[Buildroot] [PATCH 0/3] Fix ltrace autobuilder failures

[Buildroot] [PATCH 0/3] Fix ltrace autobuilder failures

[Jérôme Pouiller]

This patch serie try to fix different issues detected by autobuilder on
ltrace package:
    http://autobuild.buildroot.net/results/79dae85647526cdeddae5050db0339fbc38c9ae6
    http://autobuild.buildroot.net/results/808569f6bfac059b57eb4f1e692fbc9909e791cf
    http://autobuild.buildroot.net/results/83d0cd13d64053869708c666e2aa533202af67f7
    http://autobuild.buildroot.net/results/d4c465ea57a1954c3323234100ae08156cbd8324

J?r?me Pouiller (3):
  ltrace: Explicit enable/disable of libunwind
  ltrace: Fix a May be unitialized variable
  ltrace: Fix "strndup not declared" build failure on powerpc/uclibc

 .../ltrace-fix-may-be-uninitialized-variable.patch |   11 +++++++++++
 package/ltrace/ltrace-fix-strndup-usage.patch      |   15 +++++++++++++++
 package/ltrace/ltrace.mk                           |    9 +++++++++
 3 files changed, 35 insertions(+)
 create mode 100644 package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch
 create mode 100644 package/ltrace/ltrace-fix-strndup-usage.patch

-- 
1.7.9.5

[Buildroot] [PATCH 1/3] ltrace: Explicit enable/disable of libunwind

[Jérôme Pouiller]

ltrace automaticaly use libunwind if detected. So depending if libunwind was
already compiled or not, ltrcae compile differently. This patch add explicit
dependency to libunwind if detected.

In add, due to limitation of luclibc, ltrace cannot use libunwind with uclib. this patch explicitly compile without libunwind in this case.

Resolve issues detected here:
    http://autobuild.buildroot.net/results/79dae85647526cdeddae5050db0339fbc38c9ae6
    http://autobuild.buildroot.net/results/808569f6bfac059b57eb4f1e692fbc9909e791cf
    http://autobuild.buildroot.net/results/83d0cd13d64053869708c666e2aa533202af67f7

Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
---
 package/ltrace/ltrace.mk |    9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/package/ltrace/ltrace.mk b/package/ltrace/ltrace.mk
index 7b8d859..4d81121 100644
--- a/package/ltrace/ltrace.mk
+++ b/package/ltrace/ltrace.mk
@@ -13,4 +13,13 @@ LTRACE_CONF_OPT = --disable-werror
 LTRACE_LICENSE = GPLv2
 LTRACE_LICENSE_FILES = COPYING
 
+LTRACE_LIBUNWIDFLAG = --without-libunwind
+ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
+  ifneq ($(BR2_TOOLCHAIN_USES_UCLIBC),y)
+    LTRACE_DEPENDENCIES = libunwind
+    LTRACE_LIBUNWIDFLAG = --with-libunwind
+  endif
+endif
+LTRACE_CONF_OPT = $(LTRACE_LIBUNWIDFLAG)
+
 $(eval $(autotools-package))
-- 
1.7.9.5

[Buildroot] [PATCH 1/3] ltrace: Explicit enable/disable of libunwind

[Baruch Siach]

Hi J?r?me,

On Fri, Aug 09, 2013 at 11:36:57AM +0200, J?r?me Pouiller wrote:
> ltrace automaticaly use libunwind if detected. So depending if libunwind was
> already compiled or not, ltrcae compile differently. This patch add explicit
> dependency to libunwind if detected.
> 
> In add, due to limitation of luclibc, ltrace cannot use libunwind with uclib. this patch explicitly compile without libunwind in this case.

s/luclibc/uclibc/

Also, please warp log message lines at about 80 characters.

