[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Peter Seiderer]

Add fallback in case program_invocation_short_name is undeclared,
e.g. in case of some uclibc configurations.

Fixes [1]:
  evtest.c:730:20: error: 'program_invocation_short_name' undeclared (first use in this function)
    printf("%s %s\n", program_invocation_short_name, PACKAGE_VERSION);

[1] http://autobuild.buildroot.net/results/0fd/0fd36a0d79c4d82aedebb5aca8d3ce4214b1ed61

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
---
 ...-program_invocation_short_name-undeclared.patch | 50 ++++++++++++++++++++++
 1 file changed, 50 insertions(+)
 create mode 100644 package/evtest/0001-fix-program_invocation_short_name-undeclared.patch

diff --git a/package/evtest/0001-fix-program_invocation_short_name-undeclared.patch b/package/evtest/0001-fix-program_invocation_short_name-undeclared.patch
new file mode 100644
index 0000000..17f0fb3
--- /dev/null
+++ b/package/evtest/0001-fix-program_invocation_short_name-undeclared.patch
@@ -0,0 +1,50 @@
+From c124b50e13aaff46e7928d66b329644bf9b6dae7 Mon Sep 17 00:00:00 2001
+From: Peter Seiderer <ps.report@gmx.net>
+Date: Thu, 19 Mar 2015 23:08:34 +0100
+Subject: [PATCH] fix program_invocation_short_name undeclared
+
+Add fallback in case program_invocation_short_name is undeclared,
+e.g. in case of some uclibc configurations.
+
+Signed-off-by: Peter Seiderer <ps.report@gmx.net>
+---
+ configure.ac | 8 ++++++++
+ evtest.c     | 5 +++++
+ 2 files changed, 13 insertions(+)
+
+diff --git a/configure.ac b/configure.ac
+index 8dabc81..c13b05e 100644
+--- a/configure.ac
++++ b/configure.ac
+@@ -22,4 +22,12 @@ if test "x$XMLTO" = "x" || test "x$ASCIIDOC" = "x"; then
+ 	AC_MSG_WARN([xmlto or asciidoc not found - cannot create man pages without it])
+ fi
+ 
++AC_MSG_CHECKING([for program_invocation_short_name])
++AC_TRY_COMPILE([#define _GNU_SOURCE
++		#include <errno.h>], 
++	[strlen(program_invocation_short_name)],
++	AC_DEFINE(HAVE_PROGRAM_INVOCATION_SHORT_NAME, 1,
++		[Defined if program_invocation_short_name is defined])
++	AC_MSG_RESULT(yes), AC_MSG_RESULT(no))
++
+ AC_OUTPUT([Makefile])
+diff --git a/evtest.c b/evtest.c
+index 40bc82b..d601889 100644
+--- a/evtest.c
++++ b/evtest.c
+@@ -86,6 +86,11 @@
+ 
+ #define NAME_ELEMENT(element) [element] = #element
+ 
++/* fallback to canonical name */
++#ifndef HAVE_PROGRAM_INVOCATION_SHORT_NAME
++static char *program_invocation_short_name = "evtest";
++#endif
++
+ enum evtest_mode {
+ 	MODE_CAPTURE,
+ 	MODE_QUERY,
+-- 
+2.1.4
+
-- 
2.1.4

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Thomas Petazzoni]

Dear Peter Seiderer,

Adding Alexey from Synopsys in Cc.

On Thu, 19 Mar 2015 23:18:18 +0100, Peter Seiderer wrote:
> Add fallback in case program_invocation_short_name is undeclared,
> e.g. in case of some uclibc configurations.
> 
> Fixes [1]:
>   evtest.c:730:20: error: 'program_invocation_short_name' undeclared (first use in this function)
>     printf("%s %s\n", program_invocation_short_name, PACKAGE_VERSION);
> 
> [1] http://autobuild.buildroot.net/results/0fd/0fd36a0d79c4d82aedebb5aca8d3ce4214b1ed61
> 
> Signed-off-by: Peter Seiderer <ps.report@gmx.net>

Shouldn't we instead ask Synopsys to rebuild their toolchains with a
uClibc configuration that includes
UCLIBC_HAS_PROGRAM_INVOCATION_NAME=y, so that it matches the
default Buildroot uClibc and we don't have to add gazillions of
workarounds in multiple packages just to support the ARC external
toolchain?

Alexey ?

Thanks,

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

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Alexey Brodkin]

Hi Peter,

On Fri, 2015-03-20 at 14:09 +0100, Thomas Petazzoni wrote:
> Dear Peter Seiderer,
> 
> Adding Alexey from Synopsys in Cc.
> 
> On Thu, 19 Mar 2015 23:18:18 +0100, Peter Seiderer wrote:
> > Add fallback in case program_invocation_short_name is undeclared,
> > e.g. in case of some uclibc configurations.
> > 
> > Fixes [1]:
> >   evtest.c:730:20: error: 'program_invocation_short_name' undeclared (first use in this function)
> >     printf("%s %s\n", program_invocation_short_name, PACKAGE_VERSION);
> > 
> > [1] http://autobuild.buildroot.net/results/0fd/0fd36a0d79c4d82aedebb5aca8d3ce4214b1ed61
> > 
> > Signed-off-by: Peter Seiderer <ps.report@gmx.net>
> 
> Shouldn't we instead ask Synopsys to rebuild their toolchains with a
> uClibc configuration that includes
> UCLIBC_HAS_PROGRAM_INVOCATION_NAME=y, so that it matches the
> default Buildroot uClibc and we don't have to add gazillions of
> workarounds in multiple packages just to support the ARC external
> toolchain?

