* build inconsistency (strip)
@ 2009-10-02 7:10 Frans Meulenbroeks
2009-10-02 8:02 ` Koen Kooi
0 siblings, 1 reply; 6+ messages in thread
From: Frans Meulenbroeks @ 2009-10-02 7:10 UTC (permalink / raw)
To: openembedded-devel
Hi,
I bumped on something very odd.
I wanted to look at an issue with proftpd. As the API version had
changed, I had to remove my tmp dir.
I did bitbake proftpd, fixed the error, bumped PR, everything fine.
Then I asked Koen to verify my work and decide whether it could be
added to the angstrom feed.
At that point things became confusing. For Koen the build failed due
to install wanting to strip and indeed there was something in
Make.Rules that eventually resulted in an install -s.
My question/worry: why am I getting different results than Koen gets.
And what can we do to avoid this?
Of course it could also be dependent on my host env (b.t.w. I'm using
opensuse 11.1), but my impression was that things were supposed to be
host independent.
BTW: given some of the patches that are needed, would it perhaps be
better to have an oe/bin (or oe/native/bin or whatever) dir somewhere
with our own copies of things like strip, install, uname etc and have
this dir in front of the path when running bitbake?
Frans.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: build inconsistency (strip)
2009-10-02 7:10 build inconsistency (strip) Frans Meulenbroeks
@ 2009-10-02 8:02 ` Koen Kooi
2009-10-02 10:28 ` Frans Meulenbroeks
0 siblings, 1 reply; 6+ messages in thread
From: Koen Kooi @ 2009-10-02 8:02 UTC (permalink / raw)
To: openembedded-devel
On 02-10-09 09:10, Frans Meulenbroeks wrote:
> At that point things became confusing. For Koen the build failed due
> to install wanting to strip and indeed there was something in
> Make.Rules that eventually resulted in an install -s.
To clarify: it wanted to run 'strip', not '${TARGET_PREFIX}-strip',
which caused it to fail. Running '${TARGET_PREFIX}-strip' would have
only inconvinienced people trying to do runtime debugging.
regards,
Koen
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: build inconsistency (strip)
2009-10-02 8:02 ` Koen Kooi
@ 2009-10-02 10:28 ` Frans Meulenbroeks
2009-10-15 14:35 ` Leon Woestenberg
0 siblings, 1 reply; 6+ messages in thread
From: Frans Meulenbroeks @ 2009-10-02 10:28 UTC (permalink / raw)
To: openembedded-devel
Koen, all, did some research on this.
On my system proftpd wanted to use an INSTALL macro and added the -s flag..
This macro evaluated to
/home/frans/oe/tmp_angstrom/staging/i686-linux/usr/bin/install -c
So it did strip on my system.
What I understood from Koen was that he got an error message on strip.
Either I misunderstood or there is an incompatibility here.
(and if there is an inconsistency I'd like my system to have the error
message too).
Still the following is strange to me:
I would expect
/home/frans/oe/tmp_angstrom/staging/i686-linux/usr/bin/install -s to
call the target strip (or maybe ignore the strip command).
Frans.
2009/10/2 Koen Kooi <k.kooi@student.utwente.nl>:
> On 02-10-09 09:10, Frans Meulenbroeks wrote:
>
>> At that point things became confusing. For Koen the build failed duke
>> to install wanting to strip and indeed there was something in
>> Make.Rules that eventually resulted in an install -s.
>
> To clarify: it wanted to run 'strip', not '${TARGET_PREFIX}-strip', which
> caused it to fail. Running '${TARGET_PREFIX}-strip' would have only
> inconvinienced people trying to do runtime debugging.
>
> regards,
>
> Koen
>
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: build inconsistency (strip)
2009-10-02 10:28 ` Frans Meulenbroeks
@ 2009-10-15 14:35 ` Leon Woestenberg
2009-10-15 14:58 ` Andrea Adami
0 siblings, 1 reply; 6+ messages in thread
From: Leon Woestenberg @ 2009-10-15 14:35 UTC (permalink / raw)
To: openembedded-devel
Hello,
On Fri, Oct 2, 2009 at 12:28 PM, Frans Meulenbroeks
<fransmeulenbroeks@gmail.com> wrote:
> Still the following is strange to me:
> I would expect
> /home/frans/oe/tmp_angstrom/staging/i686-linux/usr/bin/install -s to
> call the target strip (or maybe ignore the strip command).
>
No, AFAIK install is unaware of the preferred strip executable.
Some target architecture binaries can be stripped with the host strip
command (for example, this seems to work for x86->arm, but NOT
x86->powerpc).
One proposal (from me) was to patch core-utils to use env. var. STRIP
instead of strip.
However, the OpenEmbedded solution is to patch the Makefiles to remove
the "-s" flag from the install targets.
This approach is best as the stripping is then done neatly during
packaging the debug strip-off.
Regards,
--
Leon
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: build inconsistency (strip)
2009-10-15 14:35 ` Leon Woestenberg
@ 2009-10-15 14:58 ` Andrea Adami
2009-10-15 15:09 ` Leon Woestenberg
0 siblings, 1 reply; 6+ messages in thread
From: Andrea Adami @ 2009-10-15 14:58 UTC (permalink / raw)
To: openembedded-devel
>Some target architecture binaries can be stripped with the host strip
>command (for example, this seems to work for x86->arm, but NOT
>x86->powerpc).
Don't rely on this: I had recently a recipe (dietlibc) failing when
using the host strip (i686) for arm binaries.
This is failing on Gentoo but reported to work on debian/Ubuntu
(multiarch-strip).
Regards
Andrea
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: build inconsistency (strip)
2009-10-15 14:58 ` Andrea Adami
@ 2009-10-15 15:09 ` Leon Woestenberg
0 siblings, 0 replies; 6+ messages in thread
From: Leon Woestenberg @ 2009-10-15 15:09 UTC (permalink / raw)
To: openembedded-devel
Hello,
On Thu, Oct 15, 2009 at 4:58 PM, Andrea Adami <andrea.adami@gmail.com> wrote:
>>Some target architecture binaries can be stripped with the host strip
>>command (for example, this seems to work for x86->arm, but NOT
>>x86->powerpc).
>
> Don't rely on this: I had recently a recipe (dietlibc) failing when
> using the host strip (i686) for arm binaries.
>
> This is failing on Gentoo but reported to work on debian/Ubuntu
> (multiarch-strip).
>
I didn't mean to say we should rely on this; I meant to say that for
some it may work, for some it may not, and thus MUST NOT BE USED.
So to summarize:
- The OpenEmbedded solution is to patch the Makefiles to remove the
"-s" flag from the install targets.
- This approach is best as the stripping is then done neatly during
packaging the debug strip-off.
Regards,
--
Leon
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2009-10-15 15:10 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-10-02 7:10 build inconsistency (strip) Frans Meulenbroeks
2009-10-02 8:02 ` Koen Kooi
2009-10-02 10:28 ` Frans Meulenbroeks
2009-10-15 14:35 ` Leon Woestenberg
2009-10-15 14:58 ` Andrea Adami
2009-10-15 15:09 ` Leon Woestenberg
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.