No PR for version update of gz-physics even if the new version is correctly listed in bot-versions output and no error is listed in version error

[traversaro]

At conda-forge/gz-sim-feedstock#65 we are experiencing something strange with the automatic updates that we are not sure how to further debug.

The project is https://github.com/conda-forge/gz-physics-feedstock/ . The latest version on conda-forge is 7.2.0, while there was an upstream tag for 7.3.0 (https://github.com/gazebosim/gz-physics/releases/tag/gz-physics7_7.3.0, unfortunatly the tag format gz-physics7_7.3.0 probably plays a role here).

By inspecting the bot-versions output, you can find the new version is correctly detected: (see https://github.com/regro/cf-scripts/actions/runs/11007210867/job/30562973811):

2024-09-24T05:29:51.8730580Z 2024-09-24 05:29:51,872 INFO     conda_forge_tick.update_upstream_versions || itr  2236 - eta  2226s: gz-physics - 7.2.0 -> 7.3.0

Also by manually triggering an update with conda-forge/gz-physics-feedstock#39, the correct 7.3.0 version is updated (see conda-forge/gz-physics-feedstock#40).

Furthermore, there is no error for gz-physics in https://conda-forge.org/status/#version .

However, it seems that no automatic PR to update the repository to 7.3.0 is opened by the bot.

Does anyone have some pointer to debug this issue further? Thanks a lot in advance!

[traversaro]

There is a similar issue for gz-sim that is at 8.3.0 and no PR is opened to update to 8.6.0 .

[traversaro]

Some interesting pointers:

• https://github.com/regro/cf-scripts/blob/7ea599ebf5f78bebf8df8f1d2223dd4776b50e4b/README.md#current-bot-jobs-and-structure

@zklaus was also really helpful on Element, thanks! : https://matrix.to/#/!SOyumkgPRWoXfQYIFH:matrix.org/$17273353865774xzwaO:matrix.org?via=matrix.org&via=gitter.im&via=ifisc.uib-csic.es

[traversaro]

Ok, indeed some gz- recipes (that all share the same structure and unusual style of tags [gz-<project><major>-<major>.<minor>.<patch>]) the update is working as expected:

• [bot-automerge] gz-cmake v4.0.0 conda-forge/gz-cmake-feedstock#25
• [bot-automerge] gz-math v8.0.0 conda-forge/gz-math-feedstock#19
• gz-utils v3.0.0 conda-forge/gz-utils-feedstock#8

What they have in common is that in cf-graph-countyfair they have the correct version number listed:

• https://github.com/regro/cf-graph-countyfair/blob/54145d34bd8b22d5e69795fa9bf208aeaaa4be36/version_pr_info/0/c/3/1/4/gz-cmake.json#L4
• https://github.com/regro/cf-graph-countyfair/blob/54145d34bd8b22d5e69795fa9bf208aeaaa4be36/version_pr_info/6/b/9/3/6/gz-math.json#L4
• https://github.com/regro/cf-graph-countyfair/blob/54145d34bd8b22d5e69795fa9bf208aeaaa4be36/version_pr_info/a/8/b/5/a/gz-utils.json#L4

The following package up in the stack is gz-msgs, that does not receive version update PRs, and in version_pr_info they have:

• https://github.com/regro/cf-graph-countyfair/blob/54145d34bd8b22d5e69795fa9bf208aeaaa4be36/version_pr_info/c/1/b/0/a/gz-msgs.json#L4

Note "new_version": "10_10.2.0" in place of the "new_version": "10.2.0" of working feedstocks. Note that also the new_version_attemps was wrong (i.e. 2_2.0.0) in the new_version_attempts also for gz-utils (see https://github.com/regro/cf-graph-countyfair/blob/54145d34bd8b22d5e69795fa9bf208aeaaa4be36/version_pr_info/a/8/b/5/a/gz-utils.json#L4), whiele for gz-math and gz-cmake it was always correct.

[traversaro]

The following package up in the stack is gz-msgs, that does not receive version update PRs, and in version_pr_info they have:

* https://github.com/regro/cf-graph-countyfair/blob/54145d34bd8b22d5e69795fa9bf208aeaaa4be36/version_pr_info/c/1/b/0/a/gz-msgs.json#L4

Interestingly, the latest run of bot-bot actually updated that file: https://github.com/regro/cf-graph-countyfair/blob/09a9304ce47629fffbd4c97e028e162ab202c761/version_pr_info/c/1/b/0/a/gz-msgs.json .

[traversaro]

I have no idea what changed, but now all the version update were unblocked, see for example conda-forge/gz-physics-feedstock#41 . The only one missing is gz-sim, where the "wrong" 9_9.0.0 version is shown: https://github.com/regro/cf-graph-countyfair/blob/48abfb1a9b2434bdedcf815be8dccbf123c9071d/version_pr_info/d/4/9/e/5/gz-sim.json#L4 , but this may be expected as the fix in the structure are in conda-forge/gz-sim-feedstock#64 that still needs to be merged. At this point, we can close the issue.