Skip to content

Add CI dogfood test; only Windows for this PR #7653

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 15, 2021
Merged

Conversation

Mikolaj
Copy link
Member

@Mikolaj Mikolaj commented Sep 15, 2021

A part of the effort to prevent hiccups like #7649.

I've tested that it fails with 3.6, works after reverting the offending autoconf PR, once the new CI jobs are merged, I will test if the new autoconf fix passes with it.

I've also fixed some details in he previous CI jobs from the workflow and bumped cabal used for bootstrapping from 3.2 to 3.4 in order to make list-bin available, needed for the dogfooding jobs.

Citing IRC:

I will build master, wipe out store, build master again using the exe just obtained
this is an automatic replication of the test with which @jneira originally detected #7649
let's see if that's longer than current CIs
(that was Tamar's reservation about this idea, I think)

our windows.yml CI workflow builds network as part of all cabal deps (and cabal source itself), but it uses cabal 3.2.0.0 to do so
it takes 21m to do so (BTW, the script a bit broken, the --only-dependencies part is vacuus, strangelyl)
while the whole workflow takes 50min or more
so the dog-fooding workflow would be faster than some of our existing workflows
so, if run concurrently (not always the case), it would not increase the waiting time
(that's with GHC 8.10)

doh, I've just looked at the corresponding linux and macos CI script templates and the validate.sh script they run instead of the simple Windows commands and I have no idea what's going on there
in particular, I'm not going to amend them in this commit; on the bright side, perhaps they do test dogfooding, given how smart they look
[edit: actually, they probably don't, but they have autoconf tests that seem to catch at least some of autoconf-related breakage that doesn't break Windows CI]

strangely network seems to compile fine in CI with GHC 8.6.5
I also learned what the famed zinza is for --- it's for generating CI scripts from template

@Mikolaj Mikolaj force-pushed the master branch 5 times, most recently from 1fbe4d7 to 52de3fa Compare September 15, 2021 16:43
@Mikolaj Mikolaj changed the title Add CI dogfood test; only Windows and GHC 8.6.5 for now Add CI dogfood test; only Windows for this PR Sep 15, 2021
@Mikolaj Mikolaj force-pushed the master branch 3 times, most recently from 08e3845 to ab577da Compare September 15, 2021 19:42
Copy link
Member

@jneira jneira left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the script could be simplified using a matrix of ghc versions.
It could be a separate job which can use the cabal built in the actual one via upload/download artifacts

But those things can be done in new prs after this
Thanks!

@Mikolaj
Copy link
Member Author

Mikolaj commented Sep 15, 2021

I think the script could be simplified using a matrix of ghc versions.
It could be a separate job which can use the cabal built in the actual one via upload/download artifacts

But those things can be done in new prs after this

Yep, fully agreed, there is even a ticket about a part of what you said (#7655). :)

Edit: Also, we should have dogfooding for linux and macos as well, more than 2 GHCs tested for Windows and macos, script generating code moved out of the common code base and generally the CI needs a lot of work.

@Mikolaj Mikolaj added the merge me Tell Mergify Bot to merge label Sep 15, 2021
@Mikolaj Mikolaj removed merge me Tell Mergify Bot to merge attention: needs-review labels Sep 15, 2021
@Mikolaj Mikolaj merged commit 1f0a624 into haskell:master Sep 15, 2021
@Mikolaj Mikolaj mentioned this pull request Sep 15, 2021
3 tasks
@Mikolaj
Copy link
Member Author

Mikolaj commented Sep 16, 2021

@Mergifyio backport 3.6

@mergify
Copy link
Contributor

mergify bot commented Sep 16, 2021

Command backport 3.6: failure

No backport have been created
GitHub App like Mergify are not allowed to create pull request where .github/workflows is changed.

Mikolaj added a commit that referenced this pull request Sep 16, 2021
@Mikolaj
Copy link
Member Author

Mikolaj commented Sep 16, 2021

This has been backported to 3.6 via a direct push to the branch.

Mikolaj added a commit to Mikolaj/cabal that referenced this pull request Sep 18, 2021
I guess I mistakenly removed it when resolving a conflict in backport
of haskell#7653 in c3354c9,
because on master these omitted lines are absent.
BTW, this backport was not reviewed, by my mistake.
A good example why careful reviews are essential.
Mikolaj added a commit that referenced this pull request Sep 19, 2021
Bring back cabal-install-dev call removed by backport of #7653
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants