[darcs-users] darcs patch: use gzipFormat instead of GZip to work a... (and 4 more)

Ganesh Sittampalam ganesh at earth.li
Sun Aug 30 07:28:33 UTC 2009


On Sat, 29 Aug 2009, Jason Dagit wrote:

> Sat Aug 29 19:22:09 PDT 2009  Jason Dagit <dagit at codersbase.com>
>  * use gzipFormat instead of GZip to work around deprecation warning.

This is because zlib is deprecating the constructors for abstraction 
reasons; this change is obviously correct and is something I'd been 
meaning to get around to myself.

> Sat Aug 29 19:23:54 PDT 2009  Jason Dagit <dagit at codersbase.com>
>  * remove unused argument to readPrim

Looks good, simple refactoring.

> Sat Aug 29 19:24:37 PDT 2009  Jason Dagit <dagit at codersbase.com>
>  * minor clean up in TouchesFiles

This one isn't correct because the old code compared two paths allowing 
for the '/' separator between elements, whereas the new code will make a 
change to 'foo' be thought of as touching 'food' and vice-versa - 'foo' 
should only match against 'foo/d' etc.

> Sat Aug 29 19:25:48 PDT 2009  Jason Dagit <dagit at codersbase.com>
>  * silence a warning that happens when compiling witnesses

This is fine. I've also been meaning to suggest we should just drop 
OPTIONS_GHC from being used with -cpp/-XFoo, given that our minimum GHC is 
now 6.8 which should support all that.

> Sat Aug 29 19:26:19 PDT 2009  Jason Dagit <dagit at codersbase.com>
>  * enable -Werror for now
>  On the one had, shipping release code without treating warnings as
>  errors makes a lot of pragmatic sense.  On the other hand, during
>  development it would be nice if we notice and deal with warnings as
>  soon as possible.  Therefore, I would like to reenable the treatment
>  of warnings as errors until the next release.

This is a policy decision and I'm not sure I agree with it, because of the 
variability between GHC 6.8, 6.10 and in the future GHC 6.12. But if we 
don't actually add -Werror we should all do better at cleaning up warnings 
we do notice.

Ganesh


More information about the darcs-users mailing list