baruch

> Resolve issues detected here:
>     http://autobuild.buildroot.net/results/79dae85647526cdeddae5050db0339fbc38c9ae6
>     http://autobuild.buildroot.net/results/808569f6bfac059b57eb4f1e692fbc9909e791cf
>     http://autobuild.buildroot.net/results/83d0cd13d64053869708c666e2aa533202af67f7
> 
> Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
> ---
>  package/ltrace/ltrace.mk |    9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/package/ltrace/ltrace.mk b/package/ltrace/ltrace.mk
> index 7b8d859..4d81121 100644
> --- a/package/ltrace/ltrace.mk
> +++ b/package/ltrace/ltrace.mk
> @@ -13,4 +13,13 @@ LTRACE_CONF_OPT = --disable-werror
>  LTRACE_LICENSE = GPLv2
>  LTRACE_LICENSE_FILES = COPYING
>  
> +LTRACE_LIBUNWIDFLAG = --without-libunwind
> +ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
> +  ifneq ($(BR2_TOOLCHAIN_USES_UCLIBC),y)
> +    LTRACE_DEPENDENCIES = libunwind
> +    LTRACE_LIBUNWIDFLAG = --with-libunwind
> +  endif
> +endif
> +LTRACE_CONF_OPT = $(LTRACE_LIBUNWIDFLAG)
> +
>  $(eval $(autotools-package))
> -- 
> 1.7.9.5
> 
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

-- 
     http://baruch.siach.name/blog/                  ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -

[Buildroot] [PATCH 1/3] ltrace: Explicit enable/disable of libunwind

[Thomas Petazzoni]

Dear J?r?me Pouiller,

On Fri,  9 Aug 2013 11:36:57 +0200, J?r?me Pouiller wrote:
> ltrace automaticaly use libunwind if detected. So depending if libunwind was

automatically uses

> already compiled or not, ltrcae compile differently. This patch add explicit

ltrace compiles

This patch adds an

> dependency to libunwind if detected.
> 
> In add, due to limitation of luclibc, ltrace cannot use libunwind with uclib. this patch explicitly compile without libunwind in this case.

Line should be wrapped and fixed from typos.


> diff --git a/package/ltrace/ltrace.mk b/package/ltrace/ltrace.mk
> index 7b8d859..4d81121 100644
> --- a/package/ltrace/ltrace.mk
> +++ b/package/ltrace/ltrace.mk
> @@ -13,4 +13,13 @@ LTRACE_CONF_OPT = --disable-werror
>  LTRACE_LICENSE = GPLv2
>  LTRACE_LICENSE_FILES = COPYING
>  
> +LTRACE_LIBUNWIDFLAG = --without-libunwind
> +ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
> +  ifneq ($(BR2_TOOLCHAIN_USES_UCLIBC),y)
> +    LTRACE_DEPENDENCIES = libunwind
> +    LTRACE_LIBUNWIDFLAG = --with-libunwind
> +  endif
> +endif
> +LTRACE_CONF_OPT = $(LTRACE_LIBUNWIDFLAG)

That's not how we typically write such things. We normally do something
like:

ifeq ($(BR2_PACKAGE_LIBUNWIND)$(BR2_TOOLCHAIN_USES_GLIBC),yy)
LTRACE_DEPENDENCIES += libunwind
LTRACE_CONF_OPT += --with-libunwind
else
LTRACE_CONF_OPT += --without-libunwind
endif

Thanks,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 2/3] ltrace: Fix a May be unitialized variable

[Jérôme Pouiller]

Compilation may failed if rlm is not initialized and ltrace
is compiled with -Werror.

Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
---
 .../ltrace-fix-may-be-uninitialized-variable.patch |   11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch

diff --git a/package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch b/package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch
new file mode 100644
index 0000000..23ffc94
--- /dev/null
+++ b/package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch
@@ -0,0 +1,11 @@
+--- a/sysdeps/linux-gnu/proc.c	2012-12-04 13:36:23.000000000 +0100
++++ b/sysdeps/linux-gnu/proc.c	2013-08-09 10:14:33.576464435 +0200
+@@ -483,7 +483,7 @@
+ 	arch_addr_t addr = (arch_addr_t)(uintptr_t)dbg->r_map;
+ 
+ 	while (addr != 0) {
+-		struct lt_link_map_64 rlm;
++		struct lt_link_map_64 rlm = { };
+ 		if (lm_fetcher(proc)(proc, addr, &rlm) < 0) {
+ 			debug(2, "Unable to read link map");
+ 			return;
-- 
1.7.9.5

