bug#31726: Attempt to use git reference for haskell-mode

[ludo@gnu.org (Ludovic Courtès)]

Hello Jack,

Jack Hill <jackhill@jackhill.us> skribis:

> I've returned to working on the haskell-mode fixes for Emacs
> 26.1. However, I'm afraid that I'll need a bit more guidance about the
> best path forward to prepare a patch that is suitable for inclusion in
> Guix.
>
> When I left off, in addition to your comments, I was concerned with
> test failures, so I tried to instead to define the package using a git
> reference (patch below (I have not yet worked on the commit
> message). Unfortunately, more of the tests fail than before. I was
> able to get more of the tests to pass by adding ghc as an input. With
> test disabled, I am able to install the package and load it in Emacs.
>
> A breakdown of the number of failing test is as follows:
> 16.1 version with patch  -- 4
> git version without ghc -- 11
> git version with ghc -- 7
>
> I'm not sure which version to prefer, or how to proceed making that
> decision. Version 16.1 was released in September 2016. By moving to
> the git version, we would no longer need to run a substitution in
> haskell-code-conventions.el as that file has been removed upstream.

From what you’re saying and from the previous discussion, it seems clear
to me that we’d rather use a recent Git snapshot.

> Would it be acceptable to disable the tests?

If there are 7 failing tests, perhaps that indicates a real issue.
Could you take a look at the test log to see if this is something
benign?  Sometimes it’s just hard-coded /usr/bin file names or similar
that cause tests to fail, and that is easy to fix.

Now, if haskell-mode has been failing to build for 2+ months, an update
without tests is still preferable over the status quo.  So if you don’t
see an immediate solution to the test failures, I suppose we could go
ahead and apply the patch.

Thank you,
Ludo’.