Caolan
- debug apparent missing language support
- LanguageTool supported ones are not listed because they
are not “local” to be preloaded, so those are fine, but
some explanatory logs would be good
→ https://gerrit.libreoffice.org/c/core/+/173448 - But the thesauri we load should have no relation to what
languages spell-checking supports.
→ https://gerrit.libreoffice.org/c/core/+/173467
- LanguageTool supported ones are not listed because they
- ofz#71606 png ubsan issue
→ upstream libpng takes an interest and eventual fix as:
SECURITY [libpng16] bug fix: More corrections for cHRM checks by jbowler · Pull Request #597 · pnggroup/libpng · GitHub - language tool spell checking doesn’t restart after save
→ https://gerrit.libreoffice.org/c/core/+/173564 - tdf#141577 another explicitly set current frame as active frame
→ https://gerrit.libreoffice.org/c/core/+/173624 - tdf#162538 regression
→ https://gerrit.libreoffice.org/c/core/+/173616 - missing ‘i’ in fi ligature with Karla font
→ https://gerrit.libreoffice.org/c/core/+/173665 - read logs, theorize crash related to problem that
cool#9950 fixed, which isn’t in that version? - tdf#159759 investigate misplaced svg text, probably my regression,
unclear where it goes wrong yet though. - refactoring pr#9916 for easier review
Mike
- Inconsistent preview graphic supply from ole objects
- They suspected non-blocking request for the replacement graphic, and asked for a blocking one
- In fact, it’s about a transient “no graphic” state, after already having a graphic, while another thread is requesting an update, so parallel threads doing the same would fail
- Fixed in https://gerrit.libreoffice.org/c/core/+/173690
- Fixed a regression in https://gerrit.libreoffice.org/c/core/+/173697 (I failed to see Caolan’s 8872f7121b4ae4dd0b51820366d3510a88f7aac2 when worked on the previous change)
- tdf#152298
- My regression from long ago (6f5024de2e1a5cc533527e45b33d9a415467c48d), it was some customer’s problem with large fixed-height table cells causing layout loops
- I have at least two more still unresolved regressions from that
- Created minimal reproducer
- Discovered some more related problems in the process
- It all turned out to not be the problem in the regressing commit, just another corner case uncovered
- A series of patches landed (https://gerrit.libreoffice.org/c/core/+/173595, https://gerrit.libreoffice.org/c/core/+/173596, https://gerrit.libreoffice.org/c/core/+/173674, https://gerrit.libreoffice.org/c/core/+/173600, https://gerrit.libreoffice.org/c/core/+/173601, https://gerrit.libreoffice.org/c/core/+/173675)
- Some reviews for Jaume
Tomaz
New online slideshow presentation
- VCL / Drawinglayer slide rendering
- found a way to render paragraphs individually
- in ViewObjectContactRedirector we get access to Primitive2DDecompositionVisitior, which we can convert to Primitive2DContainer
- this means we can check what is about to be rendered and potentially modify the tree
- when rendering a SdrTextObj for animation we always render only the object, so there is no other object in there which would conflict
- the issue is that to access paragraphs we need to decompose the primitives and then rely that the decomposition is buffered for this to work
- this is the ugly part of the drawing layer primitives - they are not generally meant to be modified just created and then decomposed
- then we remove “unwanted” paragraphs (nicely separated in it’s own primitive TextHierarchyParagraphPrimitive2D)
- and luckily removing the paragraphs works in this case
- next: do it properly to render each paragraph in each layer