[Buildroot] [PATCH 2/3] ltrace: Fix a May be unitialized variable

[Baruch Siach]

Hi J?r?me,

On Fri, Aug 09, 2013 at 11:36:58AM +0200, J?r?me Pouiller wrote:
> Compilation may failed if rlm is not initialized and ltrace
> is compiled with -Werror.

The right fix, in my opinion, is to disable -Werror. Otherwise, we'll have to 
add more and more fixes like this one for each bump of the toolchain.

baruch

> Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
> ---
>  .../ltrace-fix-may-be-uninitialized-variable.patch |   11 +++++++++++
>  1 file changed, 11 insertions(+)
>  create mode 100644 package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch
> 
> diff --git a/package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch b/package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch
> new file mode 100644
> index 0000000..23ffc94
> --- /dev/null
> +++ b/package/ltrace/ltrace-fix-may-be-uninitialized-variable.patch
> @@ -0,0 +1,11 @@
> +--- a/sysdeps/linux-gnu/proc.c	2012-12-04 13:36:23.000000000 +0100
> ++++ b/sysdeps/linux-gnu/proc.c	2013-08-09 10:14:33.576464435 +0200
> +@@ -483,7 +483,7 @@
> + 	arch_addr_t addr = (arch_addr_t)(uintptr_t)dbg->r_map;
> + 
> + 	while (addr != 0) {
> +-		struct lt_link_map_64 rlm;
> ++		struct lt_link_map_64 rlm = { };
> + 		if (lm_fetcher(proc)(proc, addr, &rlm) < 0) {
> + 			debug(2, "Unable to read link map");
> + 			return;
> -- 
> 1.7.9.5
> 
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

-- 
     http://baruch.siach.name/blog/                  ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -

[Buildroot] [PATCH 2/3] ltrace: Fix a May be unitialized variable

[Thomas Petazzoni]

Dear J?r?me Pouiller,

On Fri,  9 Aug 2013 11:36:58 +0200, J?r?me Pouiller wrote:
> Compilation may failed if rlm is not initialized and ltrace
> is compiled with -Werror.

In general, we try to avoid compiling things with -Werror, but ok, why
not.

Can you add a description + Signed-off-by line to the patch, and send
it upstream? Thanks!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 3/3] ltrace: Fix "strndup not declared" build failure on powerpc/uclibc

[Jérôme Pouiller]

On some systems, strndup is declared only if _GNU_SOURCE is defined

Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
---
 package/ltrace/ltrace-fix-strndup-usage.patch |   15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 package/ltrace/ltrace-fix-strndup-usage.patch

diff --git a/package/ltrace/ltrace-fix-strndup-usage.patch b/package/ltrace/ltrace-fix-strndup-usage.patch
new file mode 100644
index 0000000..ccc0e2c
--- /dev/null
+++ b/package/ltrace/ltrace-fix-strndup-usage.patch
@@ -0,0 +1,15 @@
+With some systems, strndup is declared only if __USE_GNU is defined.
+
+--- a/sysdeps/linux-gnu/ppc/plt.c	2012-11-29 18:46:57.000000000 +0100
++++ b/sysdeps/linux-gnu/ppc/plt.c	2013-08-09 10:30:07.452044496 +0200
+@@ -20,6 +20,10 @@
+  * 02110-1301 USA
+  */
+ 
++#ifndef _GNU_SOURCE
++#  define _GNU_SOURCE 1
++#endif
++
+ #include <gelf.h>
+ #include <sys/ptrace.h>
+ #include <errno.h>
-- 
1.7.9.5

[Buildroot] [PATCH 3/3] ltrace: Fix "strndup not declared" build failure on powerpc/uclibc

[Baruch Siach]

