* Cached size of off_t
@ 2016-01-12 23:23 Khem Raj
2016-01-13 0:09 ` [Openembedded-architecture] " Paul Eggleton
2016-01-14 11:33 ` Martin Jansa
0 siblings, 2 replies; 6+ messages in thread
From: Khem Raj @ 2016-01-12 23:23 UTC (permalink / raw)
To: openembedded-architecture; +Cc: OE-core
[-- Attachment #1: Type: text/plain, Size: 1117 bytes --]
Hi
We cache size of off_t in site files and the value is 4 in most cases
meta/site/arm-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
meta/site/ix86-common:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
meta/site/nios2-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
meta/site/powerpc32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
meta/site/x32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
meta/site/x86_64-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
However this value is wrong when largefile support is enabled (which is default) on
most OE based distros. It also lands us in nasty bugs which are complicated to uncover
eg. the curl issue
http://git.openembedded.org/openembedded-core/commit/?id=3c58712465494e441c4036a7cf21d2e6d343efab
and just in OE-Core more e.g. grub2,libsndfile1,pax are showing same symptoms.
I think this value should be left for configure to compute and
autotools.bbclass should pass --{enable|disable}-largefile based upon
DISTRO_FEATURES enabling largefile support
Do we see any downsides or point I might have missed
Thank you
-Khem
[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 211 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [Openembedded-architecture] Cached size of off_t
2016-01-12 23:23 Cached size of off_t Khem Raj
@ 2016-01-13 0:09 ` Paul Eggleton
2016-01-13 0:19 ` Khem Raj
2016-01-13 9:00 ` Burton, Ross
2016-01-14 11:33 ` Martin Jansa
1 sibling, 2 replies; 6+ messages in thread
From: Paul Eggleton @ 2016-01-13 0:09 UTC (permalink / raw)
To: Khem Raj; +Cc: openembedded-architecture, OE-core
On Tue, 12 Jan 2016 15:23:33 Khem Raj wrote:
> Hi
>
> We cache size of off_t in site files and the value is 4 in most cases
>
>
> meta/site/arm-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/ix86-common:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/nios2-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/powerpc32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/x32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
> meta/site/x86_64-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
>
>
> However this value is wrong when largefile support is enabled (which is
> default) on most OE based distros. It also lands us in nasty bugs which are
> complicated to uncover eg. the curl issue
>
> http://git.openembedded.org/openembedded-core/commit/?id=3c58712465494e441c4
> 036a7cf21d2e6d343efab
>
> and just in OE-Core more e.g. grub2,libsndfile1,pax are showing same
> symptoms.
>
> I think this value should be left for configure to compute and
> autotools.bbclass should pass --{enable|disable}-largefile based upon
> DISTRO_FEATURES enabling largefile support
>
> Do we see any downsides or point I might have missed
IIRC I think the issue was older configure scripts without the capability to
compute this properly (where we're not auto-reconf'ing them).
Ross, any thoughts?
Cheers,
Paul
--
Paul Eggleton
Intel Open Source Technology Centre
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [Openembedded-architecture] Cached size of off_t
2016-01-13 0:09 ` [Openembedded-architecture] " Paul Eggleton
@ 2016-01-13 0:19 ` Khem Raj
2016-01-13 0:22 ` Andre McCurdy
2016-01-13 9:00 ` Burton, Ross
1 sibling, 1 reply; 6+ messages in thread
From: Khem Raj @ 2016-01-13 0:19 UTC (permalink / raw)
To: Paul Eggleton; +Cc: openembedded-architecture, OE-core
On Tue, Jan 12, 2016 at 4:09 PM, Paul Eggleton
<paul.eggleton@linux.intel.com> wrote:
> On Tue, 12 Jan 2016 15:23:33 Khem Raj wrote:
>> Hi
>>
>> We cache size of off_t in site files and the value is 4 in most cases
>>
>>
>> meta/site/arm-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>> meta/site/ix86-common:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>> meta/site/nios2-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>> meta/site/powerpc32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>> meta/site/x32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
>> meta/site/x86_64-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
>>
>>
>> However this value is wrong when largefile support is enabled (which is
>> default) on most OE based distros. It also lands us in nasty bugs which are
>> complicated to uncover eg. the curl issue
>>
>> http://git.openembedded.org/openembedded-core/commit/?id=3c58712465494e441c4
>> 036a7cf21d2e6d343efab
>>
>> and just in OE-Core more e.g. grub2,libsndfile1,pax are showing same
>> symptoms.
>>
>> I think this value should be left for configure to compute and
>> autotools.bbclass should pass --{enable|disable}-largefile based upon
>> DISTRO_FEATURES enabling largefile support
>>
>> Do we see any downsides or point I might have missed
>
> IIRC I think the issue was older configure scripts without the capability to
> compute this properly (where we're not auto-reconf'ing them).
Thanks Paul.
I think in that case it is better to move that into the concerned
recipe and let defaults be reversed
>
> Ross, any thoughts?
