[Buildroot] [PATCH] argp-standalone: fix build with > GCC5.1

[Buildroot] [PATCH] argp-standalone: fix build with > GCC5.1

[Matt Weber]

The problem is the change of the default C standard from gnu89 to gnu11
which changes the semantics of 'inline'. The issue is described in the
Porting guide at https://gcc.gnu.org/gcc-5/porting_to.html. Adding the
'-fgnu89-inline' option fixes the issue.

Similar issue:
https://www.mail-archive.com/ptxdist at pengutronix.de/msg09746.html

Fixes:
http://autobuild.buildroot.net/results/a9c/a9cedc54829b7bd2dd7ae6ff2bd6c6db242f1c35//

Tested via test-pkg with the following cfg
BR2_PACKAGE_ARGP_STANDALONE=y

Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>
---
 package/argp-standalone/argp-standalone.mk | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/package/argp-standalone/argp-standalone.mk b/package/argp-standalone/argp-standalone.mk
index 7a7028c..977b954 100644
--- a/package/argp-standalone/argp-standalone.mk
+++ b/package/argp-standalone/argp-standalone.mk
@@ -10,7 +10,7 @@ ARGP_STANDALONE_INSTALL_STAGING = YES
 ARGP_STANDALONE_LICENSE = LGPL-2.0+
 
 ARGP_STANDALONE_CONF_ENV = \
-	CFLAGS="$(TARGET_CFLAGS) -fPIC"
+	CFLAGS="$(TARGET_CFLAGS) -fPIC -fgnu89-inline"
 
 define ARGP_STANDALONE_INSTALL_STAGING_CMDS
 	$(INSTALL) -D $(@D)/libargp.a $(STAGING_DIR)/usr/lib/libargp.a
-- 
2.7.4

[Buildroot] [PATCH] argp-standalone: fix build with > GCC5.1

[Thomas Petazzoni]

Hello,

On Tue, 31 Oct 2017 01:18:34 -0500, Matt Weber wrote:
> The problem is the change of the default C standard from gnu89 to gnu11
> which changes the semantics of 'inline'. The issue is described in the
> Porting guide at https://gcc.gnu.org/gcc-5/porting_to.html. Adding the
> '-fgnu89-inline' option fixes the issue.
> 
> Similar issue:
> https://www.mail-archive.com/ptxdist at pengutronix.de/msg09746.html
> 
> Fixes:
> http://autobuild.buildroot.net/results/a9c/a9cedc54829b7bd2dd7ae6ff2bd6c6db242f1c35//
> 
> Tested via test-pkg with the following cfg
> BR2_PACKAGE_ARGP_STANDALONE=y
> 
> Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>

Thanks for looking into this. A few questions though:

 - Why does this happen only on ARC ?

 - Why are we seeing this issue only now, while we have started to use
   gcc 5.x a long, long time ago ?

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[Buildroot] [PATCH] argp-standalone: fix build with > GCC5.1

[Thomas Petazzoni]

Hello,

On Tue, 31 Oct 2017 01:18:34 -0500, Matt Weber wrote:
> The problem is the change of the default C standard from gnu89 to gnu11
> which changes the semantics of 'inline'. The issue is described in the
> Porting guide at https://gcc.gnu.org/gcc-5/porting_to.html. Adding the
> '-fgnu89-inline' option fixes the issue.
> 
> Similar issue:
> https://www.mail-archive.com/ptxdist at pengutronix.de/msg09746.html
> 
> Fixes:
> http://autobuild.buildroot.net/results/a9c/a9cedc54829b7bd2dd7ae6ff2bd6c6db242f1c35//
> 
> Tested via test-pkg with the following cfg
> BR2_PACKAGE_ARGP_STANDALONE=y
> 
> Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>

So, I've applied, but with those tweaks. Indeed, your patch could
definitely not be fixing an issue with gcc >= 5.x, because we have been
using gcc 5.x for a while, and this build error was not popping up with
anything but gcc 7.x.

So, what happened is that in commit
a662ff7e79630ca0875dd8529fe54db27a275007, we added a patch that fixes
the build with C99 compilers (in fact this means gcc 5.x, which uses
C99 inline semantics). This changed "extern inline" to just "inline".
But that broke again with gcc 7.x, and causes the problem we're seeing
now.

So instead, I've dropped the chunk of the patch that changed from
extern inline to inline, and passsed -fgnu89-inline as you suggested. I
updated the commit log accordingly.

