Make replacement graphic management more atomicHEADmaster

Commit 8872f7121b4ae4dd0b51820366d3510a88f7aac2 (crashtesting: crash
on exporting kde274105-6.docx to .rtf, 2024-03-27) added some safety
code in EmbeddedObjectRef::GetReplacement. It mentioned, that there
are likely some bugs in the management of the graphic.

This tries to fix this management, avoiding the intermediate states,
and only changing the graphic when all the data is ready. This also
reverts the changes of the mentioned commit, obsoleted now; and of
commit 8780fa41dcd164af244742461f4e57a4bcf4c7a4 (svtools: fix lost
replacement grpahic when updating it via OLE fails, 2018-10-30);
but keeps commit 24231f2a336c50eac5e6a1621c57e60ebe1e1cf4 (svtools:
fix lost replacement non-rendered graphic when updating it fails,
2022-02-17).

This has revealed that the second part of unit test for tdf#143222
("Check export of embedded worksheet in slide"), introduced in
commit 92a407b7f90a98704a238c5ffa3a3491eaf3263a (tdf143222 Handle
alternate content of graphicData element., 2021-07-08), has never
really worked: the "pGraphic != nullptr" check would never fail;
in fact, that used to always return an empty graphic. The problem
was filed as tdf#163064, and the test was modified accordingly.

Commit 5d997c029e53c31a3651a08f5012645097cec48f (sw XHTML export:
improve dummy OLE object handling, 2018-08-30) made ReqIF export
handle missing replacement graphic. However, it had assumed that
SwOLENode::GetGraphic always returns a valid pointer even for the
missing data. That is fixed here in OutHTML_FrameFormatOLENodeGrf.

Other places, where the pointer was dereferenced unconditionally,
were fixed (keeping current behavior).

Change-Id: Ica97a691ecc11b856bdb003f89467ea3392684bd
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/173716
Tested-by: Jenkins
Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>

0 files changed, 0 insertions, 0 deletions