Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

I just noticed that after doing 'bitbake gcc' the old libstdc++ gets
replaced and the new one has:

dependency_libs='
/OE/angstrom-dev/sysroots/armv6-angstrom-linux-gnueabi/usr/lib/libstdc++.la'

instead of the expected '-lm'. It now refers to itself :(

regards,

Koen
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFL8Ne9MkyGM64RGpERAt2ZAJ96PebvOq7SBTF02DjtHxGmsHLb6wCeMeUE
Pj5h0IDCGGXN3aLLROGcDAg=
=oL/h
-----END PGP SIGNATURE-----

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Khem Raj]

On (17/05/10 07:44), Koen Kooi wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hi,
> 
> I just noticed that after doing 'bitbake gcc' the old libstdc++ gets
> replaced and the new one has:
> 
> dependency_libs='
> /OE/angstrom-dev/sysroots/armv6-angstrom-linux-gnueabi/usr/lib/libstdc++.la'
> 
> instead of the expected '-lm'. It now refers to itself :(

hmmm target gcc should not have staged libstdc++ if it was already staged
by gcc-cross but the problem is casued by
gcc-4.3.3-SYSROOT_CFLAGS_FOR_TARGET.patch it passed the native libdir which
is build_libdir even for target library builds using -rpath. I am currently
looking into the reason. If possible we can get rid of that patch.

The patch you committed to cross.bbclass will not be necessary

Thanks
-Khem

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 17-05-10 07:55, Khem Raj wrote:
> On (17/05/10 07:44), Koen Kooi wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Hi,
>>
>> I just noticed that after doing 'bitbake gcc' the old libstdc++ gets
>> replaced and the new one has:
>>
>> dependency_libs='
>> /OE/angstrom-dev/sysroots/armv6-angstrom-linux-gnueabi/usr/lib/libstdc++.la'
>>
>> instead of the expected '-lm'. It now refers to itself :(
> 
> hmmm target gcc should not have staged libstdc++ if it was already staged
> by gcc-cross but the problem is casued by
> gcc-4.3.3-SYSROOT_CFLAGS_FOR_TARGET.patch it passed the native libdir which
> is build_libdir even for target library builds using -rpath. I am currently
> looking into the reason. If possible we can get rid of that patch.

I'm getting a build failure with 'gcc' now failing to find libstc++.la.
I'm trying a build from scratch now to rule out any packaged-staging
problems (>1 recipes installing the same file, pstage removing it prior
to build).

regards,

Koen
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFL8Q57MkyGM64RGpERAvvpAJ4vR4wtskIhV6BBUN/LcOzxXYPSEwCeNNtm
FWhLXTLoZUXa+yIDjlEiNbo=
=EvPm
-----END PGP SIGNATURE-----

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 17-05-10 11:38, Koen Kooi wrote:
> On 17-05-10 07:55, Khem Raj wrote:
>> On (17/05/10 07:44), Koen Kooi wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>>
>>> Hi,
>>>
>>> I just noticed that after doing 'bitbake gcc' the old libstdc++ gets
>>> replaced and the new one has:
>>>
>>> dependency_libs='
>>> /OE/angstrom-dev/sysroots/armv6-angstrom-linux-gnueabi/usr/lib/libstdc++.la'
>>>
>>> instead of the expected '-lm'. It now refers to itself :(
> 
>> hmmm target gcc should not have staged libstdc++ if it was already staged
>> by gcc-cross but the problem is casued by
>> gcc-4.3.3-SYSROOT_CFLAGS_FOR_TARGET.patch it passed the native libdir which
>> is build_libdir even for target library builds using -rpath. I am currently
>> looking into the reason. If possible we can get rid of that patch.
> 
> I'm getting a build failure with 'gcc' now failing to find libstc++.la.
> I'm trying a build from scratch now to rule out any packaged-staging
> problems (>1 recipes installing the same file, pstage removing it prior
> to build).

Yep, it's a packaged-staging problem, or rather, a metadata problem
highlighted by packaged-staging.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFL8UoeMkyGM64RGpERAs4YAKCmiLtnoXDRCCODUCkS3EcSTCpDeACfTZoS
aQv46a+QTSUv3P+5iG/FB58=
=/KH3
-----END PGP SIGNATURE-----

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Khem Raj]

On Mon, May 17, 2010 at 2:38 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 17-05-10 07:55, Khem Raj wrote:
>> On (17/05/10 07:44), Koen Kooi wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>>
>>> Hi,
>>>
>>> I just noticed that after doing 'bitbake gcc' the old libstdc++ gets
>>> replaced and the new one has:
>>>
>>> dependency_libs='
>>> /OE/angstrom-dev/sysroots/armv6-angstrom-linux-gnueabi/usr/lib/libstdc++.la'
>>>
>>> instead of the expected '-lm'. It now refers to itself :(
>>
>> hmmm target gcc should not have staged libstdc++ if it was already staged
>> by gcc-cross but the problem is casued by
>> gcc-4.3.3-SYSROOT_CFLAGS_FOR_TARGET.patch it passed the native libdir which
>> is build_libdir even for target library builds using -rpath. I am currently
>> looking into the reason. If possible we can get rid of that patch.
>
> I'm getting a build failure with 'gcc' now failing to find libstc++.la.
> I'm trying a build from scratch now to rule out any packaged-staging
> problems (>1 recipes installing the same file, pstage removing it prior
> to build).