See
https://git.buildroot.org/buildroot/commit/?id=f0b65bd90ce4429d6b7e952ce7de2d5f92a2dd26
for more details.

Thanks for all your research behind this issue!

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[Buildroot] [PATCH] argp-standalone: fix build with > GCC5.1

[Matthew Weber]

Thomas,

On Wed, Nov 1, 2017 at 2:58 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
>
> Hello,
>
> On Tue, 31 Oct 2017 01:18:34 -0500, Matt Weber wrote:
> > The problem is the change of the default C standard from gnu89 to gnu11
> > which changes the semantics of 'inline'. The issue is described in the
> > Porting guide at https://gcc.gnu.org/gcc-5/porting_to.html. Adding the
> > '-fgnu89-inline' option fixes the issue.
> >
> > Similar issue:
> > https://www.mail-archive.com/ptxdist at pengutronix.de/msg09746.html
> >
> > Fixes:
> > http://autobuild.buildroot.net/results/a9c/a9cedc54829b7bd2dd7ae6ff2bd6c6db242f1c35//
> >
> > Tested via test-pkg with the following cfg
> > BR2_PACKAGE_ARGP_STANDALONE=y
> >
> > Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>
>
> So, I've applied, but with those tweaks. Indeed, your patch could
> definitely not be fixing an issue with gcc >= 5.x, because we have been
> using gcc 5.x for a while, and this build error was not popping up with
> anything but gcc 7.x.
>
> So, what happened is that in commit
> a662ff7e79630ca0875dd8529fe54db27a275007, we added a patch that fixes
> the build with C99 compilers (in fact this means gcc 5.x, which uses
> C99 inline semantics). This changed "extern inline" to just "inline".
> But that broke again with gcc 7.x, and causes the problem we're seeing
> now.
>
> So instead, I've dropped the chunk of the patch that changed from
> extern inline to inline, and passsed -fgnu89-inline as you suggested. I
> updated the commit log accordingly.
>
> See
> https://git.buildroot.org/buildroot/commit/?id=f0b65bd90ce4429d6b7e952ce7de2d5f92a2dd26
> for more details.
>
> Thanks for all your research behind this issue!
>

NP,  appreciate the follow-up and now the details make a lot more
sense.  Plus it's good to see that it indirectly found a toolchain
coverage issue with the test-pkg.

Matt

[Buildroot] [PATCH] argp-standalone: fix build with > GCC5.1

[Peter Korsgaard]

>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes:

 > Hello,
 > On Tue, 31 Oct 2017 01:18:34 -0500, Matt Weber wrote:
 >> The problem is the change of the default C standard from gnu89 to gnu11
 >> which changes the semantics of 'inline'. The issue is described in the
 >> Porting guide at https://gcc.gnu.org/gcc-5/porting_to.html. Adding the
 >> '-fgnu89-inline' option fixes the issue.
 >> 
 >> Similar issue:
 >> https://www.mail-archive.com/ptxdist at pengutronix.de/msg09746.html
 >> 
 >> Fixes:
 >> http://autobuild.buildroot.net/results/a9c/a9cedc54829b7bd2dd7ae6ff2bd6c6db242f1c35//
 >> 
 >> Tested via test-pkg with the following cfg
 >> BR2_PACKAGE_ARGP_STANDALONE=y
 >> 
 >> Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>

 > So, I've applied, but with those tweaks. Indeed, your patch could
 > definitely not be fixing an issue with gcc >= 5.x, because we have been
 > using gcc 5.x for a while, and this build error was not popping up with
 > anything but gcc 7.x.

 > So, what happened is that in commit
 > a662ff7e79630ca0875dd8529fe54db27a275007, we added a patch that fixes
 > the build with C99 compilers (in fact this means gcc 5.x, which uses
 > C99 inline semantics). This changed "extern inline" to just "inline".
 > But that broke again with gcc 7.x, and causes the problem we're seeing
 > now.

 > So instead, I've dropped the chunk of the patch that changed from
 > extern inline to inline, and passsed -fgnu89-inline as you suggested. I
 > updated the commit log accordingly.

 > See
 > https://git.buildroot.org/buildroot/commit/?id=f0b65bd90ce4429d6b7e952ce7de2d5f92a2dd26
 > for more details.

Committed the revised patch to 2017.02.x and 2017.08.x, thanks both!

-- 
Bye, Peter Korsgaard