All of lore.kernel.org
 help / color / mirror / Atom feed
* [Xenomai] obstack support with uClibc-ng >= 1.0.21
@ 2017-04-14 13:48 Romain Naour
  2017-04-15 15:08 ` Philippe Gerum
  0 siblings, 1 reply; 3+ messages in thread
From: Romain Naour @ 2017-04-14 13:48 UTC (permalink / raw)
  To: xenomai; +Cc: Waldemar Brodkorb, Pawel Sikora

Hi all,

Xenomai 3 doesn't build with recent uClibc-ng (since 1.0.21 release) since the
obstack support has been removed [1].

I proposed a patch [2] disabling this feature unconditionally but I don't really
understand how obstack support work with Xenomai and why it's required.

Do we need to add a requirement on Glibc system as suggested in [3] ?

Best regards,
Romain

[1]
https://cgit.openadk.org/cgi/cgit/uclibc-ng.git/commit/?id=0bd6bfb2b643ea2b4b1440dfd917ba752f0c0d15
[2] http://patchwork.ozlabs.org/patch/745792/
[3] http://lists.busybox.net/pipermail/buildroot/2017-April/189244.html


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [Xenomai] obstack support with uClibc-ng >= 1.0.21
  2017-04-14 13:48 [Xenomai] obstack support with uClibc-ng >= 1.0.21 Romain Naour
@ 2017-04-15 15:08 ` Philippe Gerum
  2017-04-15 17:58   ` Romain Naour
  0 siblings, 1 reply; 3+ messages in thread
From: Philippe Gerum @ 2017-04-15 15:08 UTC (permalink / raw)
  To: Romain Naour, xenomai; +Cc: Waldemar Brodkorb, Pawel Sikora

On 04/14/2017 03:48 PM, Romain Naour wrote:
> Hi all,
> 
> Xenomai 3 doesn't build with recent uClibc-ng (since 1.0.21 release) since the
> obstack support has been removed [1].
> 
> I proposed a patch [2] disabling this feature unconditionally but I don't really
> understand how obstack support work with Xenomai and why it's required.
> 
> Do we need to add a requirement on Glibc system as suggested in [3] ?
> 

obstacks are used internally by helpers from our fuse-based registry
system (lib/copperplate/registry.c). No rocket-science in there, but not
a wheel that should be reinvented either, so the obstack support was
pulled from a gcc-based libiberty copy IIRC. Fact is that the code
assessing the opportunity to elide that code is terminally broken, as
you found out.

I pushed a commit [1] aimed at addressing the issue, by detecting
whether the underlying libc provides for native obstack support, only
building our replacement code if not, with some cleanups. This as been
tested against uClibc 1.0.22, and common glibc releases.

PS: I pushed some changes toward supporting musl to the stable branch as
well, but since the latter hides linux-specific ABI bits such as
SIGEV_THREAD_ID, full support requires more effort. So musl support is
still not there yet.

[1]
http://git.xenomai.org/xenomai-3.git/commit/?h=stable-3.0.x&id=4c593544d2ff39ba6631cc3fa0b91493973674d4

-- 
Philippe.


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [Xenomai] obstack support with uClibc-ng >= 1.0.21
  2017-04-15 15:08 ` Philippe Gerum
@ 2017-04-15 17:58   ` Romain Naour
  0 siblings, 0 replies; 3+ messages in thread
From: Romain Naour @ 2017-04-15 17:58 UTC (permalink / raw)
  To: Philippe Gerum, xenomai; +Cc: Waldemar Brodkorb, Pawel Sikora

Hi Philippe,

Le 15/04/2017 à 17:08, Philippe Gerum a écrit :
> On 04/14/2017 03:48 PM, Romain Naour wrote:
>> Hi all,
>>
>> Xenomai 3 doesn't build with recent uClibc-ng (since 1.0.21 release) since the
>> obstack support has been removed [1].
>>
>> I proposed a patch [2] disabling this feature unconditionally but I don't really
>> understand how obstack support work with Xenomai and why it's required.
>>
>> Do we need to add a requirement on Glibc system as suggested in [3] ?
>>
> 
> obstacks are used internally by helpers from our fuse-based registry
> system (lib/copperplate/registry.c). No rocket-science in there, but not
> a wheel that should be reinvented either, so the obstack support was
> pulled from a gcc-based libiberty copy IIRC. Fact is that the code
> assessing the opportunity to elide that code is terminally broken, as
> you found out.
> 
> I pushed a commit [1] aimed at addressing the issue, by detecting
> whether the underlying libc provides for native obstack support, only
> building our replacement code if not, with some cleanups. This as been
> tested against uClibc 1.0.22, and common glibc releases.

Thanks for your quick feedback!
I sent a patch to Buildroot to backport this fix and revert my workaround.

> 
> PS: I pushed some changes toward supporting musl to the stable branch as
> well, but since the latter hides linux-specific ABI bits such as
> SIGEV_THREAD_ID, full support requires more effort. So musl support is
> still not there yet.

Last year (08/2016), I disabled Xenomai (v2.6.4) package in Buildroot for musl
based toolchain [1] after trying to fixes all issues...
Since then nobody complained that Xenomai is disabled with musl, I guess we can
wait until the musl port is complete and released :)

Best regards,
Romain

[1]
https://git.busybox.net/buildroot/commit/?id=b845a1541f4c40e028a7351ca183ea3c6db56246

> 
> [1]
> http://git.xenomai.org/xenomai-3.git/commit/?h=stable-3.0.x&id=4c593544d2ff39ba6631cc3fa0b91493973674d4
> 



^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2017-04-15 17:58 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-04-14 13:48 [Xenomai] obstack support with uClibc-ng >= 1.0.21 Romain Naour
2017-04-15 15:08 ` Philippe Gerum
2017-04-15 17:58   ` Romain Naour

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.