At this point -- not cleaning up the solution files / etc, only our actions.
Until the 32bit code itself is removed, we should still have an easy way to flip these back on.
We switched to appimage in an earlier commit. It makes more sense for us
to maintain a distribution agnostic packaging method than something
which is already forked by upstream anyways.
Remove unneeded quotation marks.
linux-workflow.yml:
Checkout Submodules:
Shorten --jobs to -j, use $(getconf _NPROCESSORS_ONLN).
Make restore-keys into one line.
Remove unneeded ./.
macos-workflow.yml:
Checkout Submodules:
Shorten --jobs to -j.
Use $(getconf _NPROCESSORS_ONLN) instead of 2.
Move HOMEBREW_NO_INSTALL_CLEANUP to env: and add HOMEBREW_NO_ANALYTICS.
Remove unneeded ./.
pr-triage.yml:
Make on: pull_request_target into one line.
Properly capitalize GitHub.
compile.sh:
Add warning when running outside of GitHub Actions CI.
Use $(getconf _NPROCESSORS_ONLN).
validate-vs-filters.ps1:
End of file newline.
windows-workflow.yml:
Rename config to configuration.
Opt out of PowerShell telemetry.
Checkout Submodules:
Shorten --jobs to -j, use $env:NUMBER_OF_PROCESSORS.
Remove unneeded shell from Verify VS Project Files.
setup-msbuild:
Use v1. (always the latest version)
Remove useless vs-version parameter.
Remove unneeded ./ and .\.
Add -p to mkdir.
.gitmodules:
Make submodules shallow.
build.sh:
Move parameters from shebang to set.
Consistent formatting for ifs.
Use $(getconf _NPROCESSORS_ONLN).
There doesn't seem to be an obvious way to detect a submodule change. If the TAG was put inside the `gitmodules` file that would be another thing, but it isn't. So after a new submodule is added, the cache is never evicted (for instance in this PR's case, where an existing submodule is bumped to a new commit)
Add PR title if applicable, limit the length of the file to avoid file-name length issues.
Also put all contents of the artifact into a folder with the same name to simplify extraction.
msbuild does not use the project in it's entirety to build the app. This means problems can slip through, so an easy solution is just to ensure the XML is well-formed. More sophisticated validation could be done with something like `xmllint` but seems overkill.
The `.git/modules` folder was not being initialized, which caused the post-cleanup of the `Checkout` step to flag an error at the end. Caching this folder and thus completely caching anything submodule related, should eliminate this warning.
Also allowed submodules to be pulled in parallel when they aren't cached.