697 | - return; | |
| 1698 | - } | |
| 1699 | - if (ch == '\n') { | |
| 1700 | - // ready to read stream data | |
| 1701 | - QTC::TC("qpdf", "QPDF stream with NL only"); | |
| 1702 | - return; | |
| 1703 | - } | |
| 1704 | - if (ch == '\r') { | |
| 1705 | - // Read another character | |
| 1706 | - if (m->file->read(&ch, 1) != 0) { | |
| 1707 | - if (ch == '\n') { | |
| 1708 | - // Ready to read stream data | |
| 1709 | - QTC::TC("qpdf", "QPDF stream with CRNL"); | |
| 1710 | - } else { | |
| 1711 | - // Treat the \r by itself as the whitespace after endstream and start reading | |
| 1712 | - // stream data in spite of not having seen a newline. | |
| 1713 | - QTC::TC("qpdf", "QPDF stream with CR only"); | |
| 1714 | - m->file->unreadCh(ch); | |
| 1715 | - warn(damagedPDF( | |
| 1716 | - m->file->tell(), "stream keyword followed by carriage return only")); | |
| 1717 | - } | |
| 1718 | - } | |
| 1719 | - return; | |
| 1720 | - } | |
| 1721 | - if (!QUtil::is_space(ch)) { | |
| 1722 | - QTC::TC("qpdf", "QPDF stream without newline"); | |
| 1723 | - m->file->unreadCh(ch); | |
| 1724 | - warn(damagedPDF( | |
| 1725 | - m->file->tell(), "stream keyword not followed by proper line terminator")); | |
| 1726 | - return; | |
| 1727 | - } | |
| 1728 | - warn(damagedPDF(m->file->tell(), "stream keyword followed by extraneous whitespace")); | |
| 1729 | - } | |
| 1730 | -} | |
| 1731 | - | |
| 1732 | -QPDFObjectHandle | |
| 1733 | -QPDF::readObjectInStream(std::shared_ptr<InputSource>& input, int obj) | |
| 1734 | -{ | |
| 1735 | - m->last_object_description.erase(7); // last_object_description starts with "object " | |
| 1736 | - m->last_object_description += std::to_string(obj); | |
| 1737 | - m->last_object_description += " 0"; | |
| 1738 | - | |
| 1739 | - bool empty = false; | |
| 1740 | - auto object = QPDFParser(*input, m->last_object_description, m->tokenizer, nullptr, this, true) | |
| 1741 | - .parse(empty, false); | |
| 1742 | - if (empty) { | |
| 1743 | - // Nothing in the PDF spec appears to allow empty objects, but they have been encountered in | |
| 1744 | - // actual PDF files and Adobe Reader appears to ignore them. | |
| 1745 | - warn(damagedPDF(*input, input->getLastOffset(), "empty object treated as null")); | |
| 1746 | - } | |
| 1747 | - return object; | |
| 1748 | -} | |
| 1749 | - | |
| 1750 | -bool | |
| 1751 | -QPDF::findEndstream() | |
| 1752 | -{ | |
| 1753 | - // Find endstream or endobj. Position the input at that token. | |
| 1754 | - auto t = readToken(*m->file, 20); | |
| 1755 | - if (t.isWord("endobj") || t.isWord("endstream")) { | |
| 1756 | - m->file->seek(m->file->getLastOffset(), SEEK_SET); | |
| 1757 | - return true; | |
| 1758 | - } | |
| 1759 | - return false; | |
| 1760 | -} | |
| 1761 | - | |
| 1762 | -size_t | |
| 1763 | -QPDF::recoverStreamLength( | |
| 1764 | - std::shared_ptr<InputSource> input, QPDFObjGen const& og, qpdf_offset_t stream_offset) | |
| 1765 | -{ | |
| 1766 | - // Try to reconstruct stream length by looking for endstream or endobj | |
| 1767 | - warn(damagedPDF(*input, stream_offset, "attempting to recover stream length")); | |
| 1768 | - | |
| 1769 | - PatternFinder ef(*this, &QPDF::findEndstream); | |
| 1770 | - size_t length = 0; | |
| 1771 | - if (m->file->findFirst("end", stream_offset, 0, ef)) { | |
| 1772 | - length = toS(m->file->tell() - stream_offset); | |
| 1773 | - // Reread endstream but, if it was endobj, don't skip that. | |
| 1774 | - QPDFTokenizer::Token t = readToken(*m->file); | |
| 1775 | - if (t.getValue() == "endobj") { | |
| 1776 | - m->file->seek(m->file->getLastOffset(), SEEK_SET); | |
| 1777 | - } | |
| 1778 | - } | |
| 1779 | - | |
| 1780 | - if (length) { | |
| 1781 | - // Make sure this is inside this object | |
| 1782 | - auto found = m->xref_table.at_offset(stream_offset + toO(length)); | |
| 1783 | - if (found == QPDFObjGen() || found == og) { | |
| 1784 | - // If we are trying to recover an XRef stream the xref table will not contain and | |
| 1785 | - // won't contain any entries, therefore we cannot check the found length. Otherwise we | |
| 1786 | - // found endstream\nendobj within the space allowed for this object, so we're probably | |
| 1787 | - // in good shape. | |
| 1788 | - } else { | |
| 1789 | - QTC::TC("qpdf", "QPDF found wrong endstream in recovery"); | |
| 1790 | - length = 0; | |
| 1791 | - } | |
| 1792 | - } | |
| 1793 | - | |
| 1794 | - if (length == 0) { | |
| 1795 | - warn(damagedPDF( | |
| 1796 | - *input, stream_offset, "unable to recover stream data; treating stream as empty")); | |
| 1797 | - } else { | |
| 1798 | - warn(damagedPDF( | |
| 1799 | - *input, stream_offset, "recovered stream length: " + std::to_string(length))); | |
| 1800 | - } | |
| 1801 | - | |
| 1802 | - QTC::TC("qpdf", "QPDF recovered stream length"); | |
| 1803 | - return length; | |
| 1804 | -} | |
| 1805 | - | |
| 1806 | 518 | QPDFTokenizer::Token |
| 1807 | 519 | QPDF::readToken(InputSource& input, size_t max_len) |
| 1808 | 520 | { |
| ... | ... | @@ -1810,367 +522,12 @@ QPDF::readToken(InputSource& input, size_t max_len) |
| 1810 | 522 | } |
| 1811 | 523 | |
| 1812 | 524 | QPDFObjectHandle |
| 1813 | -QPDF::readObjectAtOffset( | |
| 1814 | - bool try_recovery, | |
| 1815 | - qpdf_offset_t offset, | |
| 1816 | - std::string const& description, | |
| 1817 | - QPDFObjGen exp_og, | |
| 1818 | - QPDFObjGen& og, | |
| 1819 | - bool skip_cache_if_in_xref) | |
| 1820 | -{ | |
| 1821 | - bool check_og = true; | |
| 1822 | - if (exp_og.getObj() == 0) { | |
| 1823 | - // This method uses an expect object ID of 0 to indicate that we don't know or don't care | |
| 1824 | - // what the actual object ID is at this offset. This is true when we read the xref stream | |
| 1825 | - // and linearization hint streams. In this case, we don't verify the expect object | |
| 1826 | - // ID/generation against what was read from the file. There is also no reason to attempt | |
| 1827 | - // xref recovery if we get a failure in this case since the read attempt was not triggered | |
| 1828 | - // by an xref lookup. | |
| 1829 | - check_og = false; | |
| 1830 | - try_recovery = false; | |
| 1831 | - } | |
| 1832 | - setLastObjectDescription(description, exp_og); | |
| 1833 | - | |
| 1834 | - if (!m->attempt_recovery) { | |
| 1835 | - try_recovery = false; | |
| 1836 | - } | |
| 1837 | - | |
| 1838 | - // Special case: if offset is 0, just return null. Some PDF writers, in particular | |
| 1839 | - // "Mac OS X 10.7.5 Quartz PDFContext", may store deleted objects in the xref table as | |
| 1840 | - // "0000000000 00000 n", which is not correct, but it won't hurt anything for us to ignore | |
| 1841 | - // these. | |
| 1842 | - if (offset == 0) { | |
| 1843 | - QTC::TC("qpdf", "QPDF bogus 0 offset", 0); | |
| 1844 | - warn(damagedPDF(0, "object has offset 0")); | |
| 1845 | - return QPDFObjectHandle::newNull(); | |
| 1846 | - } | |
| 1847 | - | |
| 1848 | - m->file->seek(offset, SEEK_SET); | |
| 1849 | - try { | |
| 1850 | - QPDFTokenizer::Token tobjid = readToken(*m->file); | |
| 1851 | - bool objidok = tobjid.isInteger(); | |
| 1852 | - QTC::TC("qpdf", "QPDF check objid", objidok ? 1 : 0); | |
| 1853 | - if (!objidok) { | |
| 1854 | - QTC::TC("qpdf", "QPDF expected n n obj"); | |
| 1855 | - throw damagedPDF(offset, "expected n n obj"); | |
| 1856 | - } | |
| 1857 | - QPDFTokenizer::Token tgen = readToken(*m->file); | |
| 1858 | - bool genok = tgen.isInteger(); | |
| 1859 | - QTC::TC("qpdf", "QPDF check generation", genok ? 1 : 0); | |
| 1860 | - if (!genok) { | |
| 1861 | - throw damagedPDF(offset, "expected n n obj"); | |
| 1862 | - } | |
| 1863 | - QPDFTokenizer::Token tobj = readToken(*m->file); | |
| 1864 | - | |
| 1865 | - bool objok = tobj.isWord("obj"); | |
| 1866 | - QTC::TC("qpdf", "QPDF check obj", objok ? 1 : 0); | |
| 1867 | - | |
| 1868 | - if (!objok) { | |
| 1869 | - throw damagedPDF(offset, "expected n n obj"); | |
| 1870 | - } | |
| 1871 | - int objid = QUtil::string_to_int(tobjid.getValue().c_str()); | |
| 1872 | - int generation = QUtil::string_to_int(tgen.getValue().c_str()); | |
| 1873 | - og = QPDFObjGen(objid, generation); | |
| 1874 | - if (objid == 0) { | |
| 1875 | - QTC::TC("qpdf", "QPDF object id 0"); | |
| 1876 | - throw damagedPDF(offset, "object with ID 0"); | |
| 1877 | - } | |
| 1878 | - if (check_og && (exp_og != og)) { | |
| 1879 | - QTC::TC("qpdf", "QPDF err wrong objid/generation"); | |
| 1880 | - QPDFExc e = damagedPDF(offset, "expected " + exp_og.unparse(' ') + " obj"); | |
| 1881 | - if (try_recovery) { | |
| 1882 | - // Will be retried below | |
| 1883 | - throw e; | |
| 1884 | - } else { | |
| 1885 | - // We can try reading the object anyway even if the ID doesn't match. | |
| 1886 | - warn(e); | |
| 1887 | - } | |
| 1888 | - } | |
| 1889 | - } catch (QPDFExc& e) { | |
| 1890 | - if (try_recovery) { | |
| 1891 | - // Try again after reconstructing xref table | |
| 1892 | - m->xref_table.reconstruct(e); | |
| 1893 | - if (m->xref_table.type(exp_og) == 1) { | |
| 1894 | - QTC::TC("qpdf", "QPDF recovered in readObjectAtOffset"); | |
| 1895 | - return readObjectAtOffset( | |
| 1896 | - false, m->xref_table.offset(exp_og), description, exp_og, og, false); | |
| 1897 | - } else { | |
| 1898 | - QTC::TC("qpdf", "QPDF object gone after xref reconstruction"); | |
| 1899 | - warn(damagedPDF( | |
| 1900 | - "", | |
| 1901 | - 0, | |
| 1902 | - ("object " + exp_og.unparse(' ') + | |
| 1903 | - " not found in file after regenerating cross reference table"))); | |
| 1904 | - return QPDFObjectHandle::newNull(); | |
| 1905 | - } | |
| 1906 | - } else { | |
| 1907 | - throw; | |
| 1908 | - } | |
| 1909 | - } | |
| 1910 | - | |
| 1911 | - QPDFObjectHandle oh = readObject(description, og); | |
| 1912 | - | |
| 1913 | - if (isUnresolved(og)) { | |
| 1914 | - // Store the object in the cache here so it gets cached whether we first know the offset or | |
| 1915 | - // whether we first know the object ID and generation (in which we case we would get here | |
| 1916 | - // through resolve). | |
| 1917 | - | |
| 1918 | - // Determine the end offset of this object before and after white space. We use these | |
| 1919 | - // numbers to validate linearization hint tables. Offsets and lengths of objects may imply | |
