Open 3 pages DOCX attachment 71442 [details]. It's 2007 DOCX but same if resaved in MSO. LO opens 4 pages because 2nd page table is split to 3rd page. Upon click in title "Spørsmål for alle typer eiendommer", it comes back to 3 pages.
Bibisect in 6.0 for table split - click in title couldn't fix there. e5593290e3759d5cbe51a6a349d459651e3e6e81 is the first bad commit Author: Jenkins Build User <tdf@pollux.tdf> Date: Thu Jul 13 02:04:23 2017 +0200 source sha:25445d24cfa87522ee4c47e4aa7e6e816cdc9a36 previous source sha:87633bd4e3a1b37ccfe1ddf18f1f178b07092a0d commit 25445d24cfa87522ee4c47e4aa7e6e816cdc9a36 [log] author Miklos Vajna <vmiklos@collabora.co.uk> Tue Jul 11 12:44:56 2017 +0200 committer Miklos Vajna <vmiklos@collabora.co.uk> Tue Jul 11 15:23:58 2017 +0200 tree b4b88a2fb5c52f6a98fc527f1b65c41e0c7cb582 parent 87633bd4e3a1b37ccfe1ddf18f1f178b07092a0d [diff] tdf#109063 DOCX import: consider wrap space for multi-page floattables freedesktop.org doesn't work to check if this is really the only commit, but let's assume it is, I provided also the previous commit. CC: Miklos.
Another bibisect in 6.3 again for new table split, except this time click in title somehow joins the table. 4ea1db013d4b402c56758acc00182cf2702d31c0 is the first changed commit Author: Jenkins Build User <tdf@pollux.tdf> Date: Thu Sep 26 02:58:11 2019 +0200 source sha:b0c5bc47d0d170df1384dd48cee9291ce6044083 previous source sha:92267cdb1a45e1f40199136849feb692f7e06d0e author Justin Luth <justin_luth@sil.org> 2019-09-24 19:39:29 +0300 committer Xisco Faulí <xiscofauli@libreoffice.org> 2019-09-25 11:56:54 +0200 commit b0c5bc47d0d170df1384dd48cee9291ce6044083 (patch) tree 280e4bbbeb3306c9c36a62e9c756a151c5e77766 parent 92267cdb1a45e1f40199136849feb692f7e06d0e (diff) Revert "tdf#117988 writerfilter: IgnoreTabsAndBlanksForLineCalculation" This reverts LO 6.2 commit 49ddaad2f3ba4e17e1e41e94824fb94468d2b680. tdf#127617 proves it simply was not the correct solution. CC: Justin. So, we have two great devs ("usual suspects") here in this mess.
This will be some kind of layout bug. Both of these commits will just have helped this particular docx trigger a certain condition that doesn't flow well. Neither one of these is truly a regression. I don't notice anything particularly strange about this table/surrounding parargraphs. There are no "keep with next paragraphs" involved. The only strange thing is the first and last row are combined columns (gridSpan) - which I would expect to be irrelevant. A few judiciously placed page break would have helped this document immensely. And whatever is happening with those as-character shapes probably isn't helping either.
Not sure if bug 128437 (or duplicates are somehow related). Some behavior, different bibisect results. Maybe because of DOCX <-> ODT?