[Buildroot] [PATCH] ltrace: disable elfutils unwinding when libunwind is used

[Buildroot] [PATCH] ltrace: disable elfutils unwinding when libunwind is used

[Baruch Siach]

ltrace can't use both.

Fixes:
http://autobuild.buildroot.net/results/a3c/a3c1647578cdeda3aaf672084a9a893f06dbad26/
http://autobuild.buildroot.net/results/fdb/fdb7e760faeda77b263a6dac3e322a25b53e67ff/
http://autobuild.buildroot.net/results/70a/70a0614ca7baf6a46898834a87cd46865f6e3385/

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
---
 package/ltrace/ltrace.mk | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/package/ltrace/ltrace.mk b/package/ltrace/ltrace.mk
index 5c9549f4019b..61e0fb7d57be 100644
--- a/package/ltrace/ltrace.mk
+++ b/package/ltrace/ltrace.mk
@@ -21,7 +21,7 @@ LTRACE_POST_PATCH_HOOKS += LTRACE_CREATE_CONFIG_M4
 # We don't normally do so for uClibc and we can't know if it's external
 ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
 ifeq ($(BR2_TOOLCHAIN_USES_UCLIBC),)
-LTRACE_CONF_OPTS += --with-libunwind=yes
+LTRACE_CONF_OPTS += --with-libunwind=yes --with-elfutils=no
 LTRACE_DEPENDENCIES += libunwind
 else
 LTRACE_CONF_OPTS += --with-libunwind=no
-- 
2.1.3

[Buildroot] [PATCH] ltrace: disable elfutils unwinding when libunwind is used

[Thomas Petazzoni]

Dear Baruch Siach,

On Fri,  5 Dec 2014 09:09:53 +0200, Baruch Siach wrote:
> ltrace can't use both.
> 
> Fixes:
> http://autobuild.buildroot.net/results/a3c/a3c1647578cdeda3aaf672084a9a893f06dbad26/
> http://autobuild.buildroot.net/results/fdb/fdb7e760faeda77b263a6dac3e322a25b53e67ff/
> http://autobuild.buildroot.net/results/70a/70a0614ca7baf6a46898834a87cd46865f6e3385/
> 
> Signed-off-by: Baruch Siach <baruch@tkos.co.il>
> ---
>  package/ltrace/ltrace.mk | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/package/ltrace/ltrace.mk b/package/ltrace/ltrace.mk
> index 5c9549f4019b..61e0fb7d57be 100644
> --- a/package/ltrace/ltrace.mk
> +++ b/package/ltrace/ltrace.mk
> @@ -21,7 +21,7 @@ LTRACE_POST_PATCH_HOOKS += LTRACE_CREATE_CONFIG_M4
>  # We don't normally do so for uClibc and we can't know if it's external
>  ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
>  ifeq ($(BR2_TOOLCHAIN_USES_UCLIBC),)
> -LTRACE_CONF_OPTS += --with-libunwind=yes
> +LTRACE_CONF_OPTS += --with-libunwind=yes --with-elfutils=no

While I agree with the analysis, I don't agree with the proposed fix:
ltrace selects elfutils unconditionally (see Config.in), so it is quite
weird to see ltrace not using it in the end. So I believe the Config.in
*and* ltrace dependencies should be updated to reflect the fact that
either unwind or elfutils can be used.

Could you work on an updated patch?

Thanks a lot!

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

[Buildroot] [PATCH] ltrace: disable elfutils unwinding when libunwind is used

[Baruch Siach]

Hi Thomas,

On Sun, Dec 07, 2014 at 10:00:14PM +0100, Thomas Petazzoni wrote:
> On Fri,  5 Dec 2014 09:09:53 +0200, Baruch Siach wrote:
> > ltrace can't use both.
> > 
> > Fixes:
> > http://autobuild.buildroot.net/results/a3c/a3c1647578cdeda3aaf672084a9a893f06dbad26/
> > http://autobuild.buildroot.net/results/fdb/fdb7e760faeda77b263a6dac3e322a25b53e67ff/
> > http://autobuild.buildroot.net/results/70a/70a0614ca7baf6a46898834a87cd46865f6e3385/
> > 
> > Signed-off-by: Baruch Siach <baruch@tkos.co.il>
> > ---
> >  package/ltrace/ltrace.mk | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/package/ltrace/ltrace.mk b/package/ltrace/ltrace.mk
> > index 5c9549f4019b..61e0fb7d57be 100644
> > --- a/package/ltrace/ltrace.mk
> > +++ b/package/ltrace/ltrace.mk
> > @@ -21,7 +21,7 @@ LTRACE_POST_PATCH_HOOKS += LTRACE_CREATE_CONFIG_M4
> >  # We don't normally do so for uClibc and we can't know if it's external
> >  ifeq ($(BR2_PACKAGE_LIBUNWIND),y)
> >  ifeq ($(BR2_TOOLCHAIN_USES_UCLIBC),)
> > -LTRACE_CONF_OPTS += --with-libunwind=yes
> > +LTRACE_CONF_OPTS += --with-libunwind=yes --with-elfutils=no
> 
> While I agree with the analysis, I don't agree with the proposed fix:
> ltrace selects elfutils unconditionally (see Config.in), so it is quite
> weird to see ltrace not using it in the end. So I believe the Config.in
> *and* ltrace dependencies should be updated to reflect the fact that
> either unwind or elfutils can be used.

It turns out that --with-elfutils=no is a misnomer. ltrace depends on libelf 
regardless of --with-elfutils, so we must have elfutils both in Config.in and 
in LTRACE_DEPENDENCIES. The only effect of --with-elfutils is on unwinding 
back-end. Quoting configure.ac:

  AS_HELP_STRING([--with-elfutils], [Use elfutils libdwfl unwinding support]),

I'll send a revised version of this patch with a better commit log, and a 
comment in the code explaining this weirdness.

baruch

-- 
     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] ltrace: disable elfutils unwinding when libunwind is used

[Thomas Petazzoni]

Dear Baruch Siach,

On Tue, 9 Dec 2014 08:02:47 +0200, Baruch Siach wrote:

> > While I agree with the analysis, I don't agree with the proposed fix:
> > ltrace selects elfutils unconditionally (see Config.in), so it is quite
> > weird to see ltrace not using it in the end. So I believe the Config.in
> > *and* ltrace dependencies should be updated to reflect the fact that
> > either unwind or elfutils can be used.
> 
> It turns out that --with-elfutils=no is a misnomer. ltrace depends on libelf 
> regardless of --with-elfutils, so we must have elfutils both in Config.in and 
> in LTRACE_DEPENDENCIES. The only effect of --with-elfutils is on unwinding 
> back-end. Quoting configure.ac:
> 
>   AS_HELP_STRING([--with-elfutils], [Use elfutils libdwfl unwinding support]),
> 
> I'll send a revised version of this patch with a better commit log, and a 
> comment in the code explaining this weirdness.

Aah, ok. Indeed very confusing. Thanks for clarifying this and posting
an updated version with more details!

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