Created attachment 201996 [details] forum-mso-de-129184.docx: seemingly a test designed to test smart justify I ran across this beautiful example of a justified document that seems to be fairly simple - just using the Calibri font. However, our justification in 25.8 (and master) doesn't match the output I get from MS Word 2019. It started with this commit edb1e8f1a0fca2489b722877917283f616e3acd9 Author: László Németh on Mon Jun 2 19:54:43 2025 +0200 tdf#166113 sw smart justify: adjust algorithm for interoperability Reviewed-on: https://gerrit.libreoffice.org/c/core/+/186138 In my PDF comparison, both MS Word 2019 and LO bibisect were running on the same windows machine, so both used the identical Calibri font version. I found a few other examples that bisected to the same commit. I'll mention them here even though they may have a different root cause: -631-п от 15.12.2014 (18741424 v1).DOCX (attachment 123457 [details] from bug 98573) Calibri font: I focused on page 4, the line containing point "N g m". It should only be 1 line, not 2. -Le_Renard_Enjoue.docx (attachment 165596 [details] from bug 136833) Liberation Serif font: I focused on the first big paragraph on page 4 - which adds an extra line. Found by Collabora's mso-test.
Created attachment 201997 [details] forum-mso-en4-758124.docx: nice, clean document using Times New Roman font Several paragraphs show regressions. For example, should be 22 entries on the first page, not 21.
My guess: the last line should never be expanded?