Apache OpenOffice (AOO) Bugzilla – Issue 6596
Crash when trying to open a document
Last modified: 2013-08-07 14:43:23 UTC
I've got a document saved with Office2000 and fast save option enabled that makes OpenOffice 1.0 and 1.0.1 crash during loading. The same document saved with "save as" under office, thus without fast save, is opened properly.
Created attachment 2365 [details] The file that makes OOwriter crash (compressed)
Sorry, it seems I'm wrong, the file always makes OOwriter crash, with or without fast save.
moving to component word processor
Reassigned to Michael.
MRU->DVO: doc seems to loop after import. The debug version throws tons of the Assertion "redline table corrupted: overlapping redlines" SW644MI! AssertFail(char const *,char const *,unsigned short) + 258 bytes SW644MI! lcl_CheckRedline(class SwDoc const *) + 255 bytes SW644MI! SwDoc::AppendRedline(class SwRedline *,unsigned char) + 6766 bytes SW644MI! SetInDocAndDelete::operator()(class SwFltStackEntry *) + 237 bytes SW644MI! SwWW8ImplReader::LoadDoc1(class SwPaM &,class WW8Glossary *) + 4442 bytes SW644MI! WW8Reader::Read(class SwDoc &,class SwPaM &,class String const &) + 337 bytes SW644MI! SwReader::Read(class Reader const &) + 709 bytes SW644MI! SwDocShell::ConvertFrom(class SfxMedium &) + 447 bytes SFX644MI! SfxObjectShell::DoLoad(class SfxMedium *) + 3480 bytes
dvo->cmc: "The wheels have fallen off revision mark import" There are several assertions when loading the bugdoc (in SRX644d), the 'wheels' one and one complaining about the date being 0. Later on, you will find (my) assertions about the redline table being corrupted, but I assume they are consequences of the former. If not, send the bug back to me once you're done with it. Note: The bugdoc is a zipped .doc file.
Created attachment 4628 [details] little example
The little attachment shows that word does have changes that have both no author and no time/date of change. So the asserts are reporting an unexpected event that is truly the case.
And as for the core asserts of overlapping revisions, there actually *are* overlapping revisions here. There are ranges of text which have both delete and append properties in use, i.e. text that claims to be deleted before it is added :-)
Created attachment 4644 [details] overlapping example
OK, the no 1 concern is crashing. And crashes can be done away with, so we'll do that here. My asserts about the redline having no author and no date of change are actually correct behaviour in the filter, not logic gone awry. The reality is that some of the redlines have an "unknown" author and display no time of change (happens I think sometimes when redlines are imported from word95/6 to 97/2000/xp), so I must remove those asserts. I can also check for totally invalid empty redlines, which I have done, so will be no crash in apps01 workspace for 1.1 Beta. But there will still be complaints of overlapping redlines, this is because they are truly overlapping in the winword file:-(, and that requires some deeper thinking, see issue 11450 for tracking the overlap issue.
reopen to reassign
cmc->mru: It will assert in a nonproduct version because of issue 11450. But it shouldn't crash.
Checked the fix with internal build CWS_apps01.
Fix will be available with OpenOffice 1.1 Beta.
Checked integration with internal Milestone srx644m4s3