develooper Front page | perl.perl5.porters | Postings from August 2009

Re: 5.10.1-RC1 is released

From:
Christopher J. Madsen
Date:
August 15, 2009 03:28
Subject:
Re: 5.10.1-RC1 is released
Message ID:
4A866D6F.3040702@cjmweb.net
H.Merijn Brand wrote:
> On Sat, 15 Aug 2009 00:18:03 +0000 (UTC), "Christopher J. Madsen"
> <perl@cjmweb.net> wrote:
...
>> But as I pointed out before (5.10.1-RC1 build failure with PERL_UNICODE),
>> setting PERL_UNICODE can prevent Perl from building at all.  It doesn't just
>> make some tests fail.
>>
>> Ideally, make should unset PERL_UNICODE while building.  If that's impractical,
>> it would be nice if Configure warned about it.
> 
> But would it be useful/reliable? PERL_UNICODE can also be set AFTER
> Configure has finished (before make) and have the same disastrous
> effects. I think that /if/ a warning is to be issued, it should be
> during build time, and not during Configure time.

While somebody certainly could set PERL_UNICODE after running Configure,
I think that's unlikely.  It's normally set in a profile (either a
user's or system-wide).

Maybe the warning should be issued at make time, but I'm afraid it would
be easy to overlook.  You could make it an error, but that might be
overkill, as there are probably values of PERL_UNICODE that do not
prevent the build from succeeding.

Or perhaps we could just pass -C0 to miniperl every time it's called in
the Makefile.  (I'm assuming that overrides any value PERL_UNICODE might
have.)  That may be easier than altering the environment.

-- 
Chris Madsen                                          perl@cjmweb.net
  --------------------  http://www.cjmweb.net  --------------------




nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About