Conversation
|
The things I don't like in these fixes is that we risk to be changing things that did not need to be changed before if this issue was introduced by recent changes in modern mode (as I snif). That is why I said in the issue page that it would be good to check is this is not a newly created bug. A good indication of it is to see if a classic mode script shows/or not the bug |
Claude says it's an old issue:
|
joa-quim
left a comment
joa-quim
left a comment
There was a problem hiding this comment.
OK, I checked that Claude was right about the versions.
With a transparent fill, the -Sl text was painted via PSL_setcolor, which sets a global PSL_transp that is never undone, so -Gred@50 leaked into all later layers and subplot panels. Paint such text via fs instead (PSL_plottext mode 2), like other filled symbols, so the transparency stays inside V...U. This also fixes transparent text fills from segment headers being drawn black. Fixes #9234 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
I made a commit to solve the confict. |
The last command of the loop was a grep that exits 1 when the check passes, so gmtest's ERR trap flagged the test as failed even with an empty fail file. Use an if instead. Also match a global PSL_transp anywhere on the line and for any blend mode, since older PSL emitted it after the color (e.g., "1 0 0 C 0.5 0.5 /Normal PSL_transp"). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Now that the test/psxy/text_symbol_transparency.sh passed I will merge it. |
2 participants
Claude's:
Fixes #9234
Now I got this figure:
Written with Claude Opus 5.5.