| 1920 | - // the end of an object to be anywhere between these values. | |
| 1921 | - qpdf_offset_t end_before_space = m->file->tell(); | |
| 1922 | - | |
| 1923 | - // skip over spaces | |
| 1924 | - while (true) { | |
| 1925 | - char ch; | |
| 1926 | - if (m->file->read(&ch, 1)) { | |
| 1927 | - if (!isspace(static_cast<unsigned char>(ch))) { | |
| 1928 | - m->file->seek(-1, SEEK_CUR); | |
| 1929 | - break; | |
| 1930 | - } | |
| 1931 | - } else { | |
| 1932 | - throw damagedPDF(m->file->tell(), "EOF after endobj"); | |
| 1933 | - } | |
| 1934 | - } | |
| 1935 | - qpdf_offset_t end_after_space = m->file->tell(); | |
| 1936 | - if (skip_cache_if_in_xref && m->xref_table.type(og)) { | |
| 1937 | - // Ordinarily, an object gets read here when resolved through xref table or stream. In | |
| 1938 | - // the special case of the xref stream and linearization hint tables, the offset comes | |
| 1939 | - // from another source. For the specific case of xref streams, the xref stream is read | |
| 1940 | - // and loaded into the object cache very early in parsing. Ordinarily, when a file is | |
| 1941 | - // updated by appending, items inserted into the xref table in later updates take | |
| 1942 | - // precedence over earlier items. In the special case of reusing the object number | |
| 1943 | - // previously used as the xref stream, we have the following order of events: | |
| 1944 | - // | |
| 1945 | - // * reused object gets loaded into the xref table | |
| 1946 | - // * old object is read here while reading xref streams | |
| 1947 | - // * original xref entry is ignored (since already in xref table) | |
| 1948 | - // | |
| 1949 | - // It is the second step that causes a problem. Even though the xref table is correct in | |
| 1950 | - // this case, the old object is already in the cache and so effectively prevails over | |
| 1951 | - // the reused object. To work around this issue, we have a special case for the xref | |
| 1952 | - // stream (via the skip_cache_if_in_xref): if the object is already in the xref stream, | |
| 1953 | - // don't cache what we read here. | |
| 1954 | - // | |
| 1955 | - // It is likely that the same bug may exist for linearization hint tables, but the | |
| 1956 | - // existing code uses end_before_space and end_after_space from the cache, so fixing | |
| 1957 | - // that would require more significant rework. The chances of a linearization hint | |
| 1958 | - // stream being reused seems smaller because the xref stream is probably the highest | |
| 1959 | - // object in the file and the linearization hint stream would be some random place in | |
| 1960 | - // the middle, so I'm leaving that bug unfixed for now. If the bug were to be fixed, we | |
| 1961 | - // could use !check_og in place of skip_cache_if_in_xref. | |
| 1962 | - QTC::TC("qpdf", "QPDF skipping cache for known unchecked object"); | |
| 1963 | - } else { | |
| 1964 | - m->xref_table.linearization_offsets( | |
| 1965 | - toS(og.getObj()), end_before_space, end_after_space); | |
| 1966 | - updateCache(og, oh.getObj()); | |
| 1967 | - } | |
| 1968 | - } | |
| 1969 | - | |
| 1970 | - return oh; | |
| 1971 | -} | |
| 1972 | - | |
| 1973 | -QPDFObject* | |
| 1974 | -QPDF::resolve(QPDFObjGen og) | |
| 1975 | -{ | |
| 1976 | - if (!isUnresolved(og)) { | |
| 1977 | - return m->obj_cache[og].object.get(); | |
| 1978 | - } | |
| 1979 | - | |
| 1980 | - if (m->resolving.count(og)) { | |
| 1981 | - // This can happen if an object references itself directly or indirectly in some key that | |
| 1982 | - // has to be resolved during object parsing, such as stream length. | |
| 1983 | - QTC::TC("qpdf", "QPDF recursion loop in resolve"); | |
| 1984 | - warn(damagedPDF("", "loop detected resolving object " + og.unparse(' '))); | |
| 1985 | - updateCache(og, QPDF_Null::create()); | |
| 1986 | - return m->obj_cache[og].object.get(); | |
| 1987 | - } | |
| 1988 | - ResolveRecorder rr(this, og); | |
| 1989 | - | |
| 1990 | - try { | |
| 1991 | - switch (m->xref_table.type(og)) { | |
| 1992 | - case 0: | |
| 1993 | - break; | |
| 1994 | - case 1: | |
| 1995 | - { | |
| 1996 | - // Object stored in cache by readObjectAtOffset | |
| 1997 | - QPDFObjGen a_og; | |
| 1998 | - QPDFObjectHandle oh = | |
| 1999 | - readObjectAtOffset(true, m->xref_table.offset(og), "", og, a_og, false); | |
| 2000 | - } | |
| 2001 | - break; | |
| 2002 | - | |
| 2003 | - case 2: | |
| 2004 | - resolveObjectsInStream(m->xref_table.stream_number(og.getObj())); | |
| 2005 | - break; | |
| 2006 | - | |
| 2007 | - default: | |
| 2008 | - throw damagedPDF( | |
| 2009 | - "", 0, ("object " + og.unparse('/') + " has unexpected xref entry type")); | |
| 2010 | - } | |
| 2011 | - } catch (QPDFExc& e) { | |
| 2012 | - warn(e); | |
| 2013 | - } catch (std::exception& e) { | |
| 2014 | - warn(damagedPDF( | |
| 2015 | - "", 0, ("object " + og.unparse('/') + ": error reading object: " + e.what()))); | |
| 2016 | - } | |
| 2017 | - | |
| 2018 | - if (isUnresolved(og)) { | |
| 2019 | - // PDF spec says unknown objects resolve to the null object. | |
| 2020 | - QTC::TC("qpdf", "QPDF resolve failure to null"); | |
| 2021 | - updateCache(og, QPDF_Null::create()); | |
| 2022 | - } | |
| 2023 | - | |
| 2024 | - auto result(m->obj_cache[og].object); | |
| 2025 | - result->setDefaultDescription(this, og); | |
| 2026 | - return result.get(); | |
| 2027 | -} | |
| 2028 | - | |
| 2029 | -void | |
| 2030 | -QPDF::resolveObjectsInStream(int obj_stream_number) | |
| 2031 | -{ | |
| 2032 | - if (m->resolved_object_streams.count(obj_stream_number)) { | |
| 2033 | - return; | |
| 2034 | - } | |
| 2035 | - m->resolved_object_streams.insert(obj_stream_number); | |
| 2036 | - // Force resolution of object stream | |
| 2037 | - QPDFObjectHandle obj_stream = getObjectByID(obj_stream_number, 0); | |
| 2038 | - if (!obj_stream.isStream()) { | |
| 2039 | - throw damagedPDF( | |
| 2040 | - "supposed object stream " + std::to_string(obj_stream_number) + " is not a stream"); | |
| 2041 | - } | |
| 2042 | - | |
| 2043 | - QPDFObjectHandle dict = obj_stream.getDict(); | |
| 2044 | - if (!dict.isDictionaryOfType("/ObjStm")) { | |
| 2045 | - QTC::TC("qpdf", "QPDF ERR object stream with wrong type"); | |
| 2046 | - warn(damagedPDF( | |
| 2047 | - "supposed object stream " + std::to_string(obj_stream_number) + " has wrong type")); | |
| 2048 | - } | |
| 2049 | - | |
| 2050 | - if (!(dict.getKey("/N").isInteger() && dict.getKey("/First").isInteger())) { | |
| 2051 | - throw damagedPDF( | |
| 2052 | - ("object stream " + std::to_string(obj_stream_number) + " has incorrect keys")); | |
| 2053 | - } | |
| 2054 | - | |
| 2055 | - int n = dict.getKey("/N").getIntValueAsInt(); | |
| 2056 | - int first = dict.getKey("/First").getIntValueAsInt(); | |
| 2057 | - | |
| 2058 | - std::map<int, int> offsets; | |
| 2059 | - | |
| 2060 | - std::shared_ptr<Buffer> bp = obj_stream.getStreamData(qpdf_dl_specialized); | |
| 2061 | - auto input = std::shared_ptr<InputSource>( | |
| 2062 | - // line-break | |
| 2063 | - new BufferInputSource( | |
| 2064 | - (m->file->getName() + " object stream " + std::to_string(obj_stream_number)), | |
| 2065 | - bp.get())); | |
| 2066 | - | |
| 2067 | - qpdf_offset_t last_offset = -1; | |
| 2068 | - for (int i = 0; i < n; ++i) { | |
| 2069 | - QPDFTokenizer::Token tnum = readToken(*input); | |
| 2070 | - QPDFTokenizer::Token toffset = readToken(*input); | |
| 2071 | - if (!(tnum.isInteger() && toffset.isInteger())) { | |
| 2072 | - throw damagedPDF( | |
| 2073 | - *input, | |
| 2074 | - m->last_object_description, | |
| 2075 | - input->getLastOffset(), | |
| 2076 | - "expected integer in object stream header"); | |
| 2077 | - } | |
| 2078 | - | |
| 2079 | - int num = QUtil::string_to_int(tnum.getValue().c_str()); | |
| 2080 | - long long offset = QUtil::string_to_int(toffset.getValue().c_str()); | |
| 2081 | - if (num > m->xref_table.max_id()) { | |
| 2082 | - continue; | |
| 2083 | - } | |
| 2084 | - if (num == obj_stream_number) { | |
| 2085 | - QTC::TC("qpdf", "QPDF ignore self-referential object stream"); | |
| 2086 | - warn(damagedPDF( | |
| 2087 | - *input, | |
| 2088 | - m->last_object_description, | |
| 2089 | - input->getLastOffset(), | |
| 2090 | - "object stream claims to contain itself")); | |
| 2091 | - continue; | |
| 2092 | - } | |
| 2093 | - if (offset <= last_offset) { | |
| 2094 | - throw damagedPDF( | |
| 2095 | - *input, | |
| 2096 | - m->last_object_description, | |
| 2097 | - input->getLastOffset(), | |
| 2098 | - "expected offsets in object stream to be increasing"); | |
| 2099 | - } | |
| 2100 | - last_offset = offset; | |
| 2101 | - | |
| 2102 | - offsets[num] = toI(offset + first); | |
| 2103 | - } | |
| 2104 | - | |
| 2105 | - // To avoid having to read the object stream multiple times, store all objects that would be | |
| 2106 | - // found here in the cache. Remember that some objects stored here might have been overridden | |
| 2107 | - // by new objects appended to the file, so it is necessary to recheck the xref table and only | |
| 2108 | - // cache what would actually be resolved here. | |
| 2109 | - m->last_object_description.clear(); | |
| 2110 | - m->last_object_description += "object "; | |
| 2111 | - for (auto const& iter: offsets) { | |
| 2112 | - QPDFObjGen og(iter.first, 0); | |
| 2113 | - if (m->xref_table.type(og) == 2 && | |
| 2114 | - m->xref_table.stream_number(og.getObj()) == obj_stream_number) { | |
| 2115 | - int offset = iter.second; | |
| 2116 | - input->seek(offset, SEEK_SET); | |
| 2117 | - QPDFObjectHandle oh = readObjectInStream(input, iter.first); | |
| 2118 | - updateCache(og, oh.getObj()); | |
| 2119 | - } else { | |
| 2120 | - QTC::TC("qpdf", "QPDF not caching overridden objstm object"); | |
| 2121 | - } | |
| 2122 | - } | |
| 2123 | -} | |
| 2124 | - | |
| 2125 | -QPDFObjectHandle | |
| 2126 | 525 | QPDF::newIndirect(QPDFObjGen const& og, std::shared_ptr<QPDFObject> const& obj) |
| 2127 | 526 | { |
| 2128 | 527 | obj->setDefaultDescription(this, og); |
| 2129 | 528 | return {obj}; |
| 2130 | 529 | } |
| 2131 | 530 | |
| 2132 | -void | |
| 2133 | -QPDF::updateCache(QPDFObjGen const& og, std::shared_ptr<QPDFObject> const& object) | |
| 2134 | -{ | |
| 2135 | - object->setObjGen(this, og); | |
| 2136 | - if (isCached(og)) { | |
