Opened 2 months ago
Closed 7 weeks ago
#19372 closed enhancement (fixed)
asymptote-2.88
| Reported by: | Bruce Dubbs | Owned by: | |
|---|---|---|---|
| Priority: | normal | Milestone: | 12.2 |
| Component: | BOOK | Version: | git |
| Severity: | normal | Keywords: | |
| Cc: |
Description (last modified by )
New minor version. Was 2.87, now 2.88
Change History (7)
comment:1 by , 2 months ago
| Owner: | changed from to |
|---|---|
| Status: | new → assigned |
comment:2 by , 2 months ago
This is rolling out into TL pretest at the moment, but with an extra patch re the incorrect dvisvgm optimization. At the moment I plan to wait for pretest to be updated, and then test with that before I look at the source. Presumably https://github.com/vectorgraphics/asymptote/commit/b2d7d303271d137e77886bd071749bbbbbbe9c58 but the comments above the change imply suggest it has no relevance to the problem I've seen.
comment:3 by , 8 weeks ago
The problem (unable to create an svg file) seems to be caused by libgs.so.10.0.02, 10.0.01 was fine.
comment:4 by , 7 weeks ago
Neither upstream (asymptote, dvisvgm) can replicate my issue - but has found what is at least a releated issue in dvisvgm, and from that Martin has opened an issue at GS looking for feedback https://bugs.ghostscript.com/show_bug.cgi?id=707645
comment:5 by , 7 weeks ago
| Description: | modified (diff) |
|---|---|
| Summary: | asymptote-2.87 → asymptote-2.88 |
Release notes for 2.88 -
Building against specific static libraries is supported again.
An ambiguity in statistics.h was resolved.
The output of dvisvgm is optimized by default.
PDF 1.5 output is now generated.
comment:6 by , 7 weeks ago
Very strange, but then most things about this package are strange.
With gs-10.02.1, dvisvgm-3.2.2, asymptote-2.87 my MWE (create svg from the example sphere.asy shipped in asy) worked. After updating to gs-10.03.0 that svg was missing a chunk in the bottom right corner, and produced an error or warning message from ghostscript.
With asymptote-2.88 the warning was gone, but the svg was still missing part of the sphere, and then overlaid with black bands, somewhat like putting a postage stamp through the mail.
Apparently that is because my graphics driver cannot cope. The solution is to pass '-maxtile 1024,768' on the commandline, specifying a size smaller than my monitor.
What is strange is that after fixing that, my existing wild and wonderful asy tests all (including creating an svg) continue to work without adding a switch like that.
comment:7 by , 7 weeks ago
| Resolution: | → fixed |
|---|---|
| Status: | assigned → closed |
Book updated in sha:r12.1-98-gb7ef5a12f5cb
Errata created for the commented dependency in 12.1.
I can take this once we open for 12.2
Release notes at SF now visible on the project page in firefox, and nicely formatted, oh joy! No idea if the incorrect dvisvgm option is what breaks all my example svg generation, probably not (I suspect it is a ghostscript issue).
Release Notes for Version 2.87
Templated imports were implemented.
An incorrect dvisvgm optimization is avoided.
SVG gradient shading offsets were fixed. Integer are now implicitly formatted with full precision.
Inferred assignment of void type is prohibited.
An LSP segmentation fault was fixed.
Repeated stripping of file extensions is avoided.
The --output-directory option of latexmk is supported.
The OSMesa compatibility profile is used for offscreen rendering.
The fitscreen toggle is fixed. The context menu in Xasy is now only opened for GUI-drawn items.
A -v3d option allows embedding V3D content within a PDF file.
A workaround was implemented for a MacOS libc++ parsing bug.
The documentation was updated.
Source: README, updated 2024-02-27