I could not build a console-image as I ran out of disk space. Have you
considered removing
-g3 from default angstrom builds.

-Khem

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 17-05-10 18:49, Khem Raj wrote:
> On Mon, May 17, 2010 at 2:38 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:

> I could not build a console-image as I ran out of disk space. Have you
> considered removing
> -g3 from default angstrom builds.

No, that would make -dbg packages useless.

regards,

Koen
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFL8XaIMkyGM64RGpERAkfFAKCsBlf/O9fnSkUxV+pOSiV2uXiBNwCfSjuu
iZwbcen+H0YuBVf7fTPXZ9Q=
=Cuo2
-----END PGP SIGNATURE-----

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Khem Raj]

On Mon, May 17, 2010 at 10:02 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 17-05-10 18:49, Khem Raj wrote:
>> On Mon, May 17, 2010 at 2:38 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
>
>> I could not build a console-image as I ran out of disk space. Have you
>> considered removing
>> -g3 from default angstrom builds.

what is -g3 offering over -g at -O2 and how good the additional debug info is ?

>
> No, that would make -dbg packages useless.
>
> regards,
>
> Koen
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.5 (Darwin)
>
> iD8DBQFL8XaIMkyGM64RGpERAkfFAKCsBlf/O9fnSkUxV+pOSiV2uXiBNwCfSjuu
> iZwbcen+H0YuBVf7fTPXZ9Q=
> =Cuo2
> -----END PGP SIGNATURE-----
>
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Tom Rini]

On Mon, 2010-05-17 at 10:45 -0700, Khem Raj wrote: 
> On Mon, May 17, 2010 at 10:02 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > On 17-05-10 18:49, Khem Raj wrote:
> >> On Mon, May 17, 2010 at 2:38 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> >
> >> I could not build a console-image as I ran out of disk space. Have you
> >> considered removing
> >> -g3 from default angstrom builds.
> 
> what is -g3 offering over -g at -O2 and how good the additional debug info is ?

-g3 offers lots of stuff, spelled out in some doc or another is macro
expansion, for example.  I don't know if it offers enough stuff that you
say don't need the sources themselves around still too tho.  Koen?

-- 
Tom Rini <tom_rini@mentor.com>
Mentor Graphics Corporation

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Khem Raj]

On (17/05/10 11:12), Tom Rini wrote:
> On Mon, 2010-05-17 at 10:45 -0700, Khem Raj wrote: 
> > On Mon, May 17, 2010 at 10:02 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> > > -----BEGIN PGP SIGNED MESSAGE-----
> > > Hash: SHA1
> > >
> > > On 17-05-10 18:49, Khem Raj wrote:
> > >> On Mon, May 17, 2010 at 2:38 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> > >
> > >> I could not build a console-image as I ran out of disk space. Have you
> > >> considered removing
> > >> -g3 from default angstrom builds.
> > 
> > what is -g3 offering over -g at -O2 and how good the additional debug info is ?
> 
> -g3 offers lots of stuff, spelled out in some doc or another is macro
> expansion, for example.  I don't know if it offers enough stuff that you
> say don't need the sources themselves around still too tho.  Koen?

If -g3 is intended, some options to reduce debug info size should be
considered like
-feliminate-dwarf2-dups -femit-struct-debug-baseonly
-femit-struct-debug-reduced
etc.

-Khem

Re: Question: is 'gcc' supposed to replace the libstdc++.la installed by 'gcc-cross'?

[Phil Blundell]

On Mon, 2010-05-17 at 11:12 -0700, Tom Rini wrote:
> On Mon, 2010-05-17 at 10:45 -0700, Khem Raj wrote: 
> > what is -g3 offering over -g at -O2 and how good the additional debug info is ?
> 
> -g3 offers lots of stuff, spelled out in some doc or another is macro
> expansion, for example.  I don't know if it offers enough stuff that you
> say don't need the sources themselves around still too tho.  Koen?

The gcc manual is not terribly illuminating on this subject:

http://gcc.gnu.org/onlinedocs/gcc/Debugging-Options.html#Debugging-Options

just says that it "... includes extra information, such as all the macro
definitions present in the program. Some debuggers support macro
expansion when you use -g3."

The definitions in flags.h suggest that, in fact, -g3 doesn't really add
much apart from macros:

enum debug_info_level
{
  DINFO_LEVEL_NONE,     /* Write no debugging info.  */
  DINFO_LEVEL_TERSE,    /* Write minimal info to support tracebacks only.  */
  DINFO_LEVEL_NORMAL,   /* Write info for all declarations (and line table).  */
  DINFO_LEVEL_VERBOSE   /* Write normal info plus #define/#undef info.  */
};

... and, from a quick inspection of dwarf2out.c, I didn't see anything
else very obvious that is added by DINFO_LEVEL_VERBOSE.

So, all in all, I can't find a great deal of evidence to suggest that
compiling at -g3 really does enhance the debugging experience all that
much.  I've certainly never used that option myself and haven't really
felt the poorer for it.  I'd be interested to hear what exactly the
feature is that the Angstrom folks feel the lack of renders the -dbg
packages unusable with -g2.

p.