In NUT as the primary consumer of the dynamatrix, we have a fair amount of PRs that impact only documentation, so would benefit from spellcheck and probably docs builds, but not from a big matrix of compilations.
Likewise, if it were possible/easy to single out changes to shell scripts (is only a shellcheck sufficient to make a verdict?), not-interpreted data files, possibly verifiable things like systemd units and SMF manifest (svccfg validate is used) - all those can be short-circuited for much faster test runs.
Changes to recipes (Jenkinsfile, configure.ac, ci_build.sh, .m4, Makefile.am, .gitignore, ...) and actual sources/headers should cause the complete matrix check as it is now. Uncertain cases - probably also. Better safe than sorry.
In NUT as the primary consumer of the dynamatrix, we have a fair amount of PRs that impact only documentation, so would benefit from spellcheck and probably docs builds, but not from a big matrix of compilations.
Likewise, if it were possible/easy to single out changes to shell scripts (is only a shellcheck sufficient to make a verdict?), not-interpreted data files, possibly verifiable things like systemd units and SMF manifest (svccfg validate is used) - all those can be short-circuited for much faster test runs.
Changes to recipes (Jenkinsfile, configure.ac, ci_build.sh, .m4, Makefile.am, .gitignore, ...) and actual sources/headers should cause the complete matrix check as it is now. Uncertain cases - probably also. Better safe than sorry.