Hi J?r?me,

On Fri, Aug 09, 2013 at 11:36:59AM +0200, J?r?me Pouiller wrote:
> On some systems, strndup is declared only if _GNU_SOURCE is defined
> 
> Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
> ---
>  package/ltrace/ltrace-fix-strndup-usage.patch |   15 +++++++++++++++
>  1 file changed, 15 insertions(+)
>  create mode 100644 package/ltrace/ltrace-fix-strndup-usage.patch
> 
> diff --git a/package/ltrace/ltrace-fix-strndup-usage.patch b/package/ltrace/ltrace-fix-strndup-usage.patch
> new file mode 100644
> index 0000000..ccc0e2c
> --- /dev/null
> +++ b/package/ltrace/ltrace-fix-strndup-usage.patch
> @@ -0,0 +1,15 @@
> +With some systems, strndup is declared only if __USE_GNU is defined.

Please add your sign-off on this patch.

baruch

> +
> +--- a/sysdeps/linux-gnu/ppc/plt.c	2012-11-29 18:46:57.000000000 +0100
> ++++ b/sysdeps/linux-gnu/ppc/plt.c	2013-08-09 10:30:07.452044496 +0200
> +@@ -20,6 +20,10 @@
> +  * 02110-1301 USA
> +  */
> + 
> ++#ifndef _GNU_SOURCE
> ++#  define _GNU_SOURCE 1
> ++#endif
> ++
> + #include <gelf.h>
> + #include <sys/ptrace.h>
> + #include <errno.h>

-- 
     http://baruch.siach.name/blog/                  ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -

[Buildroot] [PATCH 3/3] ltrace: Fix "strndup not declared" build failure on powerpc/uclibc

[Thomas Petazzoni]

Dear J?r?me Pouiller,

On Fri,  9 Aug 2013 11:36:59 +0200, J?r?me Pouiller wrote:
> On some systems, strndup is declared only if _GNU_SOURCE is defined
> 
> Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>

Shouldn't this be globally defined by adding a -D_GNU_SOURCE to the
CFLAGS at the package level?

If not, then anyway your patch needs a Signed-off-by line, and should
be submitted upstream.

Thanks!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 3/3] ltrace: Fix "strndup not declared" build failure on powerpc/uclibc

[Thomas De Schampheleire]

On Fri, Aug 9, 2013 at 12:29 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Dear J?r?me Pouiller,
>
> On Fri,  9 Aug 2013 11:36:59 +0200, J?r?me Pouiller wrote:
>> On some systems, strndup is declared only if _GNU_SOURCE is defined
>>
>> Signed-off-by: J?r?me Pouiller <jezz@sysmic.org>
>
> Shouldn't this be globally defined by adding a -D_GNU_SOURCE to the
> CFLAGS at the package level?

When you say 'package level' you mean in ltrace.mk in buildroot?
In this case I don't think it's correct: ltrace itself is responsible
for including the right header files, and setting the right macros, so
that any used interface is available. Hence, I consider it a patch to
ltrace (that should be upstreamed), not something we should fix in
buildroot itself.

[Buildroot] [PATCH 3/3] ltrace: Fix "strndup not declared" build failure on powerpc/uclibc

[Thomas Petazzoni]

Dear Thomas De Schampheleire,

On Fri, 9 Aug 2013 13:05:53 +0200, Thomas De Schampheleire wrote:

> > Shouldn't this be globally defined by adding a -D_GNU_SOURCE to the
> > CFLAGS at the package level?
> 
> When you say 'package level' you mean in ltrace.mk in buildroot?

Yes, that's what I meant, but I agree that I might have been wrong.

> In this case I don't think it's correct: ltrace itself is responsible
> for including the right header files, and setting the right macros, so
> that any used interface is available. Hence, I consider it a patch to
> ltrace (that should be upstreamed), not something we should fix in
> buildroot itself.

Hum, yes, you're right.

J?r?me, sorry then, your original patch was OK. Just add your
Signed-off-by on it.

Can you resend your small patch series with the various minor comments
addressed?

Thanks!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com