Front page | perl.perl5.porters |
Postings from March 2013
On Sat, Mar 23, 2013 at 01:58:30AM +0100, Mark Martinec wrote:
> - created a tiny program 0.c, all it does is includes the EXTERN.h and perl.h:
> #include "EXTERN.h"
> #include "perl.h"
> int main(int argc, char **argv) {}
>
> So here is what happens:
>
> $ cd ~/perl-5.16.2/
> $ cc -I. -std=c99 -pedantic ~/0.c
Thanks. With that code, I was able to bisect the change to this commit:
commit b290562ef436d5316a2f75513def7f4f18c4ef34
Author: Father Chrysostomos <sprout@cpan.org>
AuthorDate: Sat Jul 7 17:35:10 2012 -0700
Allow CVs to point to HEKs rather than GVs
This will allow named lexical subs to exist independent of GVs.
...
-#define CvGV(sv) (0+((XPVCV*)MUTABLE_PTR(SvANY(sv)))->xcv_gv)
+#define CvGV(sv) S_CvGV((CV *)(sv))
...
+PERL_STATIC_INLINE GV *
+S_CvGV(CV *sv)
+{
+ return CvNAMED(sv)
+ ? 0
+ : ((XPVCV*)MUTABLE_PTR(SvANY(sv)))->xcv_gv_u.xcv_gv;
+}
> So it seems to me that your statement:
> | In which case, I don't think perl has ever added PERL_GCC_PEDANTIC
> | to the command line or any include file.
> is the key here, but it just so happens that one can get away
> despite this with 5.16.* .
I think (but I'm willing to be convinced otherwise), based on the above
(fairly innocuous) commit, that the situation is that if a module wants to
build itself with -pendantic, then it needs to set PERL_GCC_PEDANTIC too;
the fact that it wasn't needed in the past was just luck; the commit above
introduced some inline function definitions that made the use of
brace-groups "active" even for code that didn't directly or indirectly use
macros such as CvGV.
--
In my day, we used to edit the inodes by hand. With magnets.
Thread Next