[PATCH v1] site: enable pthread support for x86_64 APR builds

[PATCH v1] site: enable pthread support for x86_64 APR builds

[Ioan-Adrian Ratiu]

From: Jacob Champion <jacob.champion@ni.com>

The apache2 package was missing key pieces of pthread support (notably,
pthread_kill) on x86_64 platforms. Fix this by setting some cached
variables that are already set for other targets:

  apr_cv_pthreads_cflags=-pthread: enables pthread support for APR,
  which will trickle down into apache2 as well. powerpc32-linux
  currently sets this option.

  apr_cv_process_shared_works=no: several platforms already set this
  option, which conservatively disables support for PROCESS_SHARED
  mutexes.

Signed-off-by: Jacob Champion <jacob.champion@ni.com>
Signed-off-by: Ioan-Adrian Ratiu <adrian.ratiu@ni.com>
---
 meta/site/x86_64-linux | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/meta/site/x86_64-linux b/meta/site/x86_64-linux
index 573a907..9a67445 100644
--- a/meta/site/x86_64-linux
+++ b/meta/site/x86_64-linux
@@ -104,6 +104,8 @@ nfsd_cv_func_statfs=${nfsd_cv_func_statfs=statfs2_bsize}
 nfsd_cv_bsd_signals=${nfsd_cv_bsd_signals=yes}
 
 #apr
+apr_cv_process_shared_works=${apr_cv_process_shared_works=no}
+apr_cv_pthreads_cflags=${apr_cv_pthreads_cflags=-pthread}
 apr_cv_tcp_nodelay_with_cork=${apr_cv_tcp_nodelay_with_cork=yes}
 
 # lftp
-- 
2.7.1

Re: [PATCH v1] site: enable pthread support for x86_64 APR builds

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 451 bytes --]

On 29 February 2016 at 13:42, Ioan-Adrian Ratiu <adrian.ratiu@ni.com> wrote:

>   apr_cv_process_shared_works=no: several platforms already set this
>   option, which conservatively disables support for PROCESS_SHARED
>   mutexes.
>

Which is interesting as building apr-native (so the site files don't get
used) for x86-64 gives:

apr_cv_process_shared_works=yes

Can we not go around disabling stuff when it actually does work?

Ross

[-- Attachment #2: Type: text/html, Size: 1049 bytes --]

Re: [PATCH v1] site: enable pthread support for x86_64 APR builds

[Ioan-Adrian Ratiu]

On Mon, 29 Feb 2016 13:50:54 +0000
"Burton, Ross" <ross.burton@intel.com> wrote:

> On 29 February 2016 at 13:42, Ioan-Adrian Ratiu <adrian.ratiu@ni.com> wrote:
> 
> >   apr_cv_process_shared_works=no: several platforms already set this
> >   option, which conservatively disables support for PROCESS_SHARED
> >   mutexes.
> >
> 
> Which is interesting as building apr-native (so the site files don't get
> used) for x86-64 gives:
> 
> apr_cv_process_shared_works=yes
> 
> Can we not go around disabling stuff when it actually does work?

Yes, of course. It was just sloppy patch editing on my part. Sorry.
I'll remove that and resubmit v2.

> 
> Ross

Re: [PATCH v1] site: enable pthread support for x86_64 APR builds

[Khem Raj]

[-- Attachment #1.1: Type: text/plain, Size: 1117 bytes --]


> On Feb 29, 2016, at 5:50 AM, Burton, Ross <ross.burton@intel.com> wrote:
> 
> 
> On 29 February 2016 at 13:42, Ioan-Adrian Ratiu <adrian.ratiu@ni.com <mailto:adrian.ratiu@ni.com>> wrote:
>   apr_cv_process_shared_works=no: several platforms already set this
>   option, which conservatively disables support for PROCESS_SHARED
>   mutexes.
> 
> Which is interesting as building apr-native (so the site files don't get used) for x86-64 gives:
> 
> apr_cv_process_shared_works=yes
> 
> Can we not go around disabling stuff when it actually does work?

No it really doesnt work. its running a configure time test on your build host ( which I assume is x86_64 mostly now
a days) and then deciding that process shared mutexes work, which is not correct.

If we think that it should work on x86_64 then lets cache the value to ‘yes’ but not leave it
at the mercy of configure.

> 
> Ross
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core


[-- Attachment #1.2: Type: text/html, Size: 2513 bytes --]

[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 211 bytes --]

Re: [PATCH v1] site: enable pthread support for x86_64 APR builds

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 808 bytes --]

On 18 March 2016 at 00:22, Khem Raj <raj.khem@gmail.com> wrote:

> Which is interesting as building apr-native (so the site files don't get
> used) for x86-64 gives:
>
> apr_cv_process_shared_works=yes
>
> Can we not go around disabling stuff when it actually does work?
>
>
> No it really doesnt work. its running a configure time test on your build
> host ( which I assume is x86_64 mostly now
> a days) and then deciding that process shared mutexes work, which is not
> correct.
>
> If we think that it should work on x86_64 then lets cache the value to
> ‘yes’ but not leave it
> at the mercy of configure.
>

That was my point: the patch was to add works=no to x86-64 target builds,
but if I run the same test on x86-64 hardware directly then the result is
works=yes.

Ross

[-- Attachment #2: Type: text/html, Size: 1437 bytes --]

Re: [PATCH v1] site: enable pthread support for x86_64 APR builds

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1074 bytes --]

On Mar 18, 2016 4:38 AM, "Burton, Ross" <ross.burton@intel.com> wrote:
>
>
> On 18 March 2016 at 00:22, Khem Raj <raj.khem@gmail.com> wrote:
>>>
>>> Which is interesting as building apr-native (so the site files don't
get used) for x86-64 gives:
>>>
>>> apr_cv_process_shared_works=yes
>>>
>>> Can we not go around disabling stuff when it actually does work?
>>
>>
>> No it really doesnt work. its running a configure time test on your
build host ( which I assume is x86_64 mostly now
>> a days) and then deciding that process shared mutexes work, which is not
correct.
>>
>> If we think that it should work on x86_64 then lets cache the value to
‘yes’ but not leave it
>> at the mercy of configure.
>
>
> That was my point: the patch was to add works=no to x86-64 target builds,
but if I run the same test on x86-64 hardware directly then the result is
works=yes.

Actually it depends on libc implementation of process shared mutexes. So
glibc for most arches should set it yes. Musl doesnt have it. So that
should be no mostly
>
> Ross

[-- Attachment #2: Type: text/html, Size: 1419 bytes --]