Description: I have a drawing (probably originally made a long time ago) that opens fine up to LibO 6.4 but incorrectly in 7.0.x and 7.1 beta. The issue may actually be with the drawing and LibO up to 6.4 may just have been more tolerant. But please still check that there is no regression in LibO. Steps to Reproduce: Open the attached drawing both in LibO 6.4 and LibO 7.0 and 7.1. See the differences. Actual Results: Opened docs look different. The version obtained with LibO 7.x is wrong. Expected Results: LibO 7.x should open the sample doc just like LibO 6.4 Reproducible: Always User Profile Reset: No Additional Info: [Information automatically included from LibreOffice] Locale: en-US Module: DrawingDocument [Information guessed from browser] OS: Linux (All) OS is 64bit: yes
Created attachment 167620 [details] Test doc
Created attachment 167622 [details] Comparison before and after
Regression introduced by: https://cgit.freedesktop.org/libreoffice/core/commit/?id=bc886f523872d4f9845c188c7d525d72a1a60946 author Regina Henschel <rb.henschel@t-online.de> 2020-01-02 17:37:32 +0100 committer Regina Henschel <rb.henschel@t-online.de> 2020-01-17 18:03:38 +0100 commit bc886f523872d4f9845c188c7d525d72a1a60946 (patch) tree 8677487eb45e619a9dcbc135c9c934e1a8e201d4 parent 21a5c7f77dea3c880ee1808f5385205f53271338 (diff) tdf#98584 Correct import draw:transform values skewY and matrix Bisected with: bibisect-linux64-7.0 Adding Cc: to Regina Henschel
The document is faulty. It has an element <office:graphics> where it should be <office:drawing>. The different placed triangle is the element <draw:polygon> in the first <draw:g> element in content.xml. Problematic values are svg:width="-1.999cm" and svg:height="-1.599cm" and svg:viewBox="0 0 -2000 -1600". Width and height may not be negative in SVG. I cannot create such shape, tested older OOo and LO 3.5. Meta data tell, that the document was created with a Linux version of LO 3.4 in 2011-11-29. It might be, that Novell Go-oo had used these negative values. I suspect that the negative values were used to describe a 180 degree rotation. But I don't know: the center of rotation, order of this rotation in regard to other transformations, whether this was used for all kind of shapes, how skew was written by that version. So I see no obvious way how to repair such documents and keep the skew fixes. Open and save the file in LO 6.4. It repairs the document and then it will be OK in LO 7.1 too.
Thanks for the thorough analysis. Unfortunately, so much time has passed that I do not really remember how I made that drawing! In any case, it is good to know that it is not a regression, but a broken document. And to have documented a workaround in case someone encounters a similar issue.
Hi Regina, thanks for the analysis. Closing as RESOLVED NOTABUG