octoleo/qpdf
2
1
Fork 0
mirror of https://github.com/qpdf/qpdf.git synced 2025-01-08 17:24:06 +00:00
Code Issues Releases Wiki Activity

Tidy Objects::read

Browse Source
This commit is contained in:
m-holger 2024-10-14 11:12:14 +01:00
parent 6408c5c364
commit 486e68f2e5
1 changed files with 55 additions and 56 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

55
libqpdf/QPDF_objects.cc
View File

@ -1489,15 +1489,17 @@ Objects::read(
} }
QPDFObjectHandle oh = read_object(description, og); QPDFObjectHandle oh = read_object(description, og);
if (!unresolved(og)) {
return oh;
}
if (unresolved(og)) {
// Store the object in the cache here so it gets cached whether we first know the offset or // Store the object in the cache here so it gets cached whether we first know the offset or
// whether we first know the object ID and generation (in which we case we would get here // whether we first know the object ID and generation (in which we case we would get here
// through resolve). // through resolve).
// Determine the end offset of this object before and after white space. We use these // Determine the end offset of this object before and after white space. We use these numbers
// numbers to validate linearization hint tables. Offsets and lengths of objects may imply // to validate linearization hint tables. Offsets and lengths of objects may imply the end of
// the end of an object to be anywhere between these values. // an object to be anywhere between these values.
qpdf_offset_t end_before_space = m->file->tell(); qpdf_offset_t end_before_space = m->file->tell();
// skip over spaces // skip over spaces
@ -1514,39 +1516,36 @@ Objects::read(
} }
qpdf_offset_t end_after_space = m->file->tell(); qpdf_offset_t end_after_space = m->file->tell();
if (skip_cache_if_in_xref && xref.type(og)) { if (skip_cache_if_in_xref && xref.type(og)) {
// Ordinarily, an object gets read here when resolved through xref table or stream. In // Ordinarily, an object gets read here when resolved through xref table or stream. In the
// the special case of the xref stream and linearization hint tables, the offset comes // special case of the xref stream and linearization hint tables, the offset comes from
// from another source. For the specific case of xref streams, the xref stream is read // another source. For the specific case of xref streams, the xref stream is read and loaded
// and loaded into the object cache very early in parsing. Ordinarily, when a file is // into the object cache very early in parsing. Ordinarily, when a file is updated by
// updated by appending, items inserted into the xref table in later updates take // appending, items inserted into the xref table in later updates take precedence over
// precedence over earlier items. In the special case of reusing the object number // earlier items. In the special case of reusing the object number previously used as the
// previously used as the xref stream, we have the following order of events: // xref stream, we have the following order of events:
// //
// * reused object gets loaded into the xref table // * reused object gets loaded into the xref table
// * old object is read here while reading xref streams // * old object is read here while reading xref streams
// * original xref entry is ignored (since already in xref table) // * original xref entry is ignored (since already in xref table)
// //
// It is the second step that causes a problem. Even though the xref table is correct in // It is the second step that causes a problem. Even though the xref table is correct in
// this case, the old object is already in the cache and so effectively prevails over // this case, the old object is already in the cache and so effectively prevails over the
// the reused object. To work around this issue, we have a special case for the xref // reused object. To work around this issue, we have a special case for the xref stream (via
// stream (via the skip_cache_if_in_xref): if the object is already in the xref stream, // the skip_cache_if_in_xref): if the object is already in the xref stream, don't cache what
// don't cache what we read here. // we read here.
// //
// It is likely that the same bug may exist for linearization hint tables, but the // It is likely that the same bug may exist for linearization hint tables, but the existing
// existing code uses end_before_space and end_after_space from the cache, so fixing // code uses end_before_space and end_after_space from the cache, so fixing that would
// that would require more significant rework. The chances of a linearization hint // require more significant rework. The chances of a linearization hint stream being reused
// stream being reused seems smaller because the xref stream is probably the highest // seems smaller because the xref stream is probably the highest object in the file and the
// object in the file and the linearization hint stream would be some random place in // linearization hint stream would be some random place in the middle, so I'm leaving that
// the middle, so I'm leaving that bug unfixed for now. If the bug were to be fixed, we // bug unfixed for now. If the bug were to be fixed, we could use !check_og in place of
// could use !check_og in place of skip_cache_if_in_xref. // skip_cache_if_in_xref.
QTC::TC("qpdf", "QPDF skipping cache for known unchecked object"); QTC::TC("qpdf", "QPDF skipping cache for known unchecked object");
} else {
xref.linearization_offsets(toS(og.getObj()), end_before_space, end_after_space);
update_table(og.getObj(), og.getGen(), oh.getObj());
}
}
return oh; return oh;
}
xref.linearization_offsets(toS(og.getObj()), end_before_space, end_after_space);
return {update_table(og.getObj(), og.getGen(), oh.getObj())};
} }
QPDFObject* QPDFObject*