Patchwork [0/3] siteinfo.bbclass re-merge

login
register
mail settings
Submitter Tom Rini
Date July 13, 2011, 7:05 p.m.
Message ID <cover.1310583666.git.tom_rini@mentor.com>
Download mbox
Permalink /patch/7567/
State New, archived
Headers show

Pull-request

git://git.openembedded.org/openembedded-core-contrib trini/update-siteinfo

Comments

Tom Rini - July 13, 2011, 7:05 p.m.
Hey all,

This is the second of two pull requests.  This doesn't depend on the first
request, it's just splitting changes up.  There's three changes here.  First,
site/ix86-common was setting the wrong value for unsigned int.  Second,
change perl to use SITEINFO variables rather than functions.  Third, bring
in the various changes in siteinfo.bbclass that have been in oe.dev.  This
last part will, I hope, also make dealing with multilib-specific siteinfo
stuff easier (also alternative ABIs like x32).

My next step is to work on merging and verifying as much as possible
the site files in oe-core and oe.dev.

The following changes since commit e66c2999afa2b3efbce8bb46c89f9db5e15f35c7:
  Saul Wold (1):
        libx11: ensure nativesdk uses correct DEPENDS and XCB flags

are available in the git repository at:

  git://git.openembedded.org/openembedded-core-contrib trini/update-siteinfo
  http://cgit.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=trini/update-siteinfo

Tom Rini (3):
  site/ix86-common: Fix ac_cv_sizeof_unsigned_char definition
  perl: Use SITEINFO variables not functions
  siteinfo.bbclass: Port over oe.dev logic for site files

 meta/classes/autotools.bbclass             |    2 +-
 meta/classes/siteinfo.bbclass              |  202 +++++++++++++---------------
 meta/classes/toolchain-scripts.bbclass     |    2 +-
 meta/recipes-core/meta/meta-environment.bb |    2 +-
 meta/recipes-core/meta/meta-toolchain.bb   |    2 +-
 meta/recipes-devtools/perl/perl_5.12.3.bb  |    4 +-
 meta/site/ix86-common                      |    2 +-
 7 files changed, 103 insertions(+), 113 deletions(-)
 create mode 100644 meta/site/common-linux
Mark Hatle - July 13, 2011, 7:42 p.m.
On 7/13/11 2:05 PM, Tom Rini wrote:
> Hey all,
> 
> This is the second of two pull requests.  This doesn't depend on the first
> request, it's just splitting changes up.  There's three changes here.  First,
> site/ix86-common was setting the wrong value for unsigned int.  Second,
> change perl to use SITEINFO variables rather than functions.  Third, bring
> in the various changes in siteinfo.bbclass that have been in oe.dev.  This
> last part will, I hope, also make dealing with multilib-specific siteinfo
> stuff easier (also alternative ABIs like x32).

Note, this is NOT an objection to the changes.  However, one thing to keep in mind.

(In fact I like the changes being suggested)

We have a dynamic site-configuation mechanism already in place.  So it's
important that we don't duplicate things within the site/* files.

Currently the recipes: glibc, zlib, eglibc, and uclibc contain dynamic
site_config files.  (See ./recipes-core/glibc/site_config/* for an example)

The output of these files get placed into a target arch specific site config
directory: ${STAGING_DATADIR}/${TARGET_SYS}_config_site.d

These are automatically loaded as well as the predefined "common" items.  By
using the dynamically generated versions we can change the configuration options
for glibc, uclibc, etc and the site config will automatically get updated based
on the existing checks.

So instead of updating the various site/* files, if the value can be discovered
-- or is defined by a given package.. we should be using the dynamic method as
it'll be safer over the long-haul.  Someone might need to go over the existing
ones and see if there is anything that falls under this.

BTW, I did notice that ncurses is supposed to have a dynamic site_config, but it
somehow got checked into the wrong location.  I'll submit a pull request for
that shortly.

