Crash test update

[Michael Stahl]

On 10.10.19 13:51, Michael Stahl wrote:
> On 04.10.19 11:03, Caolán McNamara wrote:
>> On Wed, 2019-10-02 at 01:01 +0000, Crashtest VM wrote:
>>> New crashtest update available at
>>> http://dev-builds.libreoffice.org/crashtest/eeac5b7a424c3a3be48b1c3583171ad97d1b19cd/ 
>>>
>>
>> The export of fdo45983-1.doc to odt has been asserting in the
>> crashtesting for a while.
>>
>> https://bugs.documentfoundation.org/attachment.cgi?id=56949 is the .doc
>> and it seems to have begun asserting(in lcl_RemoveFieldMarks) since...
>>
>> commit dc9a0b124272a6dc0a8d875d51b6f882e60a8a04
>> Author: László Németh <nemeth at numbertext.org>
>> Date:   Wed Aug 7 11:25:13 2019 +0200
>>
>>      tdf#127101 Change tracking: reject format at paragraph join
> 
> i'm afraid this can't work, you can't split a redline at arbitrary 
> positions because then the user can reject one of the parts and accept 
> the other... in this case one of the parts contains a 
> CH_TXT_ATR_FIELDSTART and another the corresponding CH_TXT_ATR_FIELDEND, 
> but either one redline must contain both of them or neither (see 
> https://gerrit.libreoffice.org/#/c/80070/ for how i'm planning to 
> prevent this on the editing operations but of course i didn't notice 
> this code when i wrote that...)

it turned out that in this particular document the problem was that the 
first redline that was created was already wrong, which is fixed with 
commit 954c158acb87eb0385c51fdfff38fff84f154051 "sw: WW8: fix 
crashtesting asserts on fdo45983-1.doc export to ODT" - but in general 
the splitting of delete redlines remains a bad idea.