| 2137 | - auto& cache = m->obj_cache[og]; | |
| 2138 | - cache.object->assign(object); | |
| 2139 | - } else { | |
| 2140 | - m->obj_cache[og] = ObjCache(object); | |
| 2141 | - } | |
| 2142 | -} | |
| 2143 | - | |
| 2144 | -bool | |
| 2145 | -QPDF::isCached(QPDFObjGen const& og) | |
| 2146 | -{ | |
| 2147 | - return m->obj_cache.count(og) != 0; | |
| 2148 | -} | |
| 2149 | - | |
| 2150 | -bool | |
| 2151 | -QPDF::isUnresolved(QPDFObjGen const& og) | |
| 2152 | -{ | |
| 2153 | - return !isCached(og) || m->obj_cache[og].object->isUnresolved(); | |
| 2154 | -} | |
| 2155 | - | |
| 2156 | -QPDFObjGen | |
| 2157 | -QPDF::nextObjGen() | |
| 2158 | -{ | |
| 2159 | - int max_objid = toI(getObjectCount()); | |
| 2160 | - if (max_objid == std::numeric_limits<int>::max()) { | |
| 2161 | - throw std::range_error("max object id is too high to create new objects"); | |
| 2162 | - } | |
| 2163 | - return QPDFObjGen(max_objid + 1, 0); | |
| 2164 | -} | |
| 2165 | - | |
| 2166 | -QPDFObjectHandle | |
| 2167 | -QPDF::makeIndirectFromQPDFObject(std::shared_ptr<QPDFObject> const& obj) | |
| 2168 | -{ | |
| 2169 | - QPDFObjGen next{nextObjGen()}; | |
| 2170 | - m->obj_cache[next] = ObjCache(obj); | |
| 2171 | - return newIndirect(next, m->obj_cache[next].object); | |
| 2172 | -} | |
| 2173 | - | |
| 2174 | 531 | QPDFObjectHandle |
| 2175 | 532 | QPDF::makeIndirectObject(QPDFObjectHandle oh) |
| 2176 | 533 | { |
| ... | ... | @@ -2215,37 +572,6 @@ QPDF::newStream(std::string const& data) |
| 2215 | 572 | return result; |
| 2216 | 573 | } |
| 2217 | 574 | |
| 2218 | -std::shared_ptr<QPDFObject> | |
| 2219 | -QPDF::getObjectForParser(int id, int gen, bool parse_pdf) | |
| 2220 | -{ | |
| 2221 | - // This method is called by the parser and therefore must not resolve any objects. | |
| 2222 | - auto og = QPDFObjGen(id, gen); | |
| 2223 | - if (auto iter = m->obj_cache.find(og); iter != m->obj_cache.end()) { | |
| 2224 | - return iter->second.object; | |
| 2225 | - } | |
| 2226 | - if (m->xref_table.type(og) || !m->xref_table.initialized()) { | |
| 2227 | - return m->obj_cache.insert({og, QPDF_Unresolved::create(this, og)}).first->second.object; | |
| 2228 | - } | |
| 2229 | - if (parse_pdf) { | |
| 2230 | - return QPDF_Null::create(); | |
| 2231 | - } | |
| 2232 | - return m->obj_cache.insert({og, QPDF_Null::create(this, og)}).first->second.object; | |
| 2233 | -} | |
| 2234 | - | |
| 2235 | -std::shared_ptr<QPDFObject> | |
| 2236 | -QPDF::getObjectForJSON(int id, int gen) | |
| 2237 | -{ | |
| 2238 | - auto og = QPDFObjGen(id, gen); | |
| 2239 | - auto [it, inserted] = m->obj_cache.try_emplace(og); | |
| 2240 | - auto& obj = it->second.object; | |
| 2241 | - if (inserted) { | |
| 2242 | - obj = (m->xref_table.initialized() && !m->xref_table.type(og)) | |
| 2243 | - ? QPDF_Null::create(this, og) | |
| 2244 | - : QPDF_Unresolved::create(this, og); | |
| 2245 | - } | |
| 2246 | - return obj; | |
| 2247 | -} | |
| 2248 | - | |
| 2249 | 575 | QPDFObjectHandle |
| 2250 | 576 | QPDF::getObject(QPDFObjGen const& og) |
| 2251 | 577 | { |
| ... | ... | @@ -2284,27 +610,6 @@ QPDF::replaceObject(int objid, int generation, QPDFObjectHandle oh) |
| 2284 | 610 | } |
| 2285 | 611 | |
| 2286 | 612 | void |
| 2287 | -QPDF::replaceObject(QPDFObjGen const& og, QPDFObjectHandle oh) | |
| 2288 | -{ | |
| 2289 | - if (!oh || (oh.isIndirect() && !(oh.isStream() && oh.getObjGen() == og))) { | |
| 2290 | - QTC::TC("qpdf", "QPDF replaceObject called with indirect object"); | |
| 2291 | - throw std::logic_error("QPDF::replaceObject called with indirect object handle"); | |
| 2292 | - } | |
| 2293 | - updateCache(og, oh.getObj()); | |
| 2294 | -} | |
| 2295 | - | |
| 2296 | -void | |
| 2297 | -QPDF::removeObject(QPDFObjGen og) | |
| 2298 | -{ | |
| 2299 | - if (auto cached = m->obj_cache.find(og); cached != m->obj_cache.end()) { | |
| 2300 | - // Take care of any object handles that may be floating around. | |
| 2301 | - cached->second.object->assign(QPDF_Null::create()); | |
| 2302 | - cached->second.object->setObjGen(nullptr, QPDFObjGen()); | |
| 2303 | - m->obj_cache.erase(cached); | |
| 2304 | - } | |
| 2305 | -} | |
| 2306 | - | |
| 2307 | -void | |
| 2308 | 613 | QPDF::replaceReserved(QPDFObjectHandle reserved, QPDFObjectHandle replacement) |
| 2309 | 614 | { |
| 2310 | 615 | QTC::TC("qpdf", "QPDF replaceReserved"); |
| ... | ... | @@ -2581,15 +886,6 @@ QPDF::swapObjects(int objid1, int generation1, int objid2, int generation2) |
| 2581 | 886 | swapObjects(QPDFObjGen(objid1, generation1), QPDFObjGen(objid2, generation2)); |
| 2582 | 887 | } |
| 2583 | 888 | |
| 2584 | -void | |
| 2585 | -QPDF::swapObjects(QPDFObjGen const& og1, QPDFObjGen const& og2) | |
| 2586 | -{ | |
| 2587 | - // Force objects to be read from the input source if needed, then swap them in the cache. | |
| 2588 | - resolve(og1); | |
| 2589 | - resolve(og2); | |
| 2590 | - m->obj_cache[og1].object->swapWith(m->obj_cache[og2].object); | |
| 2591 | -} | |
| 2592 | - | |
| 2593 | 889 | unsigned long long |
| 2594 | 890 | QPDF::getUniqueId() const |
| 2595 | 891 | { |
| ... | ... | @@ -2676,137 +972,6 @@ QPDF::getXRefTable() |
| 2676 | 972 | return m->xref_table.as_map(); |
| 2677 | 973 | } |
| 2678 | 974 | |
| 2679 | -size_t | |
| 2680 | -QPDF::tableSize() | |
| 2681 | -{ | |
| 2682 | - // If obj_cache is dense, accommodate all object in tables,else accommodate only original | |
| 2683 | - // objects. | |
| 2684 | - auto max_xref = toI(m->xref_table.size()); | |
| 2685 | - if (max_xref > 0) { | |
| 2686 | - --max_xref; | |
| 2687 | - } | |
| 2688 | - auto max_obj = m->obj_cache.size() ? m->obj_cache.crbegin()->first.getObj() : 0; | |
| 2689 | - auto max_id = std::numeric_limits<int>::max() - 1; | |
| 2690 | - if (max_obj >= max_id || max_xref >= max_id) { | |
| 2691 | - // Temporary fix. Long-term solution is | |
| 2692 | - // - QPDFObjGen to enforce objgens are valid and sensible | |
| 2693 | - // - xref table and obj cache to protect against insertion of impossibly large obj ids | |
| 2694 | - stopOnError("Impossibly large object id encountered."); | |
| 2695 | - } | |
| 2696 | - if (max_obj < 1.1 * std::max(toI(m->obj_cache.size()), max_xref)) { | |
| 2697 | - return toS(++max_obj); | |
| 2698 | - } | |
| 2699 | - return toS(++max_xref); | |
| 2700 | -} | |
| 2701 | - | |
| 2702 | -std::vector<QPDFObjGen> | |
| 2703 | -QPDF::getCompressibleObjVector() | |
| 2704 | -{ | |
| 2705 | - return getCompressibleObjGens<QPDFObjGen>(); | |
| 2706 | -} | |
| 2707 | - | |
| 2708 | -std::vector<bool> | |
| 2709 | -QPDF::getCompressibleObjSet() | |
| 2710 | -{ | |
| 2711 | - return getCompressibleObjGens<bool>(); | |
| 2712 | -} | |
| 2713 | - | |
| 2714 | -template <typename T> | |
| 2715 | -std::vector<T> | |
| 2716 | -QPDF::getCompressibleObjGens() | |
| 2717 | -{ | |
| 2718 | - // Return a list of objects that are allowed to be in object streams. Walk through the objects | |
| 2719 | - // by traversing the document from the root, including a traversal of the pages tree. This | |
| 2720 | - // makes that objects that are on the same page are more likely to be in the same object stream, | |
| 2721 | - // which is slightly more efficient, particularly with linearized files. This is better than | |
| 2722 | - // iterating through the xref table since it avoids preserving orphaned items. | |
| 2723 | - | |
| 2724 | - // Exclude encryption dictionary, if any | |
| 2725 | - QPDFObjectHandle encryption_dict = m->xref_table.trailer().getKey("/Encrypt"); | |
| 2726 | - QPDFObjGen encryption_dict_og = encryption_dict.getObjGen(); | |
| 2727 | - | |
| 2728 | - const size_t max_obj = getObjectCount(); | |
| 2729 | - std::vector<bool> visited(max_obj, false); | |
| 2730 | - std::vector<QPDFObjectHandle> queue; | |
| 2731 | - queue.reserve(512); | |
| 2732 | - queue.push_back(m->xref_table.trailer()); | |
| 2733 | - std::vector<T> result; | |
| 2734 | - if constexpr (std::is_same_v<T, QPDFObjGen>) { | |
| 2735 | - result.reserve(m->obj_cache.size()); | |
| 2736 | - } else if constexpr (std::is_same_v<T, bool>) { | |
| 2737 | - result.resize(max_obj + 1U, false); | |
| 2738 | - } else { | |
| 2739 | - throw std::logic_error("Unsupported type in QPDF::getCompressibleObjGens"); | |
| 2740 | - } | |
| 2741 | - while (!queue.empty()) { | |
| 2742 | - auto obj = queue.back(); | |
| 2743 | - queue.pop_back(); | |
| 2744 | - if (obj.getObjectID() > 0) { | |
| 2745 | - QPDFObjGen og = obj.getObjGen(); | |
| 2746 | - const size_t id = toS(og.getObj() - 1); | |
| 2747 | - if (id >= max_obj) { | |
| 2748 | - throw std::logic_error( | |
| 2749 | - "unexpected object id encountered in getCompressibleObjGens"); | |
| 2750 | - } | |
| 2751 | - if (visited[id]) { | |
| 2752 | - QTC::TC("qpdf", "QPDF loop detected traversing objects"); | |
| 2753 | - continue; | |
| 2754 | - } | |
| 2755 | - | |
| 2756 | - // Check whether this is the current object. If not, remove it (which changes it into a | |
| 2757 | - // direct null and therefore stops us from revisiting it) and move on to the next object | |
| 2758 | - // in the queue. | |
| 2759 | - auto upper = m->obj_cache.upper_bound(og); | |
| 2760 | - if (upper != m->obj_cache.end() && upper->first.getObj() == og.getObj()) { | |
| 2761 | - removeObject(og); | |
| 2762 | - continue; | |
| 2763 | - } | |
| 2764 | - | |
| 2765 | - visited[id] = true; | |
| 2766 | - | |
| 2767 | - if (og == encryption_dict_og) { | |
| 2768 | - QTC::TC("qpdf", "QPDF exclude encryption dictionary"); | |
| 2769 | - } else if (!(obj.isStream() || | |
| 2770 | - (obj.isDictionaryOfType("/Sig") && obj.hasKey("/ByteRange") && | |
| 2771 | - obj.hasKey("/Contents")))) { | |
| 2772 | - if constexpr (std::is_same_v<T, QPDFObjGen>) { | |
| 2773 | - result.push_back(og); | |
| 2774 | - } else if constexpr (std::is_same_v<T, bool>) { | |
| 2775 | - result[id + 1U] = true; | |
| 2776 | - } | |
| 2777 | - } | |
| 2778 | - } | |
| 2779 | - if (obj.isStream()) { | |
| 2780 | - QPDFObjectHandle dict = obj.getDict(); | |
| 2781 | - std::set<std::string> keys = dict.getKeys(); | |
| 2782 | - for (auto iter = keys.rbegin(); iter != keys.rend(); ++iter) { | |
| 2783 | - std::string const& key = *iter; | |
| 2784 | - QPDFObjectHandle value = dict.getKey(key); | |
| 2785 | - if (key == "/Length") { | |