--Mark

> My next step is to work on merging and verifying as much as possible
> the site files in oe-core and oe.dev.
> 
> The following changes since commit e66c2999afa2b3efbce8bb46c89f9db5e15f35c7:
>   Saul Wold (1):
>         libx11: ensure nativesdk uses correct DEPENDS and XCB flags
> 
> are available in the git repository at:
> 
>   git://git.openembedded.org/openembedded-core-contrib trini/update-siteinfo
>   http://cgit.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=trini/update-siteinfo
> 
> Tom Rini (3):
>   site/ix86-common: Fix ac_cv_sizeof_unsigned_char definition
>   perl: Use SITEINFO variables not functions
>   siteinfo.bbclass: Port over oe.dev logic for site files
> 
>  meta/classes/autotools.bbclass             |    2 +-
>  meta/classes/siteinfo.bbclass              |  202 +++++++++++++---------------
>  meta/classes/toolchain-scripts.bbclass     |    2 +-
>  meta/recipes-core/meta/meta-environment.bb |    2 +-
>  meta/recipes-core/meta/meta-toolchain.bb   |    2 +-
>  meta/recipes-devtools/perl/perl_5.12.3.bb  |    4 +-
>  meta/site/ix86-common                      |    2 +-
>  7 files changed, 103 insertions(+), 113 deletions(-)
>  create mode 100644 meta/site/common-linux
> 
> 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
Tom Rini - July 13, 2011, 8:40 p.m.
On 07/13/2011 12:42 PM, Mark Hatle wrote:
> On 7/13/11 2:05 PM, Tom Rini wrote:
>> Hey all,
>>
>> This is the second of two pull requests.  This doesn't depend on the first
>> request, it's just splitting changes up.  There's three changes here.  First,
>> site/ix86-common was setting the wrong value for unsigned int.  Second,
>> change perl to use SITEINFO variables rather than functions.  Third, bring
>> in the various changes in siteinfo.bbclass that have been in oe.dev.  This
>> last part will, I hope, also make dealing with multilib-specific siteinfo
>> stuff easier (also alternative ABIs like x32).
> 
> Note, this is NOT an objection to the changes.  However, one thing to keep in mind.
> 
> (In fact I like the changes being suggested)
> 
> We have a dynamic site-configuation mechanism already in place.  So it's
> important that we don't duplicate things within the site/* files.
> 
> Currently the recipes: glibc, zlib, eglibc, and uclibc contain dynamic
> site_config files.  (See ./recipes-core/glibc/site_config/* for an example)
> 
> The output of these files get placed into a target arch specific site config
> directory: ${STAGING_DATADIR}/${TARGET_SYS}_config_site.d

I want to confirm that these files are still generated, found and
working.  But maybe part 4 of my series will be talking about some of
this stuff a bit more and if that's really the best method for the goal.
Richard Purdie - July 14, 2011, 2:42 p.m.
On Wed, 2011-07-13 at 12:05 -0700, Tom Rini wrote:
> Hey all,
> 
> This is the second of two pull requests.  This doesn't depend on the first
> request, it's just splitting changes up.  There's three changes here.  First,
> site/ix86-common was setting the wrong value for unsigned int.  Second,
> change perl to use SITEINFO variables rather than functions.  Third, bring
> in the various changes in siteinfo.bbclass that have been in oe.dev.  This
> last part will, I hope, also make dealing with multilib-specific siteinfo
> stuff easier (also alternative ABIs like x32).
> 
> My next step is to work on merging and verifying as much as possible
> the site files in oe-core and oe.dev.
> 
> The following changes since commit e66c2999afa2b3efbce8bb46c89f9db5e15f35c7:
>   Saul Wold (1):
>         libx11: ensure nativesdk uses correct DEPENDS and XCB flags
> 
> are available in the git repository at:
> 
>   git://git.openembedded.org/openembedded-core-contrib trini/update-siteinfo
>   http://cgit.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=trini/update-siteinfo
> 
> Tom Rini (3):
>   site/ix86-common: Fix ac_cv_sizeof_unsigned_char definition
>   perl: Use SITEINFO variables not functions

I've merged these two.

>   siteinfo.bbclass: Port over oe.dev logic for site files

I'm still in the process of reviewing this as its not easy to see
everything thats changing in there...

Cheers,

Richard
Tom Rini - July 14, 2011, 4:58 p.m.
On 07/14/2011 07:42 AM, Richard Purdie wrote:
> On Wed, 2011-07-13 at 12:05 -0700, Tom Rini wrote:
>> Hey all,
>>
>> This is the second of two pull requests.  This doesn't depend on the first
>> request, it's just splitting changes up.  There's three changes here.  First,
>> site/ix86-common was setting the wrong value for unsigned int.  Second,
>> change perl to use SITEINFO variables rather than functions.  Third, bring
>> in the various changes in siteinfo.bbclass that have been in oe.dev.  This
>> last part will, I hope, also make dealing with multilib-specific siteinfo
>> stuff easier (also alternative ABIs like x32).
>>
>> My next step is to work on merging and verifying as much as possible
>> the site files in oe-core and oe.dev.
>>
>> The following changes since commit e66c2999afa2b3efbce8bb46c89f9db5e15f35c7:
>>   Saul Wold (1):
>>         libx11: ensure nativesdk uses correct DEPENDS and XCB flags
>>
>> are available in the git repository at:
>>
>>   git://git.openembedded.org/openembedded-core-contrib trini/update-siteinfo
>>   http://cgit.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=trini/update-siteinfo
>>
>> Tom Rini (3):
>>   site/ix86-common: Fix ac_cv_sizeof_unsigned_char definition
>>   perl: Use SITEINFO variables not functions
> 
> I've merged these two.
> 
>>   siteinfo.bbclass: Port over oe.dev logic for site files
> 
> I'm still in the process of reviewing this as its not easy to see
> everything thats changing in there...

Yeah, sorry, I just couldn't break it down anymore.  I can say I did a
world before and after on qemux86, diff'd config.logs and didn't see
anything other than the ix86-common bugfix in differences.