WASM CI improvements
I looked at the definition of the wasm pipelines and it seems there are a few issues which need addressing:
- Why are the patches to
process
andhsc2hs
not merged into the upstream repos? - Why is the build environment provided by
ghc-wasm-meta
not provided statically in a docker image rather than fetched over the network? This introduces a large vector for failure which is not desirable on a pipeline which runs on every job. (See https://gitlab.haskell.org/ghc/ghc/-/jobs/1275577) - Why is the expiration of artifacts set to 1 year, when the standard for a validate pipeline is 2 weeks.
- Why is the CI job not generated by the same standard as all other platform jobs using
gen_ci.hs
. - Why is the trigger
when: always
, this doesn't align with how other validate/nightly/release jobs are configured. Another strong reason why these jobs need to be generated usinggen_ci.hs
.
Given the highly critical nature of the validate pipelines.. I am going to move the wasm pipelines into the nightly build and mark them as allowed to fail for now. When @TerrorJack returns he can pick up the task of making the pipelines robust enough to place into the normal validate workflow.