| 2786 | - // omit stream lengths | |
| 2787 | - if (value.isIndirect()) { | |
| 2788 | - QTC::TC("qpdf", "QPDF exclude indirect length"); | |
| 2789 | - } | |
| 2790 | - } else { | |
| 2791 | - queue.push_back(value); | |
| 2792 | - } | |
| 2793 | - } | |
| 2794 | - } else if (obj.isDictionary()) { | |
| 2795 | - std::set<std::string> keys = obj.getKeys(); | |
| 2796 | - for (auto iter = keys.rbegin(); iter != keys.rend(); ++iter) { | |
| 2797 | - queue.push_back(obj.getKey(*iter)); | |
| 2798 | - } | |
| 2799 | - } else if (obj.isArray()) { | |
| 2800 | - int n = obj.getArrayNItems(); | |
| 2801 | - for (int i = 1; i <= n; ++i) { | |
| 2802 | - queue.push_back(obj.getArrayItem(n - i)); | |
| 2803 | - } | |
| 2804 | - } | |
| 2805 | - } | |
| 2806 | - | |
| 2807 | - return result; | |
| 2808 | -} | |
| 2809 | - | |
| 2810 | 975 | bool |
| 2811 | 976 | QPDF::pipeStreamData( |
| 2812 | 977 | std::shared_ptr<EncryptionParameters> encp, | ... | ... |
libqpdf/QPDF_objects.cc
| ... | ... | @@ -3,16 +3,12 @@ |
| 3 | 3 | #include <qpdf/QPDF_private.hh> |
| 4 | 4 | |
| 5 | 5 | #include <array> |
| 6 | -#include <atomic> | |
| 7 | 6 | #include <cstring> |
| 8 | 7 | #include <limits> |
| 9 | 8 | #include <map> |
| 10 | -#include <regex> | |
| 11 | -#include <sstream> | |
| 12 | 9 | #include <vector> |
| 13 | 10 | |
| 14 | 11 | #include <qpdf/BufferInputSource.hh> |
| 15 | -#include <qpdf/FileInputSource.hh> | |
| 16 | 12 | #include <qpdf/OffsetInputSource.hh> |
| 17 | 13 | #include <qpdf/Pipeline.hh> |
| 18 | 14 | #include <qpdf/QPDFExc.hh> |
| ... | ... | @@ -28,10 +24,6 @@ |
| 28 | 24 | #include <qpdf/QTC.hh> |
| 29 | 25 | #include <qpdf/QUtil.hh> |
| 30 | 26 | |
| 31 | -// This must be a fixed value. This API returns a const reference to it, and the C API relies on its | |
| 32 | -// being static as well. | |
| 33 | -std::string const QPDF::qpdf_version(QPDF_VERSION); | |
| 34 | - | |
| 35 | 27 | namespace |
| 36 | 28 | { |
| 37 | 29 | class InvalidInputSource final: public InputSource |
| ... | ... | @@ -94,321 +86,6 @@ namespace |
| 94 | 86 | }; |
| 95 | 87 | } // namespace |
| 96 | 88 | |
| 97 | -QPDF::ForeignStreamData::ForeignStreamData( | |
| 98 | - std::shared_ptr<EncryptionParameters> encp, | |
| 99 | - std::shared_ptr<InputSource> file, | |
| 100 | - QPDFObjGen const& foreign_og, | |
| 101 | - qpdf_offset_t offset, | |
| 102 | - size_t length, | |
| 103 | - QPDFObjectHandle local_dict) : | |
| 104 | - encp(encp), | |
| 105 | - file(file), | |
| 106 | - foreign_og(foreign_og), | |
| 107 | - offset(offset), | |
| 108 | - length(length), | |
| 109 | - local_dict(local_dict) | |
| 110 | -{ | |
| 111 | -} | |
| 112 | - | |
| 113 | -QPDF::CopiedStreamDataProvider::CopiedStreamDataProvider(QPDF& destination_qpdf) : | |
| 114 | - QPDFObjectHandle::StreamDataProvider(true), | |
| 115 | - destination_qpdf(destination_qpdf) | |
| 116 | -{ | |
| 117 | -} | |
| 118 | - | |
| 119 | -bool | |
| 120 | -QPDF::CopiedStreamDataProvider::provideStreamData( | |
| 121 | - QPDFObjGen const& og, Pipeline* pipeline, bool suppress_warnings, bool will_retry) | |
| 122 | -{ | |
| 123 | - std::shared_ptr<ForeignStreamData> foreign_data = foreign_stream_data[og]; | |
| 124 | - bool result = false; | |
| 125 | - if (foreign_data.get()) { | |
| 126 | - result = destination_qpdf.pipeForeignStreamData( | |
| 127 | - foreign_data, pipeline, suppress_warnings, will_retry); | |
| 128 | - QTC::TC("qpdf", "QPDF copy foreign with data", result ? 0 : 1); | |
| 129 | - } else { | |
| 130 | - auto foreign_stream = foreign_streams[og]; | |
| 131 | - result = foreign_stream.pipeStreamData( | |
| 132 | - pipeline, nullptr, 0, qpdf_dl_none, suppress_warnings, will_retry); | |
| 133 | - QTC::TC("qpdf", "QPDF copy foreign with foreign_stream", result ? 0 : 1); | |
| 134 | - } | |
| 135 | - return result; | |
| 136 | -} | |
| 137 | - | |
| 138 | -void | |
| 139 | -QPDF::CopiedStreamDataProvider::registerForeignStream( | |
| 140 | - QPDFObjGen const& local_og, QPDFObjectHandle foreign_stream) | |
| 141 | -{ | |
| 142 | - this->foreign_streams[local_og] = foreign_stream; | |
| 143 | -} | |
| 144 | - | |
| 145 | -void | |
| 146 | -QPDF::CopiedStreamDataProvider::registerForeignStream( | |
| 147 | - QPDFObjGen const& local_og, std::shared_ptr<ForeignStreamData> foreign_stream) | |
| 148 | -{ | |
| 149 | - this->foreign_stream_data[local_og] = foreign_stream; | |
| 150 | -} | |
| 151 | - | |
| 152 | -QPDF::StringDecrypter::StringDecrypter(QPDF* qpdf, QPDFObjGen const& og) : | |
| 153 | - qpdf(qpdf), | |
| 154 | - og(og) | |
| 155 | -{ | |
| 156 | -} | |
| 157 | - | |
| 158 | -void | |
| 159 | -QPDF::StringDecrypter::decryptString(std::string& val) | |
| 160 | -{ | |
| 161 | - qpdf->decryptString(val, og); | |
| 162 | -} | |
| 163 | - | |
| 164 | -std::string const& | |
| 165 | -QPDF::QPDFVersion() | |
| 166 | -{ | |
| 167 | - // The C API relies on this being a static value. | |
| 168 | - return QPDF::qpdf_version; | |
| 169 | -} | |
| 170 | - | |
| 171 | -QPDF::EncryptionParameters::EncryptionParameters() : | |
| 172 | - encrypted(false), | |
| 173 | - encryption_initialized(false), | |
| 174 | - encryption_V(0), | |
| 175 | - encryption_R(0), | |
| 176 | - encrypt_metadata(true), | |
| 177 | - cf_stream(e_none), | |
| 178 | - cf_string(e_none), | |
| 179 | - cf_file(e_none), | |
| 180 | - user_password_matched(false), | |
| 181 | - owner_password_matched(false) | |
| 182 | -{ | |
| 183 | -} | |
| 184 | - | |
| 185 | -QPDF::Members::Members(QPDF& qpdf) : | |
| 186 | - log(QPDFLogger::defaultLogger()), | |
| 187 | - file_sp(new InvalidInputSource(no_input_name)), | |
| 188 | - file(file_sp.get()), | |
| 189 | - encp(new EncryptionParameters), | |
| 190 | - xref_table(qpdf, file) | |
| 191 | -{ | |
| 192 | -} | |
| 193 | - | |
| 194 | -QPDF::QPDF() : | |
| 195 | - m(new Members(*this)) | |
| 196 | -{ | |
| 197 | - m->tokenizer.allowEOF(); | |
| 198 | - // Generate a unique ID. It just has to be unique among all QPDF objects allocated throughout | |
| 199 | - // the lifetime of this running application. | |
| 200 | - static std::atomic<unsigned long long> unique_id{0}; | |
| 201 | - m->unique_id = unique_id.fetch_add(1ULL); | |
| 202 | -} | |
| 203 | - | |
| 204 | -QPDF::~QPDF() | |
| 205 | -{ | |
| 206 | - // If two objects are mutually referential (through each object having an array or dictionary | |
| 207 | - // that contains an indirect reference to the other), the circular references in the | |
| 208 | - // std::shared_ptr objects will prevent the objects from being deleted. Walk through all objects | |
| 209 | - // in the object cache, which is those objects that we read from the file, and break all | |
| 210 | - // resolved indirect references by replacing them with an internal object type representing that | |
| 211 | - // they have been destroyed. Note that we can't break references like this at any time when the | |
| 212 | - // QPDF object is active. The call to reset also causes all direct QPDFObjectHandle objects that | |
| 213 | - // are reachable from this object to release their association with this QPDF. Direct objects | |
| 214 | - // are not destroyed since they can be moved to other QPDF objects safely. | |
| 215 | - | |
| 216 | - for (auto const& iter: m->obj_cache) { | |
| 217 | - iter.second.object->disconnect(); | |
| 218 | - if (iter.second.object->getTypeCode() != ::ot_null) { | |
| 219 | - iter.second.object->destroy(); | |
| 220 | - } | |
| 221 | - } | |
| 222 | -} | |
| 223 | - | |
| 224 | -std::shared_ptr<QPDF> | |
| 225 | -QPDF::create() | |
| 226 | -{ | |
| 227 | - return std::make_shared<QPDF>(); | |
| 228 | -} | |
| 229 | - | |
| 230 | -void | |
| 231 | -QPDF::processFile(char const* filename, char const* password) | |
| 232 | -{ | |
| 233 | - auto* fi = new FileInputSource(filename); | |
| 234 | - processInputSource(std::shared_ptr<InputSource>(fi), password); | |
| 235 | -} | |
| 236 | - | |
| 237 | -void | |
| 238 | -QPDF::processFile(char const* description, FILE* filep, bool close_file, char const* password) | |
| 239 | -{ | |
| 240 | - auto* fi = new FileInputSource(description, filep, close_file); | |
| 241 | - processInputSource(std::shared_ptr<InputSource>(fi), password); | |
| 242 | -} | |
| 243 | - | |
| 244 | -void | |
| 245 | -QPDF::processMemoryFile( | |
| 246 | - char const* description, char const* buf, size_t length, char const* password) | |
| 247 | -{ | |
| 248 | - processInputSource( | |
| 249 | - std::shared_ptr<InputSource>( | |
| 250 | - // line-break | |
| 251 | - new BufferInputSource( | |
| 252 | - description, new Buffer(QUtil::unsigned_char_pointer(buf), length), true)), | |
| 253 | - password); | |
| 254 | -} | |
| 255 | - | |
| 256 | -void | |
| 257 | -QPDF::processInputSource(std::shared_ptr<InputSource> source, char const* password) | |
| 258 | -{ | |
| 259 | - m->file_sp = source; | |
| 260 | - m->file = source.get(); | |
| 261 | - parse(password); | |
| 262 | -} | |
| 263 | - | |
| 264 | -void | |
| 265 | -QPDF::closeInputSource() | |
| 266 | -{ | |
| 267 | - m->no_input_name = "closed input source"; | |
| 268 | - m->file_sp = std::shared_ptr<InputSource>(new InvalidInputSource(m->no_input_name)); | |
| 269 | - m->file = m->file_sp.get(); | |
| 270 | -} | |
| 271 | - | |
| 272 | -void | |
| 273 | -QPDF::setPasswordIsHexKey(bool val) | |
| 274 | -{ | |
| 275 | - m->provided_password_is_hex_key = val; | |
| 276 | -} | |
| 277 | - | |
| 278 | -void | |
| 279 | -QPDF::emptyPDF() | |
| 280 | -{ | |
| 281 | - m->pdf_version = "1.3"; | |
| 282 | - m->no_input_name = "empty PDF"; | |
| 283 | - m->xref_table.initialize_empty(); | |
| 284 | -} | |
| 285 | - | |
| 286 | -void | |
| 287 | -QPDF::registerStreamFilter( | |
| 288 | - std::string const& filter_name, std::function<std::shared_ptr<QPDFStreamFilter>()> factory) | |
| 289 | -{ | |
| 290 | - QPDF_Stream::registerStreamFilter(filter_name, factory); | |
| 291 | -} | |
| 292 | - | |
| 293 | -void | |
| 294 | -QPDF::setIgnoreXRefStreams(bool val) | |
| 295 | -{ | |
| 296 | - m->xref_table.ignore_streams(val); | |
| 297 | -} | |
| 298 | - | |
| 299 | -std::shared_ptr<QPDFLogger> | |
| 300 | -QPDF::getLogger() | |
| 301 | -{ | |
| 302 | - return m->log; | |
| 303 | -} | |
| 304 | - | |
| 305 | -void | |
| 306 | -QPDF::setLogger(std::shared_ptr<QPDFLogger> l) | |
