Conversation
…platforms not on Windows elm_playground_web uses js_of_ocaml directly (Typed_array) and vdom does not bring it: the package, and elm_playground_3d_web and _webgl on it, failed wherever nothing else had installed it. caps said by the two packages that use it directly. The packages implementing a virtual library (the platforms, and tinybox on them) do not build on Windows, as elm_playground_native already says: dune's check of the implementation fails there (ocamlobjinfo, 0xc0000135).
The lower-bounds job built elm_playground_3d_opengl with dune 3.10.0, which does not find the implementation of a virtual library among the installed packages (3.17.2 does). On Fedora, tiny_libs' TLS tests found no openssl program.
0.3.1 fixes what this CI found in 0.3.0 that the opam files could not: tinybox's build failing on a file gone while its generator walks the build tree (ar's temporary files), and tiny_libs' socket tests failing where a sandbox forbids sockets.
tsdl 1.3.0 binds SDL_RenderGetWindow (SDL 2.0.22), looked up when a program starts; Ubuntu 22.04 has an older SDL.
One of them failed on a shared builder (ppc64, OCaml 5.5): a ratio of two running times, asserted. The tests skip themselves when CI is set. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
I think the other errors seems to be all in "experimental" CI test, so they are probably ok to fail. |
|
I've fixed the issue in the CI test on the ppc64 build in mtelvers/day10@795b46c. This was a write/rename ordering issue. From what I can see, the s390 failure is a genuine cross-architecture difference in floating point. I see that you already use 1e-12. 1e-6 would clear the error, or accept the discrepancy on this hardware. The other CI failures are due to the tests waiting a fixed period for the service to start. e.g. |
What opam's CI found in 0.3.1: the socket tests wait for what they expect instead of a fixed 50 ms; a floating-point check that was off by 1.5e-7 on s390x is made to a millionth. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
This opam-ci is hard to satisfy ... |
|
tough guy |
|
ok the error seems outside this package now, a curl issue I have nothing to do with. |
|
@shonfeder what is the "needs reporter action" label? I think the remaining errors in opam-ci are ok no? We could merge this? |
| bug-reports: "https://github.com/aryx/ocaml-elm-playground/issues" | ||
| depends: [ | ||
| "ocaml" {>= "4.14.0"} | ||
| "dune" {>= "3.0" & >= "3.17"} |
There was a problem hiding this comment.
| "dune" {>= "3.0" & >= "3.17"} | |
| "dune" {>= "3.17"} |
I suggest cleaning up these double bounds - both here and in the other files. They arise because you both have a (lang dune ...) bound and an simultaneously list a dune dependency. I suggest removing the latter and regenerating the .opam-files, rather than me suggesting 14 manual edits or you doing 14 manual edits...
| # claude: not on Windows (as elm_playground_native): dune's check of a | ||
| # virtual library's implementation fails there (ocamlobjinfo exits with | ||
| # 0xc0000135, opam's Windows CI, 0.3.0); tiny_libs, tiny_languages, | ||
| # tiny_appkits, elm_playground and elm_playground_3d build on it |
There was a problem hiding this comment.
These Claude comments bug me a bit:
- No need to list
claude:as you are responsible for the PR and release - It mentions "a virtual library's implementation" but not which one, so it is not very informative ("oh sdl/gl/foobar doesn't work on Cygwin/MSys2") nor very actionable ("let me try to get sdl/gl/foobar working on Cygwin/MSys2")
- Repeatedly listing
tiny_libs, tiny_languages, tiny_appkits, elm_playground and elm_playground_3d build on itas a comment in several other packages doesn't add anything
🤷
There was a problem hiding this comment.
I forgot a fourth one (a nit I admit):
- They also mention version 0.3.0, but now the PR's release version is at 0.3.5.
Since the remark concerns the enclosing package version, I'd just omit it.
A comment inpackages/elm_playground_3d_opengl/elm_playground_3d_opengl.0.3.5/opamconcerns 0.3.5 🙂
Hi! Sorry for not being more explicit. It was because there were apparent CI tests failures in the packages at the time (which it seems you addressed after I added the label). This is just a tracking mechanism to coordinate between maintainers on what needs our attention. |
|
I think it's ready for merge. The only remaining error in opam-ci looks unrelated. |
This pull-request concerns:
elm_playground.0.3.0: Interface of the Playground: pictures, animations and gameselm_playground_3d.0.3.0: Interface of the 3D Playgroundelm_playground_3d_opengl.0.3.0: Native 3D playground platform on the GPU (OpenGL 3.3 core)elm_playground_3d_software.0.3.0: Native 3D playground platform with a from-scratch software rasterizerelm_playground_3d_web.0.3.0: Web 3D playground platform drawing through SVGelm_playground_3d_webgl.0.3.0: Web 3D playground platform on the GPU (WebGL 1)elm_playground_gamekits.0.3.0: Genre kits for Playground games: platformers, racing, shoot 'em ups, fighting andmore
elm_playground_native.0.3.0: Native playground platform: SDL window, Cairo drawingelm_playground_software.0.3.0: Native playground platform with a from-scratch software rasterizerelm_playground_web.0.3.0: Web playground platform: js_of_ocaml, vdom and SVGtiny_appkits.0.3.0: Application engines from scratch: a spreadsheet, rich text, paint, draw, CAD, editors and moretiny_languages.0.3.0: Small languages from scratch: Scheme, Lisp, Smalltalk-80, Pascal, BASIC, JavaScript, HTML,CSS and more
tiny_libs.0.3.0: From-scratch libraries for teaching: graphics, audio, compression, crypto, networking and moretinybox.0.3.0: Every game and app of elm-playground in one binary🐫 Pull-request generated by opam-publish v3.0.1