failure to load CDR file with large bitmap (libxml2 >= 2.9.0)

[su_v]

With libxml2 >= 2.9.0 loading of a CDR files with large embedded bitmap images fails in inkscape trunk with a parser error, for example:
  Entity: line 12: parser error : AttValue length too long

A sample CDR file can be downloaded from this page:
<http://www.openaccessweek.org/photo/oaweek-2013-poster>
Direct download link:
<https://www.dropbox.com/s/f9mn0d6f4df8wvm/OAWeek%202013.cdr>

With rev >= 12575, this parser error triggers a crash (so far confirmed for CDR files, possibly affects other file formats too):

Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: KERN_INVALID_ADDRESS at address: 0x0000000000000030
0x000000010007731c in SPDocument::getRoot (this=0x0) at document.h:125
125 SPRoot *getRoot() { return root; }
(gdb) bt
#0 0x000000010007731c in SPDocument::getRoot (this=0x0) at document.h:125
#1 0x000000010059927b in Inkscape::Extension::Internal::CdrInput::open (this=0x1084282c0, unnamed_arg=0x1084282d0, uri=0x143b3ab58 "/Users/su_v/Downloads/_img/_misc/OAWeek 2013.cdr") at cdr-input.cpp:260
#2 0x000000010057c253 in Inkscape::Extension::Input::open (this=0x1084282d0, uri=0x143b3ab58 "/Users/su_v/Downloads/_img/_misc/OAWeek 2013.cdr") at input.cpp:153
#3 0x0000000100576e14 in Inkscape::Extension::open (key=0x0, filename=0x143b3ab58 "/Users/su_v/Downloads/_img/_misc/OAWeek 2013.cdr") at system.cpp:117
#4 0x000000010010460f in sp_file_open (uri=@0x7fff5fbfe288, key=0x0, add_to_recent=true, replace_empty=true) at file.cpp:274
#5 0x000000010018dbc6 in sp_recent_open (recent_menu=0x1088480f0, unnamed_arg=0x0) at interface.cpp:718

Confirmed with r12708 on OS X 10.7.5 (libxml2 2.9.1, libcdr git master @b279f96).

Based on tests with archived builds (on OS X 10.7.5):
- crash not reproduced with rev <= 12571,
- crash reproduced with rev >= 12576,
the regression (crash instead of notification about failure to load) was likely introduced in
  Revision 12575: Fix adding viewBox to new documents.
  <http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/12575>

Note: AFAICT this regression does only occur with libxml2 >= 2.9.0 installed, which introduced default limits for attribute lengths:
<https://git.gnome.org/browse/libxml2/commit/?id=e17db9946c09af709d6b37c598b336b1d2ef18a5>

The fix likely requires two steps:
1) add 'XML_PARSE_HUGE' (or whatever parser option is required) to support long attributes used for embedded bitmap images
2) add checks to prevent crash if parsed document is not valid (?) - fixed in rev 12733.

[su_v (suv-lp)]

Attaching the output of cdr2xhtml (command line tool installed with libcdr which wraps the SVG output into an XHTML file): AFAIU it contains the same SVG output which is read in by Inkcape's internal CDR input (and which inkscape fails to load due to a parser error if using libxml2 2.9.x).

[su_v (suv-lp)]

> Revision 12733: add some nullptr checking, should fix crash (couldn't test)
> <http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/12733>

@Johan - crash fix confirmed with r12734 on OS X 10.7.5: inkscape no longer crashes it if fails to parse the input.

Remaining issue:
- add adequate XMl parser flags to allow long attribute values.

[Patrick Storz (ede123)]

Crashing issue seems to have been fixed already.

Fix for the underlying import failure due to long attribute length pushed in
http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/15654

Please test the change thoroughly. If no regressions are found I think this would be a good candidate for backporting to 0.92.x.

[jazzynico (jazzynico)]

Fix confirmed on Xubuntu 16.04, lp:inkscape rev. 15657. The file now loads correctly, with no console message.

[Patrick Storz (ede123)]

Committed to stable branch in
http://bazaar.launchpad.net/~inkscape.dev/inkscape/0.92.x/revision/15427