| 307 | -{ | |
| 308 | - m->log = l; | |
| 309 | -} | |
| 310 | - | |
| 311 | -void | |
| 312 | -QPDF::setOutputStreams(std::ostream* out, std::ostream* err) | |
| 313 | -{ | |
| 314 | - setLogger(QPDFLogger::create()); | |
| 315 | - m->log->setOutputStreams(out, err); | |
| 316 | -} | |
| 317 | - | |
| 318 | -void | |
| 319 | -QPDF::setSuppressWarnings(bool val) | |
| 320 | -{ | |
| 321 | - m->suppress_warnings = val; | |
| 322 | -} | |
| 323 | - | |
| 324 | -void | |
| 325 | -QPDF::setMaxWarnings(size_t val) | |
| 326 | -{ | |
| 327 | - m->max_warnings = val; | |
| 328 | -} | |
| 329 | - | |
| 330 | -void | |
| 331 | -QPDF::setAttemptRecovery(bool val) | |
| 332 | -{ | |
| 333 | - m->attempt_recovery = val; | |
| 334 | - m->xref_table.attempt_recovery(val); | |
| 335 | -} | |
| 336 | - | |
| 337 | -void | |
| 338 | -QPDF::setImmediateCopyFrom(bool val) | |
| 339 | -{ | |
| 340 | - m->immediate_copy_from = val; | |
| 341 | -} | |
| 342 | - | |
| 343 | -std::vector<QPDFExc> | |
| 344 | -QPDF::getWarnings() | |
| 345 | -{ | |
| 346 | - std::vector<QPDFExc> result = m->warnings; | |
| 347 | - m->warnings.clear(); | |
| 348 | - return result; | |
| 349 | -} | |
| 350 | - | |
| 351 | -bool | |
| 352 | -QPDF::anyWarnings() const | |
| 353 | -{ | |
| 354 | - return !m->warnings.empty(); | |
| 355 | -} | |
| 356 | - | |
| 357 | -size_t | |
| 358 | -QPDF::numWarnings() const | |
| 359 | -{ | |
| 360 | - return m->warnings.size(); | |
| 361 | -} | |
| 362 | - | |
| 363 | -bool | |
| 364 | -QPDF::validatePDFVersion(char const*& p, std::string& version) | |
| 365 | -{ | |
| 366 | - bool valid = QUtil::is_digit(*p); | |
| 367 | - if (valid) { | |
| 368 | - while (QUtil::is_digit(*p)) { | |
| 369 | - version.append(1, *p++); | |
| 370 | - } | |
| 371 | - if ((*p == '.') && QUtil::is_digit(*(p + 1))) { | |
| 372 | - version.append(1, *p++); | |
| 373 | - while (QUtil::is_digit(*p)) { | |
| 374 | - version.append(1, *p++); | |
| 375 | - } | |
| 376 | - } else { | |
| 377 | - valid = false; | |
| 378 | - } | |
| 379 | - } | |
| 380 | - return valid; | |
| 381 | -} | |
| 382 | - | |
| 383 | -bool | |
| 384 | -QPDF::findHeader() | |
| 385 | -{ | |
| 386 | - qpdf_offset_t global_offset = m->file->tell(); | |
| 387 | - std::string line = m->file->readLine(1024); | |
| 388 | - char const* p = line.c_str(); | |
| 389 | - if (strncmp(p, "%PDF-", 5) != 0) { | |
| 390 | - throw std::logic_error("findHeader is not looking at %PDF-"); | |
| 391 | - } | |
| 392 | - p += 5; | |
| 393 | - std::string version; | |
| 394 | - // Note: The string returned by line.c_str() is always null-terminated. The code below never | |
| 395 | - // overruns the buffer because a null character always short-circuits further advancement. | |
| 396 | - bool valid = validatePDFVersion(p, version); | |
| 397 | - if (valid) { | |
| 398 | - m->pdf_version = version; | |
| 399 | - if (global_offset != 0) { | |
| 400 | - // Empirical evidence strongly suggests that when there is leading material prior to the | |
| 401 | - // PDF header, all explicit offsets in the file are such that 0 points to the beginning | |
| 402 | - // of the header. | |
| 403 | - QTC::TC("qpdf", "QPDF global offset"); | |
| 404 | - m->file_sp = | |
| 405 | - std::shared_ptr<InputSource>(new OffsetInputSource(m->file_sp, global_offset)); | |
| 406 | - m->file = m->file_sp.get(); | |
| 407 | - } | |
| 408 | - } | |
| 409 | - return valid; | |
| 410 | -} | |
| 411 | - | |
| 412 | 89 | bool |
| 413 | 90 | QPDF::findStartxref() |
| 414 | 91 | { |
| ... | ... | @@ -421,64 +98,6 @@ QPDF::findStartxref() |
| 421 | 98 | } |
| 422 | 99 | |
| 423 | 100 | void |
| 424 | -QPDF::parse(char const* password) | |
| 425 | -{ | |
| 426 | - if (password) { | |
| 427 | - m->encp->provided_password = password; | |
| 428 | - } | |
| 429 | - | |
| 430 | - // Find the header anywhere in the first 1024 bytes of the file. | |
| 431 | - PatternFinder hf(*this, &QPDF::findHeader); | |
| 432 | - if (!m->file->findFirst("%PDF-", 0, 1024, hf)) { | |
| 433 | - QTC::TC("qpdf", "QPDF not a pdf file"); | |
| 434 | - warn(damagedPDF("", 0, "can't find PDF header")); | |
| 435 | - // QPDFWriter writes files that usually require at least version 1.2 for /FlateDecode | |
| 436 | - m->pdf_version = "1.2"; | |
| 437 | - } | |
| 438 | - | |
| 439 | - m->xref_table.initialize(); | |
| 440 | - initializeEncryption(); | |
| 441 | - if (m->xref_table.size() > 0 && !getRoot().getKey("/Pages").isDictionary()) { | |
| 442 | - // QPDFs created from JSON have an empty xref table and no root object yet. | |
| 443 | - throw damagedPDF("", 0, "unable to find page tree"); | |
| 444 | - } | |
| 445 | -} | |
| 446 | - | |
| 447 | -void | |
| 448 | -QPDF::inParse(bool v) | |
| 449 | -{ | |
| 450 | - if (m->in_parse == v) { | |
| 451 | - // This happens if QPDFParser::parse tries to resolve an indirect object while it is | |
| 452 | - // parsing. | |
| 453 | - throw std::logic_error("QPDF: re-entrant parsing detected. This is a qpdf bug." | |
| 454 | - " Please report at https://github.com/qpdf/qpdf/issues."); | |
| 455 | - } | |
| 456 | - m->in_parse = v; | |
| 457 | -} | |
| 458 | - | |
| 459 | -void | |
| 460 | -QPDF::warn(QPDFExc const& e) | |
| 461 | -{ | |
| 462 | - if (m->max_warnings > 0 && m->warnings.size() >= m->max_warnings) { | |
| 463 | - stopOnError("Too many warnings - file is too badly damaged"); | |
| 464 | - } | |
| 465 | - m->warnings.push_back(e); | |
| 466 | - if (!m->suppress_warnings) { | |
| 467 | - *m->log->getWarn() << "WARNING: " << m->warnings.back().what() << "\n"; | |
| 468 | - } | |
| 469 | -} | |
| 470 | - | |
| 471 | -void | |
| 472 | -QPDF::warn( | |
| 473 | - qpdf_error_code_e error_code, | |
| 474 | - std::string const& object, | |
| 475 | - qpdf_offset_t offset, | |
| 476 | - std::string const& message) | |
| 477 | -{ | |
| 478 | - warn(QPDFExc(error_code, getFilename(), object, offset, message)); | |
| 479 | -} | |
| 480 | - | |
| 481 | -void | |
| 482 | 101 | QPDF::Xref_table::initialize_empty() |
| 483 | 102 | { |
| 484 | 103 | initialized_ = true; |
| ... | ... | @@ -1384,7 +1003,7 @@ QPDF::Xref_table::insert(int obj, int f0, qpdf_offset_t f1, int f2) |
| 1384 | 1003 | if (!old_type && entry.gen() > 0) { |
| 1385 | 1004 | // At the moment we are processing the updates last to first and therefore the gen doesn't |
| 1386 | 1005 | // matter as long as it > 0 to distinguish it from an uninitialized entry. This will need |
| 1387 | - // to be revisited when we want to support incremental updates or more comprhensive | |
| 1006 | + // to be revisited when we want to support incremental updates or more comprehensive | |
| 1388 | 1007 | // checking. |
| 1389 | 1008 | QTC::TC("qpdf", "QPDF xref deleted object"); |
| 1390 | 1009 | return; |
| ... | ... | @@ -1425,7 +1044,7 @@ QPDF::Xref_table::insert_free(QPDFObjGen og) |
| 1425 | 1044 | { |
| 1426 | 1045 | // At the moment we are processing the updates last to first and therefore the gen doesn't |
| 1427 | 1046 | // matter as long as it > 0 to distinguish it from an uninitialized entry. This will need to be |
| 1428 | - // revisited when we want to support incremental updates or more comprhensive checking. | |
| 1047 | + // revisited when we want to support incremental updates or more comprehensive checking. | |
| 1429 | 1048 | if (og.getObj() < 1) { |
| 1430 | 1049 | return; |
| 1431 | 1050 | } |
| ... | ... | @@ -1480,12 +1099,6 @@ QPDF::Xref_table::as_map() const |
| 1480 | 1099 | } |
| 1481 | 1100 | |
| 1482 | 1101 | void |
| 1483 | -QPDF::showXRefTable() | |
| 1484 | -{ | |
| 1485 | - m->xref_table.show(); | |
| 1486 | -} | |
| 1487 | - | |
| 1488 | -void | |
| 1489 | 1102 | QPDF::Xref_table::show() |
| 1490 | 1103 | { |
| 1491 | 1104 | auto& cout = *qpdf.m->log->getInfo(); |
| ... | ... | @@ -1533,35 +1146,6 @@ QPDF::Xref_table::resolve() |
| 1533 | 1146 | return true; |
| 1534 | 1147 | } |
| 1535 | 1148 | |
| 1536 | -// Ensure all objects in the pdf file, including those in indirect references, appear in the object | |
| 1537 | -// cache. | |
| 1538 | -void | |
| 1539 | -QPDF::fixDanglingReferences(bool force) | |
| 1540 | -{ | |
| 1541 | - if (m->fixed_dangling_refs) { | |
| 1542 | - return; | |
| 1543 | - } | |
| 1544 | - if (!m->xref_table.resolve()) { | |
| 1545 | - QTC::TC("qpdf", "QPDF fix dangling triggered xref reconstruction"); | |
| 1546 | - m->xref_table.resolve(); | |
| 1547 | - } | |
| 1548 | - m->fixed_dangling_refs = true; | |
| 1549 | -} | |
| 1550 | - | |
| 1551 | -size_t | |
| 1552 | -QPDF::getObjectCount() | |
| 1553 | -{ | |
| 1554 | - // This method returns the next available indirect object number. makeIndirectObject uses it for | |
| 1555 | - // this purpose. After fixDanglingReferences is called, all objects in the xref table will also | |
| 1556 | - // be in obj_cache. | |
| 1557 | - fixDanglingReferences(); | |
| 1558 | - QPDFObjGen og; | |
| 1559 | - if (!m->obj_cache.empty()) { | |
| 1560 | - og = (*(m->obj_cache.rbegin())).first; | |
| 1561 | - } | |
| 1562 | - return toS(og.getObj()); | |
| 1563 | -} | |
| 1564 | - | |
| 1565 | 1149 | std::vector<QPDFObjectHandle> |
| 1566 | 1150 | QPDF::getAllObjects() |
| 1567 | 1151 | { |
| ... | ... | @@ -1574,21 +1158,6 @@ QPDF::getAllObjects() |
| 1574 | 1158 | return result; |
| 1575 | 1159 | } |
| 1576 | 1160 | |
| 1577 | -void | |
| 1578 | -QPDF::setLastObjectDescription(std::string const& description, QPDFObjGen const& og) | |
| 1579 | -{ | |
| 1580 | - m->last_object_description.clear(); | |
| 1581 | - if (!description.empty()) { | |
| 1582 | - m->last_object_description += description; | |
| 1583 | - if (og.isIndirect()) { | |
| 1584 | - m->last_object_description += ": "; | |
| 1585 | - } | |
| 1586 | - } | |
| 1587 | - if (og.isIndirect()) { | |
| 1588 | - m->last_object_description += "object " + og.unparse(' '); | |
| 1589 | - } | |
| 1590 | -} | |
| 1591 | - | |
| 1592 | 1161 | QPDFObjectHandle |
| 1593 | 1162 | QPDF::Xref_table::read_trailer() |
| 1594 | 1163 | { |
| ... | ... | @@ -1783,7 +1352,7 @@ QPDF::recoverStreamLength( |
| 1783 | 1352 | if (found == QPDFObjGen() || found == og) { |
| 1784 | 1353 | // If we are trying to recover an XRef stream the xref table will not contain and |
