False positives in skip check (after merge on top of skipped commit)

[h-vetinari]

I've run into the following kind of situation a few times now, where a merge commit gets skipped if the last commit on main contained [ci skip] etc. The following is a shortened & annotated git log --decorate --oneline --graph --all from https://github.com/conda-forge/ctng-compiler-activation-feedstock/, where

*   ddc2d49 (up/main) Merge pull request #136 from minrk/isystem-not-enough     # CI ran
|\
| * 4e19877 add -I $PREFIX/include to FFLAGS
* |   3648ba7 Merge pull request #131 from h-vetinari/clang19                   # CI skipped!
|\ \
| * | 68e8776 (origin/clang19, clang19) readd libstdcxx-devel to clangxx_impl; fix typo
| * | 111e074 [...]
| |/
* / 777282f [ci skip] [skip ci] [cf admin skip] ***NO_CI*** admin migration RemoveAutomergeAndRerender
|/
*   9833f59 Merge pull request #135 from apmorton/am/gcc-14.2                   # CI ran

In particular, the merge after https://github.com/conda-forge/ctng-compiler-activation-feedstock/pull/131 seems to falsely determine that the job should be skipped. Presumably there's an issue with how skips get identified for merge commits, in that it wrongly picks up on the last non-merge commit on main (or something like that)?

CC @beckermr @jaimergp

[jaimergp]

I think @beckermr had removed the skip checks in main 🤔 See https://github.com/conda-forge/conda-smithy/pull/2110. It looks like that file has not been rerendered in a while, though.

[beckermr]

Yes please rerender and try again.

[h-vetinari]

That feedstock currently cannot be rerendered due to https://github.com/conda-forge/conda-smithy/issues/2100. I did see it elsewhere as well, but will definitely check if a rerender solves it when I next encounter it.