>
> Cheers,
> Paul
>
> --
>
> Paul Eggleton
> Intel Open Source Technology Centre
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [Openembedded-architecture] Cached size of off_t
2016-01-13 0:19 ` Khem Raj
@ 2016-01-13 0:22 ` Andre McCurdy
0 siblings, 0 replies; 6+ messages in thread
From: Andre McCurdy @ 2016-01-13 0:22 UTC (permalink / raw)
To: Khem Raj; +Cc: Paul Eggleton, openembedded-architecture, OE-core
On Tue, Jan 12, 2016 at 4:19 PM, Khem Raj <raj.khem@gmail.com> wrote:
> On Tue, Jan 12, 2016 at 4:09 PM, Paul Eggleton
> <paul.eggleton@linux.intel.com> wrote:
>> On Tue, 12 Jan 2016 15:23:33 Khem Raj wrote:
>>> Hi
>>>
>>> We cache size of off_t in site files and the value is 4 in most cases
>>>
>>>
>>> meta/site/arm-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>>> meta/site/ix86-common:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>>> meta/site/nios2-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>>> meta/site/powerpc32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
>>> meta/site/x32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
>>> meta/site/x86_64-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
>>>
>>>
>>> However this value is wrong when largefile support is enabled (which is
>>> default) on most OE based distros. It also lands us in nasty bugs which are
>>> complicated to uncover eg. the curl issue
>>>
>>> http://git.openembedded.org/openembedded-core/commit/?id=3c58712465494e441c4
>>> 036a7cf21d2e6d343efab
>>>
>>> and just in OE-Core more e.g. grub2,libsndfile1,pax are showing same
>>> symptoms.
>>>
>>> I think this value should be left for configure to compute and
>>> autotools.bbclass should pass --{enable|disable}-largefile based upon
>>> DISTRO_FEATURES enabling largefile support
>>>
>>> Do we see any downsides or point I might have missed
>>
>> IIRC I think the issue was older configure scripts without the capability to
>> compute this properly (where we're not auto-reconf'ing them).
>
> Thanks Paul.
> I think in that case it is better to move that into the concerned
> recipe and let defaults be reversed
How do these older configure scripts work today for MIPS and SH4 which
don't seem to define ac_cv_sizeof_off_t in their site files?
>>
>> Ross, any thoughts?
>>
>> Cheers,
>> Paul
>>
>> --
>>
>> Paul Eggleton
>> Intel Open Source Technology Centre
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Openembedded-architecture] Cached size of off_t
2016-01-13 0:09 ` [Openembedded-architecture] " Paul Eggleton
2016-01-13 0:19 ` Khem Raj
@ 2016-01-13 9:00 ` Burton, Ross
1 sibling, 0 replies; 6+ messages in thread
From: Burton, Ross @ 2016-01-13 9:00 UTC (permalink / raw)
To: Paul Eggleton; +Cc: openembedded-architecture, OE-core
[-- Attachment #1: Type: text/plain, Size: 1012 bytes --]
On 13 January 2016 at 00:09, Paul Eggleton <paul.eggleton@linux.intel.com>
wrote:
> > I think this value should be left for configure to compute and
> > autotools.bbclass should pass --{enable|disable}-largefile based upon
> > DISTRO_FEATURES enabling largefile support
> >
> > Do we see any downsides or point I might have missed
>
> IIRC I think the issue was older configure scripts without the capability
> to
> compute this properly (where we're not auto-reconf'ing them).
>
The fact that it changes based on configure options says that it really
shouldn't be a cached value.
It's very unlikely that there's configure scripts out there still that
don't use the "modern" autoconf type size detection and if there are
they're trivially patched to use the new macros.
Remove it, as this is going to cause very strange problems as large files
get enabled more globally. In particular I'm now worrying about DBus which
just recently got a patch to enable large file support.
Ross
[-- Attachment #2: Type: text/html, Size: 1673 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Openembedded-architecture] Cached size of off_t
2016-01-12 23:23 Cached size of off_t Khem Raj
2016-01-13 0:09 ` [Openembedded-architecture] " Paul Eggleton
@ 2016-01-14 11:33 ` Martin Jansa
1 sibling, 0 replies; 6+ messages in thread
From: Martin Jansa @ 2016-01-14 11:33 UTC (permalink / raw)
To: Khem Raj; +Cc: openembedded-architecture, OE-core
[-- Attachment #1: Type: text/plain, Size: 1940 bytes --]
On Tue, Jan 12, 2016 at 03:23:33PM -0800, Khem Raj wrote:
> Hi
>
> We cache size of off_t in site files and the value is 4 in most cases
>
>
> meta/site/arm-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/ix86-common:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/nios2-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/powerpc32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=4}
> meta/site/x32-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
> meta/site/x86_64-linux:ac_cv_sizeof_off_t=${ac_cv_sizeof_off_t=8}
>
>
> However this value is wrong when largefile support is enabled (which is default) on
> most OE based distros. It also lands us in nasty bugs which are complicated to uncover
> eg. the curl issue
>
> http://git.openembedded.org/openembedded-core/commit/?id=3c58712465494e441c4036a7cf21d2e6d343efab
>
> and just in OE-Core more e.g. grub2,libsndfile1,pax are showing same symptoms.
It's even more complicated in curl case, which bumps SONAME version when
off_t is changed.
So with dizzy we had libcurl5, but after the commit above we get
libcurl4 which looks strange considering the curl version in oe-core is newer now.
You can force the bump with --enable-soname-bump, but that doesn't make
it ABI compatible, because SIZEOF_OFF_T was really changed inside curl
build.
Regards,
> I think this value should be left for configure to compute and
> autotools.bbclass should pass --{enable|disable}-largefile based upon
> DISTRO_FEATURES enabling largefile support
>
> Do we see any downsides or point I might have missed
>
> Thank you
> -Khem
>
> _______________________________________________
> Openembedded-architecture mailing list
> Openembedded-architecture@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-architecture
--
Martin 'JaMa' Jansa jabber: Martin.Jansa@gmail.com
[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2016-01-14 11:32 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-01-12 23:23 Cached size of off_t Khem Raj
2016-01-13 0:09 ` [Openembedded-architecture] " Paul Eggleton
2016-01-13 0:19 ` Khem Raj
2016-01-13 0:22 ` Andre McCurdy
2016-01-13 9:00 ` Burton, Ross
2016-01-14 11:33 ` Martin Jansa
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.