We install GHC using apt, so stack shouldn't install it, but should it
ever end up installing it anyway (version mismatch, version unavailable
in hvr/ghc ppa, …), we don't want it wasting valuable cache space.
GitHub Actions writes caches on the first miss and then never updates
them again. If the CI is used frequently, the caches never expire and
as they get old, become less useful.
To avoid this, force refreshing the caches once a month.
Related: 52751f47d0
When building with an LTS version that has exactly the dependencies we
need (X11-1.9.2), stack doesn't need to download the Hackage index. If
GitHub Actions cache locking chooses this job as the one that writes the
cache, then the "stack-pantry-Linux" cache entry stays empty, possibly
forever.
Force Hackage index update to prevent this from happening.
Related: d1a4820b55
These past months I've spent a lot of time working on xmonad¹ and I feel
like I've done a lot. This is, however, not sustainable long term. :-(
¹) primarily xmonad-contrib and the community in general
I'd like to try making my GitHub Sponsors profile a bit more visible,
hoping that would allow me to continue dedicating time to xmonad.
I know that the correct approach probably is for the xmonad project to
find a fiscal sponsor like the Software Freedom Conservancy or Open
Collective or something and accept donations as a project, and then
redistribute that to people, but I don't think the project has enough
momentum to do something as complicated as that. :-/
Related: https://github.com/xmonad/xmonad-contrib/pull/544
Since actions/virtual-environments#3268, the GHC PPA is no longer
enabled by default. :-(
(It would be better to use the haskell/actions/setup action, but it
insists on doing `cabal update` even if all we need is to install GHC
and Stack, which adds noticeable delay:
https://github.com/haskell/actions/issues/29)
pandoc's API changes often enough that distros like Debian were patching
our GenerateManpage.hs to work with their version of pandoc, and it
doesn't build against any Stackage LTS except the recently released
LTS-17. Also, building pandoc from source takes quite some time and
resources.
But for what benefit? We're not using any special pandoc functionality
whatsoever. It's just that it was all in Haskell and the entire build
was orchestrated by cabal, but is all that complexity and resource
consumption worth it? I think not.
(Frankly, this whole thing was just a massive waste of time as the help
text in Config.hs isn't generated at all, so we still need to do this
manually. And then, the default key-bindings in core are unlikely to
change ever again.)
Let's simplify this:
* drop all dependencies except base and just run it through runhaskell
* add a Makefile and GH Actions workflow that invokes this after push
* only ship the results in release tarball, not the scripts which are
considered our dev infrastructure
Closes: https://github.com/xmonad/xmonad/issues/283
Related: https://github.com/xmonad/xmonad/pull/260
Related: https://github.com/xmonad/xmonad/pull/261
Related: https://github.com/xmonad/xmonad/pull/265
Related: https://github.com/xmonad/xmonad/pull/266
Related: https://github.com/xmonad/xmonad/pull/267
Related: https://github.com/xmonad/xmonad/pull/268
* rename workflow to Stack
* test all branches (provides feedback before opening a PR)
* split into buildenv deps and stack build
* make the deps list more git-friendly; tweak the deps
* use stack from github environment (I've seen the curl fail intermittently)
* use `*.cabal` to minimize diff between our repos
Related: https://github.com/xmonad/X11/pull/75
Related: https://github.com/xmonad/xmonad/issues/283
Unfortunately, our CI builds have a hard-coded assumption that all kinds of
LTS-X builds can be performed with the same stack.yaml file, which is not true.
We need to use specialized build instructions for each version of the
environment we're building in, e.g. stack-lts-16.yaml, stack-nightly.yaml, and
so on. Making that change is not super hard, but it's not completely trivial
either because the hooks use a hash of the stack file to determine whether the
cash is still valid, etc., and I'm not entirely sure about the best way to
adapt that code to support multiple YAML files.
For the time being, the most important build IMHO is Stackage Nightly,
because (a) we really want to support the nightly snapshot since (b) it can
compile the genmanpage executable relatively easily.
