Re: CVS commit: pkgsrc/www/webkit-gtk

[Leonardo Taccari <iamleot%gmail.com@localhost>]

Hello Jörg and the entire pkgsrc community,

"Joerg Sonnenberger" writes:
> [...]
> Log Message:
> Disable optimisations only when compiling for debugging.
> [...]
As discussed on #pkgsrc this will probably breaks www/webkit-gtk on
platforms that uses lang/gcc48 (at least NetBSD 6.1.* where I had
originally worked during the update from webkit1-gtk to the current
version). According to one of my commit message in pkgsrc-wip:

 revision 1.4
 date: 2014/05/25 21:32:22;  author: leot1990;  state: Exp;  lines: +2 -1
 Add --disable-optimizations and update TODO...
 Please note that --enable-optimizations is by default yes and, at least in
 pkgsrc, should not be:
  1. usually the user should decide her favourite CFLAGS
  2. --enable-optimizations also add -D_FORTIFY_SOURCE=2 without including
     -I/usr/include/ssp nor passing to the linker -lssp (seems)
 The (2) is more critical and you can reproduce it, at least with lang/gcc48,
 adding --enable-optimizations and watching that webkit-gtk will fail in the
 configure phase when checking for pthread... Then please see the config.log
 and you will find an error about <ssp.h>.

A possible solution to this problem (also suggested by Jörg) is to use
BUILDLINK_TRANSFORM.
I tryed to write a patch to avoid this regression but I need some
help... How it can be distinguished lang/gcc48 vs a native gcc
4.8.*? (USE_PKGSRC_GCC seems only a user settable variable or am I
missing something?)
So in pkgsrc's Makefile-ese with a pseudo-code part in the condition,
just after the CONFIGURE_ARGS in Makefile.common we should add:

 .include "../../mk/compiler.mk"
 .if !empty(PKGSRC_COMPILER:Mgcc) && XXX_GCC_IS_FROM_PKGSRC
 BUILDLINK_TRANSFORM+=   rm:-D_FORTIFY_SOURCE=2
 .endif


Sorry for the incorrectness of the patch but I tryed to write a possible
quick workaround in order to have it in pkgsrc-2015Q1.
Thank you for your attention!
Ciao,
L.