Thanks for this report about missing feature.
Indeed this one was disabled (not yet enabled) by 2014.12 release.
These are uClibc defconfigs we used for building GNU tools for ARC:

 [ARC 700]
https://github.com/foss-for-synopsys-dwc-arc-processors/uClibc/blob/arc-2014.12/extra/Configs/defconfigs/arc/defconfig
 [ARC HS]
https://github.com/foss-for-synopsys-dwc-arc-processors/uClibc/blob/arc-2014.12/extra/Configs/defconfigs/arc/arcv2_defconfig

And unfortunately chances are low to rebuild 2014.12 tools with modified
configs.

But! We faced some issues when dealing with perf and elfutils and
understood that we need to enable both
UCLIBC_HAS_PROGRAM_INVOCATION_NAME and UCLIBC_HAS_GLIBC_CUSTOM_STREAMS -
https://github.com/foss-for-synopsys-dwc-arc-processors/uClibc/commit/c2460b9b7b8c76098dfb1313be4aa4a4a65ff619

So in the next release (that we plan to prepare later this spring) we'll
have those enabled in prebuilt tools.

And for now I would propose to ignore mentioned build errors. Because
indeed there's no point to throw in tons of temporary patches - if
anybody really wants those "problematic" packages to be built there's
always a possibility to build toolchain in Buildroot and it will use
Buildroot's default uClibc config.

Will it work for you?

-Alexey

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Thomas Petazzoni]

Dear Alexey Brodkin,

On Fri, 20 Mar 2015 20:02:58 +0000, Alexey Brodkin wrote:

> And for now I would propose to ignore mentioned build errors. Because
> indeed there's no point to throw in tons of temporary patches - if
> anybody really wants those "problematic" packages to be built there's
> always a possibility to build toolchain in Buildroot and it will use
> Buildroot's default uClibc config.
> 
> Will it work for you?

Yes, it works. In the mean time, there is a way to avoid the build
errors: add an exception to the autobuilder script. See
http://git.buildroot.net/buildroot-test/tree/scripts/autobuild-run#n346
for a bunch of examples.

Thanks,

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

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Peter Korsgaard]

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

 > Dear Alexey Brodkin,
 > On Fri, 20 Mar 2015 20:02:58 +0000, Alexey Brodkin wrote:

 >> And for now I would propose to ignore mentioned build errors. Because
 >> indeed there's no point to throw in tons of temporary patches - if
 >> anybody really wants those "problematic" packages to be built there's
 >> always a possibility to build toolchain in Buildroot and it will use
 >> Buildroot's default uClibc config.
 >> 
 >> Will it work for you?

 > Yes, it works. In the mean time, there is a way to avoid the build
 > errors: add an exception to the autobuilder script. See
 > http://git.buildroot.net/buildroot-test/tree/scripts/autobuild-run#n346
 > for a bunch of examples.

Or simply make the packages depend on
!BR2_TOOLCHAIN_EXTERNAL_SYNOPSYS_ARC_2014_12. Peter will you send
patches to do that for evtest and libevdev?

-- 
Bye, Peter Korsgaard

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Thomas Petazzoni]

Dear Peter Korsgaard,

On Mon, 23 Mar 2015 23:32:52 +0100, Peter Korsgaard wrote:

>  > Yes, it works. In the mean time, there is a way to avoid the build
>  > errors: add an exception to the autobuilder script. See
>  > http://git.buildroot.net/buildroot-test/tree/scripts/autobuild-run#n346
>  > for a bunch of examples.
> 
> Or simply make the packages depend on
> !BR2_TOOLCHAIN_EXTERNAL_SYNOPSYS_ARC_2014_12. Peter will you send
> patches to do that for evtest and libevdev?

This is indeed OK for packages such as evtest that don't have, or don't
have many, reverse dependencies. But this is hardly a sustainable
solution in general, especially for packages that have a high number of
reverse dependencies.

We simply can't handle all the little possible variations of uClibc
configurations that exist in every external toolchain, it's not
workable in a general fashion.

I know we're already doing it to some extent, but I'd like to avoid
doing it when possible, especially when we have a good cooperation with
the upstream developers (in this case, the Synopsys folks).

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

[Buildroot] [PATCH v1] evtest: fix program_invocation_short_name undeclared

[Peter Korsgaard]

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

 > Dear Peter Korsgaard,
 > On Mon, 23 Mar 2015 23:32:52 +0100, Peter Korsgaard wrote:

 >> > Yes, it works. In the mean time, there is a way to avoid the build
 >> > errors: add an exception to the autobuilder script. See
 >> > http://git.buildroot.net/buildroot-test/tree/scripts/autobuild-run#n346
 >> > for a bunch of examples.
 >> 
 >> Or simply make the packages depend on
 >> !BR2_TOOLCHAIN_EXTERNAL_SYNOPSYS_ARC_2014_12. Peter will you send
 >> patches to do that for evtest and libevdev?

 > This is indeed OK for packages such as evtest that don't have, or don't
 > have many, reverse dependencies. But this is hardly a sustainable
 > solution in general, especially for packages that have a high number of
 > reverse dependencies.

True, with lots of reverse deps it becomes very painful.

 > We simply can't handle all the little possible variations of uClibc
 > configurations that exist in every external toolchain, it's not
 > workable in a general fashion.

We don't support that many external uClibc based toolchains (only arc,
bfin and sh) so I think it's still doable, but ok.

 > I know we're already doing it to some extent, but I'd like to avoid
 > doing it when possible, especially when we have a good cooperation with
 > the upstream developers (in this case, the Synopsys folks).

If we can remove the arc 2014.12 toolchain and add the new 2015 one
before 2015.05-rc1 (E.G. end of April) then I'm fine with just disabling
it in the autobuilders for now, otherwise I still think we should add
the depends on !arc-2014-12 to not break it for users.

-- 
Bye, Peter Korsgaard