Add missing dependency on cuda-cupti for CUDA 12 jaxlib builds

[traversaro]

Fix for:

• run_exports missing in cuda-cupti-dev 12.0.* cuda-cupti-feedstock#14
• Presence or absence of run_exports in cuda-cupti-dev cuda-cupti-feedstock#9

Checklist

• Used a static YAML file for the patch if possible (instructions).
• Only wrote code directly into generate_patch_json.py if absolutely necessary.
• Ran pre-commit run -a and ensured all files pass the linting checks.
• Ran python show_diff.py and posted the output as part of the PR.
• Modifications won't affect packages built in the future.

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[traversaro]

python show_diff.py (not sure why it contains clang-related changes):

(repodata) traversaro@IITBMP014LW012:~/conda-forge-repodata-patches-feedstock/recipe$ python3 show_diff.py
================================================================================
================================================================================
linux-armv7l
================================================================================
================================================================================
win-32
================================================================================
================================================================================
osx-arm64
osx-arm64::clang-18.1.5-default_h095aff0_0.conda
+  "constrains": [
+    "clang-tools 18.1.5.*",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
+  ],
osx-arm64::clang-tools-18.1.5-default_h7c89ad4_0.conda
-    "clangdev 18.1.5"
+    "clang 18.1.5.*",
+    "clangdev 18.1.5",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
================================================================================
================================================================================
linux-ppc64le
linux-ppc64le::clang-18.1.5-default_h60175c1_0.conda
+  "constrains": [
+    "clang-tools 18.1.5.*",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
+  ],
linux-ppc64le::clang-tools-18.1.5-default_hf6035ce_0.conda
-    "clangdev 18.1.5"
+    "clang 18.1.5.*",
+    "clangdev 18.1.5",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
================================================================================
================================================================================
linux-aarch64
linux-aarch64::clang-18.1.5-default_h33a044b_0.conda
+  "constrains": [
+    "clang-tools 18.1.5.*",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
+  ],
linux-aarch64::clang-tools-18.1.5-default_hb368394_0.conda
-    "clangdev 18.1.5"
+    "clang 18.1.5.*",
+    "clangdev 18.1.5",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
================================================================================
================================================================================
noarch
================================================================================
================================================================================
win-64
win-64::clang-18.1.5-default_hb53fc94_0.conda
+  "constrains": [
+    "clang-tools 18.1.5.*",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
+  ],
win-64::clang-tools-18.1.5-default_h3a3e6c3_0.conda
-    "clangdev 18.1.5"
+    "clang 18.1.5.*",
+    "clangdev 18.1.5",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
================================================================================
================================================================================
osx-64
osx-64::clang-18.1.5-default_h69aec76_0.conda
+  "constrains": [
+    "clang-tools 18.1.5.*",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
+  ],
osx-64::clang-tools-18.1.5-default_h69aec76_0.conda
-    "clangdev 18.1.5"
+    "clang 18.1.5.*",
+    "clangdev 18.1.5",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
================================================================================
================================================================================
linux-64
linux-64::clang-18.1.5-default_h90ac42e_0.conda
+  "constrains": [
+    "clang-tools 18.1.5.*",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
+  ],
linux-64::clang-tools-18.1.5-default_h127d8a8_0.conda
-    "clangdev 18.1.5"
+    "clang 18.1.5.*",
+    "clangdev 18.1.5",
+    "llvm 18.1.5.*",
+    "llvm-tools 18.1.5.*",
+    "llvmdev 18.1.5.*"
linux-64::jaxlib-0.4.23-cuda120py312hc008a70_201.conda
linux-64::jaxlib-0.4.23-cuda120py311hfb00743_200.conda
linux-64::jaxlib-0.4.23-cuda120py312h6027bbc_201.conda
linux-64::jaxlib-0.4.23-cuda120py310h3cc97ca_200.conda
linux-64::jaxlib-0.4.23-cuda120py312hc008a70_200.conda
linux-64::jaxlib-0.4.23-cuda120py39h443a19e_201.conda
linux-64::jaxlib-0.4.23-cuda120py310h3cc97ca_201.conda
linux-64::jaxlib-0.4.23-cuda120py311hfb00743_201.conda
linux-64::jaxlib-0.4.23-cuda120py311h5a49573_201.conda
linux-64::jaxlib-0.4.23-cuda120py39hfad2e6e_200.conda
linux-64::jaxlib-0.4.23-cuda120py310hb2a6d54_201.conda
linux-64::jaxlib-0.4.23-cuda120py39hfad2e6e_201.conda
+    "cuda-cupti >=12.0.90,<13.0a0",

[traversaro]

I rebased on top of main, and now the diff is the expected one:

(repodata) traversaro@IITBMP014LW012:~/conda-forge-repodata-patches-feedstock/recipe$ python show_diff.py
================================================================================
================================================================================
linux-armv7l
================================================================================
================================================================================
win-32
================================================================================
================================================================================
osx-arm64
================================================================================
================================================================================
linux-ppc64le
================================================================================
================================================================================
linux-aarch64
================================================================================
================================================================================
noarch
================================================================================
================================================================================
win-64
================================================================================
================================================================================
osx-64
================================================================================
================================================================================
linux-64
linux-64::jaxlib-0.4.23-cuda120py312hc008a70_200.conda
linux-64::jaxlib-0.4.23-cuda120py310hb2a6d54_201.conda
linux-64::jaxlib-0.4.23-cuda120py311h5a49573_201.conda
linux-64::jaxlib-0.4.23-cuda120py311hfb00743_200.conda
linux-64::jaxlib-0.4.23-cuda120py311hfb00743_201.conda
linux-64::jaxlib-0.4.23-cuda120py39hfad2e6e_200.conda
linux-64::jaxlib-0.4.23-cuda120py312h6027bbc_201.conda
linux-64::jaxlib-0.4.23-cuda120py312hc008a70_201.conda
linux-64::jaxlib-0.4.23-cuda120py39hfad2e6e_201.conda
linux-64::jaxlib-0.4.23-cuda120py39h443a19e_201.conda
linux-64::jaxlib-0.4.23-cuda120py310h3cc97ca_201.conda
linux-64::jaxlib-0.4.23-cuda120py310h3cc97ca_200.conda
+    "cuda-cupti >=12.0.90,<13.0a0",

The PR is ready for review @conda-forge/core, thanks!

[traversaro]

fyi @conda-forge/jax @conda-forge/cuda-cupti if you have any comment.

[xhochy]

This looks good but shouldn't we also have a PR that add this dependency to the jaxlib recipe, too?

[traversaro]

This looks good but shouldn't we also have a PR that add this dependency to the jaxlib recipe, too?

In theory the right way to handle would be to add a run_exports to cuda-cupti, we already did that on 12.4 but 12.0 is actually used, see conda-forge/cuda-cupti-feedstock#14 . However, if we do not get feedback on conda-forge/cuda-cupti-feedstock#14 from @conda-forge/cuda-cupti probably we can just add the dependency on jaxlib itself.

[xhochy]

We should definitely add it immediately to jaxlib as I would like to continue fixing other things there and that would probably produce broken packages again.

[traversaro]

We should definitely add it immediately to jaxlib as I would like to continue fixing other things there and that would probably produce broken packages again.

Done: conda-forge/jaxlib-feedstock#251 . Or if you are working on another PR, feel free to add it in that PR.