Use cabal-cache #4717
Use cabal-cache #4717
Conversation
newhoggy
force-pushed
the
newhoggy/use-cabal-cache
branch
17 times, most recently
from
22e09ef to
c94fe3e
Compare
| echo "cabal-store=$(dirname "$cabal_config_file")/store" | tee -a "$GITHUB_OUTPUT" | ||
| else | ||
| echo "cabal-store=C:\\cabal\\store" | tee -a "$GITHUB_OUTPUT" | ||
| fi |
There was a problem hiding this comment.
Unify the computation of the cabal store directory to a single step. This makes referencing the cabal store in later steps easier.
This also fixes the problem where on MacOS, the cabal-store=/home/runner/.cabal/store is incorrect. This should be /Users/runner/.cabal/store.
To avoid this problem in future, on Linux and MacOS, the step queries cabal for the some information to compute the location of the cabal store.
There was a problem hiding this comment.
Can you put this as a comment in the file?
| store-path: ${{ steps.cabal-store.outputs.cabal-store }} | ||
| threads: 16 | ||
| archive-uri: ${{ secrets.BINARY_CACHE_URI }} | ||
| skip: "${{ secrets.BINARY_CACHE_URI == '' }}" |
There was a problem hiding this comment.
This step syncs the cabal store with S3, an object store service (think a service for storing files).
The S3 bucket is set to s3://iohk.cache.haskellworks.io.
| threads: 16 | ||
| archive-uri: https://iohk.cache.haskellworks.io | ||
| skip: "${{ secrets.BINARY_CACHE_URI != '' }}" | ||
| enable-save: false |
There was a problem hiding this comment.
It's important to ensure that people who fork this repository can not only successfully build in CI by default, but also have meaning cabal store caching.
Because syncing with S3 requires credentials, we cannot rely on S3 for this. For this reason a https fallback is used. The https server mirrors the content of the S3 bucket. The https cabal store archive is read-only for security reasons.
Users who fork this repository who want to have a writable cabal store archive are encouraged to set up their own S3 bucket.
| @@ -288,6 +291,8 @@ jobs: | |||
| # - name: Setup tmate session | |||
| # if: ${{ failure() }} | |||
| # uses: mxschmitt/action-tmate@v3 | |||
| # with: | |||
| # limit-access-to-actor: true | |||
There was a problem hiding this comment.
Adding this to the commented out tmate step for security reasons in case someone decides to use it.
newhoggy
force-pushed
the
newhoggy/use-cabal-cache
branch
5 times, most recently
from
ecac7f4 to
54e5140
Compare
newhoggy
force-pushed
the
newhoggy/use-cabal-cache
branch
2 times, most recently
from
33dde0c to
f49bfd4
Compare
newhoggy
force-pushed
the
newhoggy/use-cabal-cache
branch
from
f49bfd4 to
0d75329
Compare
Use
cabal-cacheto cachecabalpackages.This is better than using
actions/cache@v2becausecabal-cacheis a fine grained cache that caches each built cabal package individually.This means that if any packages were built successfully in CI, even if CI fails, we can still upload those packages that successfully built. This is far superior to the situation we have now which is if CI fails we cache nothing.
This is especially useful for those occasions where we update the index state or change our package dependencies (including package dependency bounds) because currently for those cases, a failed build means we cache nothing which hurts iteration times very badly.
The change keeps the build time more reliably tight:
Compare this to the old caching method:
The PR also fixes a bug which mean that MacOS builds weren't cached at all (ie. the
89mMacOS build time in the old caching method).