[ghc-steering-committee] #571: -Wsevere, Shepherd: Adam (rec: accept)

[Moritz Angermann]

This all sounds to me like we are trying to fix PVP by adding errors. I'm
ok to declare
PVP bankruptcy. If it's too hard to be practically used, we need to find a
different solution.

But this problem to me sounds more like we rather need some automation?
What we
seem to essentially want to achieve is guarantee that component integration
is not
faulty? It all sounds to me like what we really want is an automated
solution that flags
packages that were broken by changes in their dependencies. And then be
trivially able
to address those (and open PRs). Though I guess you'd also need to deal with
version bounds, the inverse of missing-fields would be overspecified
fields, and thus
the library that depends on something that added more fields, now needs a
hard lower
bound after adding the fields?

https://github.com/ghc-proposals/ghc-proposals/pull/617 likely also would
benefit from
automation. Maybe reboot hackage matrix in some form? Can someone point me
in
the right direction?

Best,
 Moritz

On Mon, 9 Oct 2023 at 23:57, Joachim Breitner <mail at joachim-breitner.de>
wrote:

> Hi,
>
> Am Montag, dem 09.10.2023 um 16:53 +0100 schrieb Simon Marlow:
> > On Mon, 9 Oct 2023 at 15:22, Joachim Breitner <mail at joachim-breitner.de>
> wrote:
> > >
> > > That is my understanding, yes.  Oleg, who suggested this originally (I
> > > just wrote it up) says (see motivation section of the proposal)
> > >
> > > > not having -Wsevere=missing-methods by default essentially prevents
> > > > any (true) breakage assessment of adding new, non-defaulted members
> > > > to existing type-classes.
> >
> >
> > If the goal is to do a breakage assessment, couldn't you make the
> > change to your library and then build all of Hackage with `--ghc-
> > option=-Werror=severe`?
>
> Not if Hackage already fails with -Werror=severe even without the
> change under assessment. But we should ask Oleg on Github, I am just
> relaying what I thought his motivation was.
>
> Should we send this back for revision? It seems there is plenty of
> discussion going on here and on github. (Is there an equivalent of
> Wadler’s law about “compiler flags” instead of “whitespace”?)
>
> Cheers,
> Joachim
>
>
> --
> Joachim Breitner
>   mail at joachim-breitner.de
>   http://www.joachim-breitner.de/
>
> _______________________________________________
> ghc-steering-committee mailing list
> ghc-steering-committee at haskell.org
> https://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-steering-committee
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-steering-committee/attachments/20231010/8cdf91c7/attachment.html>