| 1785 | 1354 | // won't contain any entries, therefore we cannot check the found length. Otherwise we |
| 1786 | - // found endstream\nendobj within the space allowed for this object, so we're probably | |
| 1355 | + // found endstream\endobj within the space allowed for this object, so we're probably | |
| 1787 | 1356 | // in good shape. |
| 1788 | 1357 | } else { |
| 1789 | 1358 | QTC::TC("qpdf", "QPDF found wrong endstream in recovery"); |
| ... | ... | @@ -1803,12 +1372,6 @@ QPDF::recoverStreamLength( |
| 1803 | 1372 | return length; |
| 1804 | 1373 | } |
| 1805 | 1374 | |
| 1806 | -QPDFTokenizer::Token | |
| 1807 | -QPDF::readToken(InputSource& input, size_t max_len) | |
| 1808 | -{ | |
| 1809 | - return m->tokenizer.readToken(input, m->last_object_description, true, max_len); | |
| 1810 | -} | |
| 1811 | - | |
| 1812 | 1375 | QPDFObjectHandle |
| 1813 | 1376 | QPDF::readObjectAtOffset( |
| 1814 | 1377 | bool try_recovery, |
| ... | ... | @@ -2122,13 +1685,6 @@ QPDF::resolveObjectsInStream(int obj_stream_number) |
| 2122 | 1685 | } |
| 2123 | 1686 | } |
| 2124 | 1687 | |
| 2125 | -QPDFObjectHandle | |
| 2126 | -QPDF::newIndirect(QPDFObjGen const& og, std::shared_ptr<QPDFObject> const& obj) | |
| 2127 | -{ | |
| 2128 | - obj->setDefaultDescription(this, og); | |
| 2129 | - return {obj}; | |
| 2130 | -} | |
| 2131 | - | |
| 2132 | 1688 | void |
| 2133 | 1689 | QPDF::updateCache(QPDFObjGen const& og, std::shared_ptr<QPDFObject> const& object) |
| 2134 | 1690 | { |
| ... | ... | @@ -2171,50 +1727,6 @@ QPDF::makeIndirectFromQPDFObject(std::shared_ptr<QPDFObject> const& obj) |
| 2171 | 1727 | return newIndirect(next, m->obj_cache[next].object); |
| 2172 | 1728 | } |
| 2173 | 1729 | |
| 2174 | -QPDFObjectHandle | |
| 2175 | -QPDF::makeIndirectObject(QPDFObjectHandle oh) | |
| 2176 | -{ | |
| 2177 | - if (!oh) { | |
| 2178 | - throw std::logic_error("attempted to make an uninitialized QPDFObjectHandle indirect"); | |
| 2179 | - } | |
| 2180 | - return makeIndirectFromQPDFObject(oh.getObj()); | |
| 2181 | -} | |
| 2182 | - | |
| 2183 | -QPDFObjectHandle | |
| 2184 | -QPDF::newReserved() | |
| 2185 | -{ | |
| 2186 | - return makeIndirectFromQPDFObject(QPDF_Reserved::create()); | |
| 2187 | -} | |
| 2188 | - | |
| 2189 | -QPDFObjectHandle | |
| 2190 | -QPDF::newIndirectNull() | |
| 2191 | -{ | |
| 2192 | - return makeIndirectFromQPDFObject(QPDF_Null::create()); | |
| 2193 | -} | |
| 2194 | - | |
| 2195 | -QPDFObjectHandle | |
| 2196 | -QPDF::newStream() | |
| 2197 | -{ | |
| 2198 | - return makeIndirectFromQPDFObject( | |
| 2199 | - QPDF_Stream::create(this, nextObjGen(), QPDFObjectHandle::newDictionary(), 0, 0)); | |
| 2200 | -} | |
| 2201 | - | |
| 2202 | -QPDFObjectHandle | |
| 2203 | -QPDF::newStream(std::shared_ptr<Buffer> data) | |
| 2204 | -{ | |
| 2205 | - auto result = newStream(); | |
| 2206 | - result.replaceStreamData(data, QPDFObjectHandle::newNull(), QPDFObjectHandle::newNull()); | |
| 2207 | - return result; | |
| 2208 | -} | |
| 2209 | - | |
| 2210 | -QPDFObjectHandle | |
| 2211 | -QPDF::newStream(std::string const& data) | |
| 2212 | -{ | |
| 2213 | - auto result = newStream(); | |
| 2214 | - result.replaceStreamData(data, QPDFObjectHandle::newNull(), QPDFObjectHandle::newNull()); | |
| 2215 | - return result; | |
| 2216 | -} | |
| 2217 | - | |
| 2218 | 1730 | std::shared_ptr<QPDFObject> |
| 2219 | 1731 | QPDF::getObjectForParser(int id, int gen, bool parse_pdf) |
| 2220 | 1732 | { |
| ... | ... | @@ -2246,43 +1758,6 @@ QPDF::getObjectForJSON(int id, int gen) |
| 2246 | 1758 | return obj; |
| 2247 | 1759 | } |
| 2248 | 1760 | |
| 2249 | -QPDFObjectHandle | |
| 2250 | -QPDF::getObject(QPDFObjGen const& og) | |
| 2251 | -{ | |
| 2252 | - if (auto it = m->obj_cache.find(og); it != m->obj_cache.end()) { | |
| 2253 | - return {it->second.object}; | |
| 2254 | - } else if (m->xref_table.initialized() && !m->xref_table.type(og)) { | |
| 2255 | - return QPDF_Null::create(); | |
| 2256 | - } else { | |
| 2257 | - auto result = m->obj_cache.try_emplace(og, QPDF_Unresolved::create(this, og)); | |
| 2258 | - return {result.first->second.object}; | |
| 2259 | - } | |
| 2260 | -} | |
| 2261 | - | |
| 2262 | -QPDFObjectHandle | |
| 2263 | -QPDF::getObject(int objid, int generation) | |
| 2264 | -{ | |
| 2265 | - return getObject(QPDFObjGen(objid, generation)); | |
| 2266 | -} | |
| 2267 | - | |
| 2268 | -QPDFObjectHandle | |
| 2269 | -QPDF::getObjectByObjGen(QPDFObjGen const& og) | |
| 2270 | -{ | |
| 2271 | - return getObject(og); | |
| 2272 | -} | |
| 2273 | - | |
| 2274 | -QPDFObjectHandle | |
| 2275 | -QPDF::getObjectByID(int objid, int generation) | |
| 2276 | -{ | |
| 2277 | - return getObject(QPDFObjGen(objid, generation)); | |
| 2278 | -} | |
| 2279 | - | |
| 2280 | -void | |
| 2281 | -QPDF::replaceObject(int objid, int generation, QPDFObjectHandle oh) | |
| 2282 | -{ | |
| 2283 | - replaceObject(QPDFObjGen(objid, generation), oh); | |
| 2284 | -} | |
| 2285 | - | |
| 2286 | 1761 | void |
| 2287 | 1762 | QPDF::replaceObject(QPDFObjGen const& og, QPDFObjectHandle oh) |
| 2288 | 1763 | { |
| ... | ... | @@ -2305,283 +1780,6 @@ QPDF::removeObject(QPDFObjGen og) |
| 2305 | 1780 | } |
| 2306 | 1781 | |
| 2307 | 1782 | void |
| 2308 | -QPDF::replaceReserved(QPDFObjectHandle reserved, QPDFObjectHandle replacement) | |
| 2309 | -{ | |
| 2310 | - QTC::TC("qpdf", "QPDF replaceReserved"); | |
| 2311 | - auto tc = reserved.getTypeCode(); | |
| 2312 | - if (!(tc == ::ot_reserved || tc == ::ot_null)) { | |
| 2313 | - throw std::logic_error("replaceReserved called with non-reserved object"); | |
| 2314 | - } | |
| 2315 | - replaceObject(reserved.getObjGen(), replacement); | |
| 2316 | -} | |
| 2317 | - | |
| 2318 | -QPDFObjectHandle | |
| 2319 | -QPDF::copyForeignObject(QPDFObjectHandle foreign) | |
| 2320 | -{ | |
| 2321 | - // Here's an explanation of what's going on here. | |
| 2322 | - // | |
| 2323 | - // A QPDFObjectHandle that is an indirect object has an owning QPDF. The object ID and | |
| 2324 | - // generation refers to an object in the owning QPDF. When we copy the QPDFObjectHandle from a | |
| 2325 | - // foreign QPDF into the local QPDF, we have to replace all indirect object references with | |
| 2326 | - // references to the corresponding object in the local file. | |
| 2327 | - // | |
| 2328 | - // To do this, we maintain mappings from foreign object IDs to local object IDs for each foreign | |
| 2329 | - // QPDF that we are copying from. The mapping is stored in an ObjCopier, which contains a | |
| 2330 | - // mapping from the foreign ObjGen to the local QPDFObjectHandle. | |
| 2331 | - // | |
| 2332 | - // To copy, we do a deep traversal of the foreign object with loop detection to discover all | |
| 2333 | - // indirect objects that are encountered, stopping at page boundaries. Whenever we encounter an | |
| 2334 | - // indirect object, we check to see if we have already created a local copy of it. If not, we | |
| 2335 | - // allocate a "reserved" object (or, for a stream, just a new stream) and store in the map the | |
| 2336 | - // mapping from the foreign object ID to the new object. While we | |
| 2337 | - // do this, we keep a list of objects to copy. | |
| 2338 | - // | |
| 2339 | - // Once we are done with the traversal, we copy all the objects that we need to copy. However, | |
| 2340 | - // the copies will contain indirect object IDs that refer to objects in the foreign file. We | |
| 2341 | - // need to replace them with references to objects in the local file. This is what | |
| 2342 | - // replaceForeignIndirectObjects does. Once we have created a copy of the foreign object with | |
| 2343 | - // all the indirect references replaced with new ones in the local context, we can replace the | |
| 2344 | - // local reserved object with the copy. This mechanism allows us to copy objects with circular | |
| 2345 | - // references in any order. | |
| 2346 | - | |
| 2347 | - // For streams, rather than copying the objects, we set up the stream data to pull from the | |
| 2348 | - // original stream by using a stream data provider. This is done in a manner that doesn't | |
| 2349 | - // require the original QPDF object but may require the original source of the stream data with | |
| 2350 | - // special handling for immediate_copy_from. This logic is also in | |
| 2351 | - // replaceForeignIndirectObjects. | |
| 2352 | - | |
| 2353 | - // Note that we explicitly allow use of copyForeignObject on page objects. It is a documented | |
| 2354 | - // use case to copy pages this way if the intention is to not update the pages tree. | |
| 2355 | - if (!foreign.isIndirect()) { | |
| 2356 | - QTC::TC("qpdf", "QPDF copyForeign direct"); | |
| 2357 | - throw std::logic_error("QPDF::copyForeign called with direct object handle"); | |
| 2358 | - } | |
| 2359 | - QPDF& other = foreign.getQPDF(); | |
| 2360 | - if (&other == this) { | |
| 2361 | - QTC::TC("qpdf", "QPDF copyForeign not foreign"); | |
| 2362 | - throw std::logic_error("QPDF::copyForeign called with object from this QPDF"); | |
| 2363 | - } | |
| 2364 | - | |
| 2365 | - ObjCopier& obj_copier = m->object_copiers[other.m->unique_id]; | |
| 2366 | - if (!obj_copier.visiting.empty()) { | |
| 2367 | - throw std::logic_error("obj_copier.visiting is not empty" | |
| 2368 | - " at the beginning of copyForeignObject"); | |
| 2369 | - } | |
| 2370 | - | |
| 2371 | - // Make sure we have an object in this file for every referenced object in the old file. | |
| 2372 | - // obj_copier.object_map maps foreign QPDFObjGen to local objects. For everything new that we | |
| 2373 | - // have to copy, the local object will be a reservation, unless it is a stream, in which case | |
| 2374 | - // the local object will already be a stream. | |
| 2375 | - reserveObjects(foreign, obj_copier, true); | |
| 2376 | - | |
| 2377 | - if (!obj_copier.visiting.empty()) { | |
| 2378 | - throw std::logic_error("obj_copier.visiting is not empty after reserving objects"); | |
| 2379 | - } | |
| 2380 | - | |
| 2381 | - // Copy any new objects and replace the reservations. | |
| 2382 | - for (auto& to_copy: obj_copier.to_copy) { | |
| 2383 | - QPDFObjectHandle copy = replaceForeignIndirectObjects(to_copy, obj_copier, true); | |
| 2384 | - if (!to_copy.isStream()) { | |
| 2385 | - QPDFObjGen og(to_copy.getObjGen()); | |
