[PATCH 1/1] gcc: libgcc-common.inc: symlink gthr-posix -> gthr-default.h

[PATCH 1/1] gcc: libgcc-common.inc: symlink gthr-posix -> gthr-default.h

[Ioan-Adrian Ratiu]

libgfortran needs gthr-default.h to exist (included by libgcc/gthr.h),
but for some reason it is not symlinked in configure as defined by
AC_CONFIG_LINKS. This patch creates it at the end of do_configure; thus
allowing libgfortran to build without erros.

Signed-off-by: Ioan-Adrian Ratiu <adrian.ratiu@ni.com>
---
 meta/recipes-devtools/gcc/libgcc-common.inc | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/meta/recipes-devtools/gcc/libgcc-common.inc b/meta/recipes-devtools/gcc/libgcc-common.inc
index b09ea65..2922219 100644
--- a/meta/recipes-devtools/gcc/libgcc-common.inc
+++ b/meta/recipes-devtools/gcc/libgcc-common.inc
@@ -13,6 +13,9 @@ do_configure () {
 	cd ${B}/${BPN}
 	chmod a+x ${S}/${BPN}/configure
 	${S}/${BPN}/configure ${CONFIGUREOPTS} ${EXTRA_OECONF}
+
+	#fix libgfortran build which expects gthr-default.h -> gthr-posix.h
+	ln -sf ${S}/${BPN}/gthr-posix.h ${S}/${BPN}/gthr-default.h
 }
 
 do_compile () {
-- 
2.7.0

Re: [PATCH 1/1] gcc: libgcc-common.inc: symlink gthr-posix -> gthr-default.h

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 1286 bytes --]

On 26 January 2016 at 11:25, Ioan-Adrian Ratiu <adrian.ratiu@ni.com> wrote:

> libgfortran needs gthr-default.h to exist (included by libgcc/gthr.h),
> but for some reason it is not symlinked in configure as defined by
> AC_CONFIG_LINKS. This patch creates it at the end of do_configure; thus
> allowing libgfortran to build without erros.
>

Are you sure?  libgfortran builds here and gthr-default.h is in the build
tree:

In
.../work/corei7-64-poky-linux/libgcc/5.3.0-r0/gcc-5.3.0/build.x86_64-poky-linux.x86_64-poky-linux/libgcc
$ ls -l gthr-default.h
lrwxrwxrwx 1 ross ross 76 Jan 25 16:20 gthr-default.h ->
/data/poky-master/tmp/work-shared/gcc-5.3.0-r0/gcc-5.3.0/libgcc/gthr-posix.h

I'll note that libgfortran actually then errors out in package QA:

ERROR: QA Issue: libgfortran: Files/directories were installed but not
shipped in any package:
  /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude
  /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude/ieee_features.mod
  /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude/ieee_arithmetic.mod
  /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude/ieee_exceptions.mod

If you actually use fortran would you be able to submit the right fix?  I'd
guess at adding those to FILES_PN would be right, but that's a guess.

Ross

[-- Attachment #2: Type: text/html, Size: 2253 bytes --]

Re: [PATCH 1/1] gcc: libgcc-common.inc: symlink gthr-posix -> gthr-default.h

[Ioan-Adrian Ratiu]

On Tue, 26 Jan 2016 12:20:43 +0000
"Burton, Ross" <ross.burton@intel.com> wrote:

> On 26 January 2016 at 11:25, Ioan-Adrian Ratiu <adrian.ratiu@ni.com> wrote:
> 
> > libgfortran needs gthr-default.h to exist (included by libgcc/gthr.h),
> > but for some reason it is not symlinked in configure as defined by
> > AC_CONFIG_LINKS. This patch creates it at the end of do_configure; thus
> > allowing libgfortran to build without erros.
> >
> 
> Are you sure?  libgfortran builds here and gthr-default.h is in the build
> tree:
> 
> In
> .../work/corei7-64-poky-linux/libgcc/5.3.0-r0/gcc-5.3.0/build.x86_64-poky-linux.x86_64-poky-linux/libgcc
> $ ls -l gthr-default.h
> lrwxrwxrwx 1 ross ross 76 Jan 25 16:20 gthr-default.h ->
> /data/poky-master/tmp/work-shared/gcc-5.3.0-r0/gcc-5.3.0/libgcc/gthr-posix.h
> 
> I'll note that libgfortran actually then errors out in package QA:
> 
> ERROR: QA Issue: libgfortran: Files/directories were installed but not
> shipped in any package:
>   /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude
>   /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude/ieee_features.mod
>   /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude/ieee_arithmetic.mod
>   /usr/lib/gcc/x86_64-poky-linux/5.3.0/finclude/ieee_exceptions.mod
> 
> If you actually use fortran would you be able to submit the right fix?  I'd
> guess at adding those to FILES_PN would be right, but that's a guess.

I'm using fortran to build a recipe for python-scipy on the fido branch. The header symlink is definitely not created in fido.
Also I'm not getting any QA errors; I don't see why the symlink gets created on master while not in fido.

I'll try to test more extensively on the master branch and then cherry-pick if there are any fixes there to fido.

Thanks,
Adrian

> 
> Ross