Hi,

Thanks for all the work on error messages, this is important.

Whatever you do please remember that not only humans are recipients of
these messages. Recently GHC changed 'Warning' into 'warning' and even that
caused some issues:

https://github.com/haskell/haskell-mode/issues/1513


2017-06-03 9:22 GMT-07:00 Ben Gamari <[email protected]>:

> Alan & Kim Zimmerman <[email protected]> writes:
>
> > Thanks Ben, a great summary.  Is there a Wiki page for this? It feels
> like
> > it should be on one, so we can easily comment/update the individual
> points.
> >
> Here you are: https://ghc.haskell.org/trac/ghc/wiki/PrettyErrors
>
> In the interest of time I essentially just pasted my response. Feel free
> to hack it up to your heart's content.
>
> > In terms of the pretty-printer and its string type. Perhaps we could
> > backpackify it to use http://next.hackage.haskell.
> org:8080/package/str-sig,
> > and then specialise the GHC version to FastString etc.
> >
> Sounds plausible.
>
> Relatedly, I always found it a bit odd that we used
> FastString for Doc literals. Namely we have to pay the cost of
> hashing the string, despite the fact that we will never use the hash.
> It's likely a tiny effect (hashing is quite quick), but it still seems
> odd.
>
> Cheers,
>
> _______________________________________________
> ghc-devs mailing list
> [email protected]
> http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
>
>
_______________________________________________
ghc-devs mailing list
[email protected]
http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

Reply via email to