[Libreoffice-bugs] [Bug 129553] Crash replacing in a specific text

[bugzilla-daemon at bugs.documentfoundation.org]

https://bugs.documentfoundation.org/show_bug.cgi?id=129553

Mike Kaganski <mikekaganski at hotmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |michael.stahl at cib.de

--- Comment #3 from Mike Kaganski <mikekaganski at hotmail.com> ---
Michael: could you please take a look?

The SwTextFrame::MapModelToViewPos code with the assert that own node must be
passed to the function was added in
https://git.libreoffice.org/core/+/c5abcfdf0c2e98c8f3445a92dd2d71eab23a7268
"sw_redlinehide: add TextFrameIndex type & SwTextFrame mapping functions"
(2018-06-08).

The two calls to the function from ReplaceBackReferences were added in
https://git.libreoffice.org/core/+/5e81b966778d82692b4763d892b457186a7f269d
"sw_redlinehide_4b: FindText() adapt the replace part as well" (2018-12-18).
And the calls are guaranteed to pass a node other than the frame's node: they
are made for a PaM that must start at end of a previous paragraph, and end at
beginning of the next - for single "$" regex that is treated specially: see
[1].

Simply commenting out the assert "solves" the problem. And release builds don't
give crashes here (so I initially thought it's a regression in master, but now
I suspect it only happens in debug builds). Could you see if this is OK, or
does it actually violates some of your assumptions in the code?

[1] https://help.libreoffice.org/latest/en-US/text/shared/01/02100001.html

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-bugs/attachments/20191223/b7e3d7d2/attachment.htm>