| 2386 | - replaceReserved(obj_copier.object_map[og], copy); | |
| 2387 | - } | |
| 2388 | - } | |
| 2389 | - obj_copier.to_copy.clear(); | |
| 2390 | - | |
| 2391 | - auto og = foreign.getObjGen(); | |
| 2392 | - if (!obj_copier.object_map.count(og)) { | |
| 2393 | - warn(damagedPDF("unexpected reference to /Pages object while copying foreign object; " | |
| 2394 | - "replacing with null")); | |
| 2395 | - return QPDFObjectHandle::newNull(); | |
| 2396 | - } | |
| 2397 | - return obj_copier.object_map[foreign.getObjGen()]; | |
| 2398 | -} | |
| 2399 | - | |
| 2400 | -void | |
| 2401 | -QPDF::reserveObjects(QPDFObjectHandle foreign, ObjCopier& obj_copier, bool top) | |
| 2402 | -{ | |
| 2403 | - auto foreign_tc = foreign.getTypeCode(); | |
| 2404 | - if (foreign_tc == ::ot_reserved) { | |
| 2405 | - throw std::logic_error("QPDF: attempting to copy a foreign reserved object"); | |
| 2406 | - } | |
| 2407 | - | |
| 2408 | - if (foreign.isPagesObject()) { | |
| 2409 | - QTC::TC("qpdf", "QPDF not copying pages object"); | |
| 2410 | - return; | |
| 2411 | - } | |
| 2412 | - | |
| 2413 | - if (foreign.isIndirect()) { | |
| 2414 | - QPDFObjGen foreign_og(foreign.getObjGen()); | |
| 2415 | - if (!obj_copier.visiting.add(foreign_og)) { | |
| 2416 | - QTC::TC("qpdf", "QPDF loop reserving objects"); | |
| 2417 | - return; | |
| 2418 | - } | |
| 2419 | - if (obj_copier.object_map.count(foreign_og) > 0) { | |
| 2420 | - QTC::TC("qpdf", "QPDF already reserved object"); | |
| 2421 | - if (!(top && foreign.isPageObject() && obj_copier.object_map[foreign_og].isNull())) { | |
| 2422 | - obj_copier.visiting.erase(foreign); | |
| 2423 | - return; | |
| 2424 | - } | |
| 2425 | - } else { | |
| 2426 | - QTC::TC("qpdf", "QPDF copy indirect"); | |
| 2427 | - obj_copier.object_map[foreign_og] = | |
| 2428 | - foreign.isStream() ? newStream() : newIndirectNull(); | |
| 2429 | - if ((!top) && foreign.isPageObject()) { | |
| 2430 | - QTC::TC("qpdf", "QPDF not crossing page boundary"); | |
| 2431 | - obj_copier.visiting.erase(foreign_og); | |
| 2432 | - return; | |
| 2433 | - } | |
| 2434 | - } | |
| 2435 | - obj_copier.to_copy.push_back(foreign); | |
| 2436 | - } | |
| 2437 | - | |
| 2438 | - if (foreign_tc == ::ot_array) { | |
| 2439 | - QTC::TC("qpdf", "QPDF reserve array"); | |
| 2440 | - int n = foreign.getArrayNItems(); | |
| 2441 | - for (int i = 0; i < n; ++i) { | |
| 2442 | - reserveObjects(foreign.getArrayItem(i), obj_copier, false); | |
| 2443 | - } | |
| 2444 | - } else if (foreign_tc == ::ot_dictionary) { | |
| 2445 | - QTC::TC("qpdf", "QPDF reserve dictionary"); | |
| 2446 | - for (auto const& key: foreign.getKeys()) { | |
| 2447 | - reserveObjects(foreign.getKey(key), obj_copier, false); | |
| 2448 | - } | |
| 2449 | - } else if (foreign_tc == ::ot_stream) { | |
| 2450 | - QTC::TC("qpdf", "QPDF reserve stream"); | |
| 2451 | - reserveObjects(foreign.getDict(), obj_copier, false); | |
| 2452 | - } | |
| 2453 | - | |
| 2454 | - obj_copier.visiting.erase(foreign); | |
| 2455 | -} | |
| 2456 | - | |
| 2457 | -QPDFObjectHandle | |
| 2458 | -QPDF::replaceForeignIndirectObjects(QPDFObjectHandle foreign, ObjCopier& obj_copier, bool top) | |
| 2459 | -{ | |
| 2460 | - auto foreign_tc = foreign.getTypeCode(); | |
| 2461 | - QPDFObjectHandle result; | |
| 2462 | - if ((!top) && foreign.isIndirect()) { | |
| 2463 | - QTC::TC("qpdf", "QPDF replace indirect"); | |
| 2464 | - auto mapping = obj_copier.object_map.find(foreign.getObjGen()); | |
| 2465 | - if (mapping == obj_copier.object_map.end()) { | |
| 2466 | - // This case would occur if this is a reference to a Pages object that we didn't | |
| 2467 | - // traverse into. | |
| 2468 | - QTC::TC("qpdf", "QPDF replace foreign indirect with null"); | |
| 2469 | - result = QPDFObjectHandle::newNull(); | |
| 2470 | - } else { | |
| 2471 | - result = mapping->second; | |
| 2472 | - } | |
| 2473 | - } else if (foreign_tc == ::ot_array) { | |
| 2474 | - QTC::TC("qpdf", "QPDF replace array"); | |
| 2475 | - result = QPDFObjectHandle::newArray(); | |
| 2476 | - int n = foreign.getArrayNItems(); | |
| 2477 | - for (int i = 0; i < n; ++i) { | |
| 2478 | - result.appendItem( | |
| 2479 | - // line-break | |
| 2480 | - replaceForeignIndirectObjects(foreign.getArrayItem(i), obj_copier, false)); | |
| 2481 | - } | |
| 2482 | - } else if (foreign_tc == ::ot_dictionary) { | |
| 2483 | - QTC::TC("qpdf", "QPDF replace dictionary"); | |
| 2484 | - result = QPDFObjectHandle::newDictionary(); | |
| 2485 | - std::set<std::string> keys = foreign.getKeys(); | |
| 2486 | - for (auto const& iter: keys) { | |
| 2487 | - result.replaceKey( | |
| 2488 | - iter, replaceForeignIndirectObjects(foreign.getKey(iter), obj_copier, false)); | |
| 2489 | - } | |
| 2490 | - } else if (foreign_tc == ::ot_stream) { | |
| 2491 | - QTC::TC("qpdf", "QPDF replace stream"); | |
| 2492 | - result = obj_copier.object_map[foreign.getObjGen()]; | |
| 2493 | - result.assertStream(); | |
| 2494 | - QPDFObjectHandle dict = result.getDict(); | |
| 2495 | - QPDFObjectHandle old_dict = foreign.getDict(); | |
| 2496 | - std::set<std::string> keys = old_dict.getKeys(); | |
| 2497 | - for (auto const& iter: keys) { | |
| 2498 | - dict.replaceKey( | |
| 2499 | - iter, replaceForeignIndirectObjects(old_dict.getKey(iter), obj_copier, false)); | |
| 2500 | - } | |
| 2501 | - copyStreamData(result, foreign); | |
| 2502 | - } else { | |
| 2503 | - foreign.assertScalar(); | |
| 2504 | - result = foreign; | |
| 2505 | - result.makeDirect(); | |
| 2506 | - } | |
| 2507 | - | |
| 2508 | - if (top && (!result.isStream()) && result.isIndirect()) { | |
| 2509 | - throw std::logic_error("replacement for foreign object is indirect"); | |
| 2510 | - } | |
| 2511 | - | |
| 2512 | - return result; | |
| 2513 | -} | |
| 2514 | - | |
| 2515 | -void | |
| 2516 | -QPDF::copyStreamData(QPDFObjectHandle result, QPDFObjectHandle foreign) | |
| 2517 | -{ | |
| 2518 | - // This method was originally written for copying foreign streams, but it is used by | |
| 2519 | - // QPDFObjectHandle to copy streams from the same QPDF object as well. | |
| 2520 | - | |
| 2521 | - QPDFObjectHandle dict = result.getDict(); | |
| 2522 | - QPDFObjectHandle old_dict = foreign.getDict(); | |
| 2523 | - if (m->copied_stream_data_provider == nullptr) { | |
| 2524 | - m->copied_stream_data_provider = new CopiedStreamDataProvider(*this); | |
| 2525 | - m->copied_streams = | |
| 2526 | - std::shared_ptr<QPDFObjectHandle::StreamDataProvider>(m->copied_stream_data_provider); | |
| 2527 | - } | |
| 2528 | - QPDFObjGen local_og(result.getObjGen()); | |
| 2529 | - // Copy information from the foreign stream so we can pipe its data later without keeping the | |
| 2530 | - // original QPDF object around. | |
| 2531 | - | |
| 2532 | - QPDF& foreign_stream_qpdf = | |
| 2533 | - foreign.getQPDF("unable to retrieve owning qpdf from foreign stream"); | |
| 2534 | - | |
| 2535 | - auto stream = foreign.getObjectPtr()->as<QPDF_Stream>(); | |
| 2536 | - if (stream == nullptr) { | |
| 2537 | - throw std::logic_error("unable to retrieve underlying" | |
| 2538 | - " stream object from foreign stream"); | |
| 2539 | - } | |
| 2540 | - std::shared_ptr<Buffer> stream_buffer = stream->getStreamDataBuffer(); | |
| 2541 | - if ((foreign_stream_qpdf.m->immediate_copy_from) && (stream_buffer == nullptr)) { | |
| 2542 | - // Pull the stream data into a buffer before attempting the copy operation. Do it on the | |
| 2543 | - // source stream so that if the source stream is copied multiple times, we don't have to | |
| 2544 | - // keep duplicating the memory. | |
| 2545 | - QTC::TC("qpdf", "QPDF immediate copy stream data"); | |
| 2546 | - foreign.replaceStreamData( | |
| 2547 | - foreign.getRawStreamData(), | |
| 2548 | - old_dict.getKey("/Filter"), | |
| 2549 | - old_dict.getKey("/DecodeParms")); | |
| 2550 | - stream_buffer = stream->getStreamDataBuffer(); | |
| 2551 | - } | |
| 2552 | - std::shared_ptr<QPDFObjectHandle::StreamDataProvider> stream_provider = | |
| 2553 | - stream->getStreamDataProvider(); | |
| 2554 | - if (stream_buffer.get()) { | |
| 2555 | - QTC::TC("qpdf", "QPDF copy foreign stream with buffer"); | |
| 2556 | - result.replaceStreamData( | |
| 2557 | - stream_buffer, dict.getKey("/Filter"), dict.getKey("/DecodeParms")); | |
| 2558 | - } else if (stream_provider.get()) { | |
| 2559 | - // In this case, the remote stream's QPDF must stay in scope. | |
| 2560 | - QTC::TC("qpdf", "QPDF copy foreign stream with provider"); | |
| 2561 | - m->copied_stream_data_provider->registerForeignStream(local_og, foreign); | |
| 2562 | - result.replaceStreamData( | |
| 2563 | - m->copied_streams, dict.getKey("/Filter"), dict.getKey("/DecodeParms")); | |
| 2564 | - } else { | |
| 2565 | - auto foreign_stream_data = std::make_shared<ForeignStreamData>( | |
| 2566 | - foreign_stream_qpdf.m->encp, | |
| 2567 | - foreign_stream_qpdf.m->file_sp, | |
| 2568 | - foreign.getObjGen(), | |
| 2569 | - stream->getParsedOffset(), | |
| 2570 | - stream->getLength(), | |
| 2571 | - dict); | |
| 2572 | - m->copied_stream_data_provider->registerForeignStream(local_og, foreign_stream_data); | |
| 2573 | - result.replaceStreamData( | |
| 2574 | - m->copied_streams, dict.getKey("/Filter"), dict.getKey("/DecodeParms")); | |
| 2575 | - } | |
| 2576 | -} | |
| 2577 | - | |
| 2578 | -void | |
| 2579 | -QPDF::swapObjects(int objid1, int generation1, int objid2, int generation2) | |
| 2580 | -{ | |
| 2581 | - swapObjects(QPDFObjGen(objid1, generation1), QPDFObjGen(objid2, generation2)); | |
| 2582 | -} | |
| 2583 | - | |
| 2584 | -void | |
| 2585 | 1783 | QPDF::swapObjects(QPDFObjGen const& og1, QPDFObjGen const& og2) |
| 2586 | 1784 | { |
| 2587 | 1785 | // Force objects to be read from the input source if needed, then swap them in the cache. |
| ... | ... | @@ -2590,92 +1788,6 @@ QPDF::swapObjects(QPDFObjGen const& og1, QPDFObjGen const& og2) |
| 2590 | 1788 | m->obj_cache[og1].object->swapWith(m->obj_cache[og2].object); |
| 2591 | 1789 | } |
| 2592 | 1790 | |
| 2593 | -unsigned long long | |
| 2594 | -QPDF::getUniqueId() const | |
| 2595 | -{ | |
| 2596 | - return m->unique_id; | |
| 2597 | -} | |
| 2598 | - | |
| 2599 | -std::string | |
| 2600 | -QPDF::getFilename() const | |
| 2601 | -{ | |
