mhx-perl at gmx
Nov 7, 2009, 4:58 PM
Post #5 of 10
(284 views)
Permalink
|
|
Re: [PATCH] buildperl.pl: allow older perl to compile with newer gcc
[In reply to]
|
|
Hi Philippe,
On 2009-11-08, at 01:01:14 +0100, Philippe Bruhat (BooK) wrote:
> On Sat, Nov 07, 2009 at 07:26:00AM +0100, Marcus Holland-Moritz wrote:
> > >
> > > diff --git a/cpan/Devel-PPPort/devel/buildperl.pl b/cpan/Devel-PPPort/devel/buildperl.pl
> > > index 49b7fbb..fefd480 100644
> > > --- a/cpan/Devel-PPPort/devel/buildperl.pl
> > > +++ b/cpan/Devel-PPPort/devel/buildperl.pl
> > > @@ -99,7 +99,7 @@ my @patch = (
> > > },
> > > {
> > > perl => [
> > > - qr/^5\.004_0[1234]$/,
> > > + qr/^5\.004_0[12345]$/,
> > > ],
> > > subs => [
> > > [ \&patch_doio ],
> >
> > any special reason why you added 5.005_05 for the doio patch?
>
> My patch is for 5.004_05, not 5.005_05.
Yes. My bad. I meant 5.004_05, but my fingers dared not to
follow my brain ;)
> > While this might sound like a dumb question at first -- I guess
> > you added it because it didn't compile for you -- I'd be glad
> > for some details, as *with* your change, it doesn't compile for
> > me...
>
> Indeed, I added that because it didn't compile for me.
>
> First of all, I create the archives using
>
> git archive --format=tar --prefix=$tag/ $tag^{tree} \
> | gzip > /tmp/perl/source/$tag.tar.gz
>
> With $tag being perl-5.004_05 in this case.
>
> Maybe this tag doesn't generate the same archive as the one you're using?
Can you compare yours against the one from CPAN (which is
what I'm using):
http://www.cpan.org/src/perl5.004_05.tar.gz
> > `sh cflags libperl.a doio.o` doio.c
> > CCCMD = cc -DPERL_CORE -c -Dbool=char -DHAS_BOOL -DDEBUGGING -g
> > doio.c:86: error: redefinition of 'union semun'
> > make: *** [doio.o] Error 1
>
> The error I get *without* my change is:
>
> `sh cflags libperl.a doio.o` doio.c
> CCCMD = cc -DPERL_CORE -c -Dbool=char -DHAS_BOOL -I/usr/local/include -O2
> doio.c: In function ‘Perl_nextargv’:
> doio.c:505: attention : ignoring return value of ‘fchown’, declared with attribute warn_unused_result
> doio.c: In function ‘Perl_do_ipcctl’:
> doio.c:1445: erreur: storage size of ‘unsemds’ isn’t known
> doio.c:1486: erreur: storage size of ‘semun’ isn’t known
> make: *** [doio.o] Erreur 1
> make all: 512
>
> > ...which is because this piece of code had been added between
> > 5.005_04 and 5.005_05 in perl.h...
> >
> > # ifndef HAS_UNION_SEMUN /* Provide the union semun. */
> > union semun {
> > int val;
> > struct semid_ds *buf;
> > unsigned short *array;
> > };
> > # endif
> >
> > ...which was precisely why I excluded 5.004_05 from the doio
> > patch. So, maybe you can provide me with some details on why
> > exactly it doesn't work for you so we can find a solution
> > that works for both systems.
>
> When looking at the logs in the git tree, it seems that the patch
> you're talking about was added in bd89102fe7784367e3eeddc7cefae29defae903a:
>
> $ git log -n1 -p bd89102fe7784367e3eeddc7cefae29defae903a -- perl.h
> commit bd89102fe7784367e3eeddc7cefae29defae903a
> Author: Andy Dougherty <doughera [at] lafayette>
> Date: Thu May 28 09:27:25 1998 -0400
>
> [PATCH 5.004_65] Config_65-02-03.diff: SunOS and Solaris hints
> Date: Thu, 28 May 1998 13:27:25 -0400 (EDT)
> Subject: [PATCH 5.004_65] Config_65-03-04.diff: semctl probing
> Date: Thu, 28 May 1998 13:28:21 -0400 (EDT)
>
> p4raw-id: //depot/perl [at] 104
>
> diff --git a/perl.h b/perl.h
> index fc96064..8645c39 100644
> --- a/perl.h
> +++ b/perl.h
> @@ -2090,6 +2090,26 @@ enum {
> UNLOCK_SV_MUTEX; \
> } while (0)
>
> +#ifdef HAS_SEM
> +# include <sys/ipc.h>
> +# include <sys/sem.h>
> +# ifndef HAS_UNION_SEMUN /* Provide the union semun. */
> + union semun {
> + int val;
> + struct semid_ds *buf;
> + unsigned short *array;
> + };
> +# endif
> +# ifdef USE_SEMCTL_SEMUN
> +# define Semctl(id, num, cmd, semun) semctl(id, num, cmd, semun)
> +# else
> +# ifdef USE_SEMCTL_SEMID_DS
> +# define Semctl(id, num, cmd, semun) semctl(id, num, cmd, semun.buf)
> +# endif
> +# endif
> +# ifndef Semctl /* Place our bets on the semun horse. */
> +# define Semctl(id, num, cmd, semun) semctl(id, num, cmd, semun)
> +# endif
> +#endif
>
> #endif /* Include guard */
> -
>
> What worries me is that this patch is for Perl 5.004_65.
> But the only 5.004_xx tags in the git tree are:
>
> $ git tag -l 'perl-5.004*'
> perl-5.004
> perl-5.004_01
> perl-5.004_02
> perl-5.004_03
> perl-5.004_04
> perl-5.004_05
>
> Since the versions I see in the log messages go up to 5.004_71, it seems
> that some tags are missing...
IIRC, 5.004_50 and onwards were development releases (just
like 5.(7|9|11).x today). And just as 5.8.9 was released
after 5.9.5, for example, 5.004_05 was most probably
released after 5.004_65. (Indeed, checking in the APC I
can see a 5.004_65 tarball from 1998-05-15 while 5.004_05
is from 1999-04-29.)
The question is whether the 5.004_05 tag in the repo is
correct. I'd say it should match the contents of the
tarball that's available from CPAN.
> The best way to go with my patch is probably to ignore the parts related
> to 5.004_05, as obviously I didn't know what I was doing.
I indeed changed the behaviour of the doio patch back to
*not* patching 5.004_05. With this, I can build all perls
on my box.
Thanks for your work!
Cheers,
Marcus
--
Computer programmers do it byte by byte.
|
signature.asc
