Re: [Guile-commits] GNU Guile branch, stable-2.0, updated. v2.0.6-97-ge8772a9

Re: [Guile-commits] GNU Guile branch, stable-2.0, updated. v2.0.6-97-ge8772a9

[Mark H Weaver]

Hi Ludovic,

"Ludovic Courtès" <ludo@gnu.org> writes:
> commit 9ee0455738f90086894d602075915d49a5044fb7
> Author: Ludovic Courtès <ludo@gnu.org>
> Date:   Sat Nov 24 00:16:14 2012 +0100
>
>     Turn on the `case' warnings in auto-compilation.
>     
>     * module/ice-9/boot-9.scm (%auto-compilation-options): Add
>       `duplicate-case-datum' and `bad-case-datum'.
>
> commit 5cd10307866e6e6c44cb46b366f71d4118fa6aed
> Author: Ludovic Courtès <ludo@gnu.org>
> Date:   Fri Nov 23 23:56:01 2012 +0100
>
>     Add tests for `-Wduplicate-case-datum' and `-Wbad-case-datum'.
>     
>     * test-suite/tests/tree-il.test (%opts-w-duplicate-case-datum,
>       %opts-w-bad-case-datum): New variables.
>       ("warnings")["duplicate-case-datum", "bad-case-datum"]: New tests.
>
> commit 679a35567dbb0a467e99f19d3f513fac28317f26
> Author: Ludovic Courtès <ludo@gnu.org>
> Date:   Fri Nov 23 23:41:32 2012 +0100
>
>     doc: Mention the `duplicate-case-datum' and `bad-case-datum' warnings.
>     
>     * doc/ref/api-evaluation.texi (Compilation): List the
>       `duplicate-case-datum' and `bad-case-datum' warnings.

Sorry for not noticing this earlier, but I should mention that although
I use some of the warnings infrastructure for reporting the
'duplicate-case-datum' and 'bad-case-datum' warnings, I never check for
those warning flags.  The warnings are reported unconditionally.

I had started to work on a patch set to make them conditional, but that
work was halted due to an unresolved disagreement about how warnings
should be specified.

  http://lists.gnu.org/archive/html/guile-devel/2012-02/msg00080.html

I felt, and continue to strongly feel, that we should not require the
user to provide a complete list of warning types that they want.  If we
do that, then users will be forced to hard-code that list into their
build systems (and/or code that uses 'compile').  If they do this, then
whenever we add a new warning type, no one will see the new warnings
until they modify their build system.

Can we revisit this issue?

   Regards,
     Mark

Re: [Guile-commits] GNU Guile branch, stable-2.0, updated. v2.0.6-97-ge8772a9

[Ludovic Courtès]

Hi,

Mark H Weaver <mhw@netris.org> skribis:

> Sorry for not noticing this earlier, but I should mention that although
> I use some of the warnings infrastructure for reporting the
> 'duplicate-case-datum' and 'bad-case-datum' warnings, I never check for
> those warning flags.  The warnings are reported unconditionally.

Oops, I had overlooked that.

> I had started to work on a patch set to make them conditional, but that
> work was halted due to an unresolved disagreement about how warnings
> should be specified.
>
>   http://lists.gnu.org/archive/html/guile-devel/2012-02/msg00080.html

As an aside, it’s really hard for me to deal with work that’s “halted”.
I’d rather resolve them quickly, than just let them be forgotten, and
eventually revive them.

> I felt, and continue to strongly feel, that we should not require the
> user to provide a complete list of warning types that they want.  If we
> do that, then users will be forced to hard-code that list into their
> build systems (and/or code that uses 'compile').  If they do this, then
> whenever we add a new warning type, no one will see the new warnings
> until they modify their build system.
>
> Can we revisit this issue?

Sure.  Can you reply to my last message in the thread?  :-)

Thanks,
Ludo’.