| 2602 | - return m->file->getName(); | |
| 2603 | -} | |
| 2604 | - | |
| 2605 | -PDFVersion | |
| 2606 | -QPDF::getVersionAsPDFVersion() | |
| 2607 | -{ | |
| 2608 | - int major = 1; | |
| 2609 | - int minor = 3; | |
| 2610 | - int extension_level = getExtensionLevel(); | |
| 2611 | - | |
| 2612 | - std::regex v("^[[:space:]]*([0-9]+)\\.([0-9]+)"); | |
| 2613 | - std::smatch match; | |
| 2614 | - if (std::regex_search(m->pdf_version, match, v)) { | |
| 2615 | - major = QUtil::string_to_int(match[1].str().c_str()); | |
| 2616 | - minor = QUtil::string_to_int(match[2].str().c_str()); | |
| 2617 | - } | |
| 2618 | - | |
| 2619 | - return {major, minor, extension_level}; | |
| 2620 | -} | |
| 2621 | - | |
| 2622 | -std::string | |
| 2623 | -QPDF::getPDFVersion() const | |
| 2624 | -{ | |
| 2625 | - return m->pdf_version; | |
| 2626 | -} | |
| 2627 | - | |
| 2628 | -int | |
| 2629 | -QPDF::getExtensionLevel() | |
| 2630 | -{ | |
| 2631 | - int result = 0; | |
| 2632 | - QPDFObjectHandle obj = getRoot(); | |
| 2633 | - if (obj.hasKey("/Extensions")) { | |
| 2634 | - obj = obj.getKey("/Extensions"); | |
| 2635 | - if (obj.isDictionary() && obj.hasKey("/ADBE")) { | |
| 2636 | - obj = obj.getKey("/ADBE"); | |
| 2637 | - if (obj.isDictionary() && obj.hasKey("/ExtensionLevel")) { | |
| 2638 | - obj = obj.getKey("/ExtensionLevel"); | |
| 2639 | - if (obj.isInteger()) { | |
| 2640 | - result = obj.getIntValueAsInt(); | |
| 2641 | - } | |
| 2642 | - } | |
| 2643 | - } | |
| 2644 | - } | |
| 2645 | - return result; | |
| 2646 | -} | |
| 2647 | - | |
| 2648 | -QPDFObjectHandle | |
| 2649 | -QPDF::getTrailer() | |
| 2650 | -{ | |
| 2651 | - return m->xref_table.trailer(); | |
| 2652 | -} | |
| 2653 | - | |
| 2654 | -QPDFObjectHandle | |
| 2655 | -QPDF::getRoot() | |
| 2656 | -{ | |
| 2657 | - QPDFObjectHandle root = m->xref_table.trailer().getKey("/Root"); | |
| 2658 | - if (!root.isDictionary()) { | |
| 2659 | - throw damagedPDF("", 0, "unable to find /Root dictionary"); | |
| 2660 | - } else if ( | |
| 2661 | - // Check_mode is an interim solution to request #810 pending a more comprehensive review of | |
| 2662 | - // the approach to more extensive checks and warning levels. | |
| 2663 | - m->check_mode && !root.getKey("/Type").isNameAndEquals("/Catalog")) { | |
| 2664 | - warn(damagedPDF("", 0, "catalog /Type entry missing or invalid")); | |
| 2665 | - root.replaceKey("/Type", "/Catalog"_qpdf); | |
| 2666 | - } | |
| 2667 | - return root; | |
| 2668 | -} | |
| 2669 | - | |
| 2670 | -std::map<QPDFObjGen, QPDFXRefEntry> | |
| 2671 | -QPDF::getXRefTable() | |
| 2672 | -{ | |
| 2673 | - if (!m->xref_table.initialized()) { | |
| 2674 | - throw std::logic_error("QPDF::getXRefTable called before parsing."); | |
| 2675 | - } | |
| 2676 | - return m->xref_table.as_map(); | |
| 2677 | -} | |
| 2678 | - | |
| 2679 | 1791 | size_t |
| 2680 | 1792 | QPDF::tableSize() |
| 2681 | 1793 | { |
| ... | ... | @@ -2806,192 +1918,3 @@ QPDF::getCompressibleObjGens() |
| 2806 | 1918 | |
| 2807 | 1919 | return result; |
| 2808 | 1920 | } |
| 2809 | - | |
| 2810 | -bool | |
| 2811 | -QPDF::pipeStreamData( | |
| 2812 | - std::shared_ptr<EncryptionParameters> encp, | |
| 2813 | - std::shared_ptr<InputSource> file, | |
| 2814 | - QPDF& qpdf_for_warning, | |
| 2815 | - QPDFObjGen const& og, | |
| 2816 | - qpdf_offset_t offset, | |
| 2817 | - size_t length, | |
| 2818 | - QPDFObjectHandle stream_dict, | |
| 2819 | - Pipeline* pipeline, | |
| 2820 | - bool suppress_warnings, | |
| 2821 | - bool will_retry) | |
| 2822 | -{ | |
| 2823 | - std::unique_ptr<Pipeline> to_delete; | |
| 2824 | - if (encp->encrypted) { | |
| 2825 | - decryptStream(encp, file, qpdf_for_warning, pipeline, og, stream_dict, to_delete); | |
| 2826 | - } | |
| 2827 | - | |
| 2828 | - bool attempted_finish = false; | |
| 2829 | - try { | |
| 2830 | - file->seek(offset, SEEK_SET); | |
| 2831 | - auto buf = std::make_unique<char[]>(length); | |
| 2832 | - if (auto read = file->read(buf.get(), length); read != length) { | |
| 2833 | - throw damagedPDF(*file, "", offset + toO(read), "unexpected EOF reading stream data"); | |
| 2834 | - } | |
| 2835 | - pipeline->write(buf.get(), length); | |
| 2836 | - attempted_finish = true; | |
| 2837 | - pipeline->finish(); | |
| 2838 | - return true; | |
| 2839 | - } catch (QPDFExc& e) { | |
| 2840 | - if (!suppress_warnings) { | |
| 2841 | - qpdf_for_warning.warn(e); | |
| 2842 | - } | |
| 2843 | - } catch (std::exception& e) { | |
| 2844 | - if (!suppress_warnings) { | |
| 2845 | - QTC::TC("qpdf", "QPDF decoding error warning"); | |
| 2846 | - qpdf_for_warning.warn( | |
| 2847 | - // line-break | |
| 2848 | - damagedPDF( | |
| 2849 | - *file, | |
| 2850 | - "", | |
| 2851 | - file->getLastOffset(), | |
| 2852 | - ("error decoding stream data for object " + og.unparse(' ') + ": " + | |
| 2853 | - e.what()))); | |
| 2854 | - if (will_retry) { | |
| 2855 | - qpdf_for_warning.warn( | |
| 2856 | - // line-break | |
| 2857 | - damagedPDF( | |
| 2858 | - *file, | |
| 2859 | - "", | |
| 2860 | - file->getLastOffset(), | |
| 2861 | - "stream will be re-processed without filtering to avoid data loss")); | |
| 2862 | - } | |
| 2863 | - } | |
| 2864 | - } | |
| 2865 | - if (!attempted_finish) { | |
| 2866 | - try { | |
| 2867 | - pipeline->finish(); | |
| 2868 | - } catch (std::exception&) { | |
| 2869 | - // ignore | |
| 2870 | - } | |
| 2871 | - } | |
| 2872 | - return false; | |
| 2873 | -} | |
| 2874 | - | |
| 2875 | -bool | |
| 2876 | -QPDF::pipeStreamData( | |
| 2877 | - QPDFObjGen const& og, | |
| 2878 | - qpdf_offset_t offset, | |
| 2879 | - size_t length, | |
| 2880 | - QPDFObjectHandle stream_dict, | |
| 2881 | - Pipeline* pipeline, | |
| 2882 | - bool suppress_warnings, | |
| 2883 | - bool will_retry) | |
| 2884 | -{ | |
| 2885 | - return pipeStreamData( | |
| 2886 | - m->encp, | |
| 2887 | - m->file_sp, | |
| 2888 | - *this, | |
| 2889 | - og, | |
| 2890 | - offset, | |
| 2891 | - length, | |
| 2892 | - stream_dict, | |
| 2893 | - pipeline, | |
| 2894 | - suppress_warnings, | |
| 2895 | - will_retry); | |
| 2896 | -} | |
| 2897 | - | |
| 2898 | -bool | |
| 2899 | -QPDF::pipeForeignStreamData( | |
| 2900 | - std::shared_ptr<ForeignStreamData> foreign, | |
| 2901 | - Pipeline* pipeline, | |
| 2902 | - bool suppress_warnings, | |
| 2903 | - bool will_retry) | |
| 2904 | -{ | |
| 2905 | - if (foreign->encp->encrypted) { | |
| 2906 | - QTC::TC("qpdf", "QPDF pipe foreign encrypted stream"); | |
| 2907 | - } | |
| 2908 | - return pipeStreamData( | |
| 2909 | - foreign->encp, | |
| 2910 | - foreign->file, | |
| 2911 | - *this, | |
| 2912 | - foreign->foreign_og, | |
| 2913 | - foreign->offset, | |
| 2914 | - foreign->length, | |
| 2915 | - foreign->local_dict, | |
| 2916 | - pipeline, | |
| 2917 | - suppress_warnings, | |
| 2918 | - will_retry); | |
| 2919 | -} | |
| 2920 | - | |
| 2921 | -// Throw a generic exception when we lack context for something more specific. New code should not | |
| 2922 | -// use this. This method exists to improve somewhat from calling assert in very old code. | |
| 2923 | -void | |
| 2924 | -QPDF::stopOnError(std::string const& message) | |
| 2925 | -{ | |
| 2926 | - throw damagedPDF("", message); | |
| 2927 | -} | |
| 2928 | - | |
| 2929 | -// Return an exception of type qpdf_e_damaged_pdf. | |
| 2930 | -QPDFExc | |
| 2931 | -QPDF::damagedPDF( | |
| 2932 | - InputSource& input, std::string const& object, qpdf_offset_t offset, std::string const& message) | |
| 2933 | -{ | |
| 2934 | - return {qpdf_e_damaged_pdf, input.getName(), object, offset, message}; | |
| 2935 | -} | |
| 2936 | - | |
| 2937 | -// Return an exception of type qpdf_e_damaged_pdf. The object is taken from | |
| 2938 | -// m->last_object_description. | |
| 2939 | -QPDFExc | |
| 2940 | -QPDF::damagedPDF(InputSource& input, qpdf_offset_t offset, std::string const& message) | |
| 2941 | -{ | |
| 2942 | - return damagedPDF(input, m->last_object_description, offset, message); | |
| 2943 | -} | |
| 2944 | - | |
| 2945 | -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file. | |
| 2946 | -QPDFExc | |
| 2947 | -QPDF::damagedPDF(std::string const& object, qpdf_offset_t offset, std::string const& message) | |
| 2948 | -{ | |
| 2949 | - return {qpdf_e_damaged_pdf, m->file->getName(), object, offset, message}; | |
| 2950 | -} | |
| 2951 | - | |
| 2952 | -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file and the | |
| 2953 | -// offset from .m->file->getLastOffset(). | |
| 2954 | -QPDFExc | |
| 2955 | -QPDF::damagedPDF(std::string const& object, std::string const& message) | |
| 2956 | -{ | |
| 2957 | - return damagedPDF(object, m->file->getLastOffset(), message); | |
| 2958 | -} | |
| 2959 | - | |
| 2960 | -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file and the object | |
| 2961 | -// from .m->last_object_description. | |
| 2962 | -QPDFExc | |
| 2963 | -QPDF::damagedPDF(qpdf_offset_t offset, std::string const& message) | |
| 2964 | -{ | |
| 2965 | - return damagedPDF(m->last_object_description, offset, message); | |
| 2966 | -} | |
| 2967 | - | |
| 2968 | -// Return an exception of type qpdf_e_damaged_pdf. The filename is taken from m->file, the object | |
| 2969 | -// from m->last_object_description and the offset from m->file->getLastOffset(). | |
| 2970 | -QPDFExc | |
| 2971 | -QPDF::damagedPDF(std::string const& message) | |
| 2972 | -{ | |
| 2973 | - return damagedPDF(m->last_object_description, m->file->getLastOffset(), message); | |
| 2974 | -} | |
| 2975 | - | |
| 2976 | -bool | |
| 2977 | -QPDF::everCalledGetAllPages() const | |
| 2978 | -{ | |
| 2979 | - return m->ever_called_get_all_pages; | |
| 2980 | -} | |
| 2981 | - | |
| 2982 | -bool | |
| 2983 | -QPDF::everPushedInheritedAttributesToPages() const | |
| 2984 | -{ | |
| 2985 | - return m->ever_pushed_inherited_attributes_to_pages; | |
| 2986 | -} | |
| 2987 | - | |
| 2988 | -void | |
| 2989 | -QPDF::removeSecurityRestrictions() | |
| 2990 | -{ | |
| 2991 | - auto root = getRoot(); | |
| 2992 | - root.removeKey("/Perms"); | |
| 2993 | - auto acroform = root.getKey("/AcroForm"); | |
| 2994 | - if (acroform.isDictionary() && acroform.hasKey("/SigFlags")) { | |
| 2995 | - acroform.replaceKey("/SigFlags", QPDFObjectHandle::newInteger(0)); | |
| 2996 | - } | |
| 2997 | -} | ... | ... |