[OE-core] Fwd: [oe] Source Archiver Class

Saul Wold sgw at linux.intel.com
Tue Feb 21 16:57:48 UTC 2012


On 02/19/2012 10:40 PM, Xiaofeng Yan wrote:
> On 2012年02月18日 07:52, Saul Wold wrote:
>>
>> Xiaofeng & Community:
>>
>> We had a chat with Chris Larson and Mark Hatle here at ELC.
>>
>> We focused this issue down to a fewer number of options that meet the
>> needs of the licenses. Please review this and let us know if there are
>> any issues or concerns.
>>
>>
> Hi Saul,
>
> I comment my understanding as follow:
>> Best Practices:
>> - Archive during build, we do not support nor recommend post archiving
>> - Original tarball and patches w/ordering file (maybe comment the
>> series), grab non-patch files also.
> for example:
> file://a.patch
> file://b.config.in
> we also archive b.config.in.

Yes, correct.

>> - Unless requesting configured sources, which will just be the
>> configured source tarball
>> - Grab temp as postfunc of do_package - latest logs via links & pid
>>
> That means logs package include all of logs in temp. for example.
> You has described a following function we will realize at the previous
> email.
> 3 - Original Source code & Patches & temp (scripts & logs)
>
> source codes and patches should be in the stage do_patch[postfunc] = "
> do_get_source"
> logs should be in the stage do_package[postfunc] = " do_get_logs"
> So logs file shouldn't be archived in do_patch[postfunc] = " do_get_logs"
> The logs in the stage of do_packge are more than do_patch.
>
Correct, you want to get the logs and scripts from temp when they are 
most complete (after do_package not after do_patch).

As mentioned below, you would use the variable SOURCE_ARCHIVE_LOG in the 
do_package[postfunc] to determine if you need to archive the logs or not.


>> For sstate Builds (LATER):
>> - need to verify that fetch/patch/configure will get re-run for archiving
>> - Add temp dir to sstate capture (without links)
>>
>> 3 Implementations classes
>> - Filter in oe/lib/license.py
>> - source_archive.bb
>> - archives tarballs to ${BP}/...
>> - reuse copyleft_compliance for patch handling
>> - srpm
>> - take output from above and create SRPM - (LATER)
>>
> We will do the archive work according the license . I will implement two
> functions, one is for left, the other for right.
I am not sure what you mean by one is for left and the other is for right?

Can you explain your thoughts here.


>> 2 Configuration classes
>> - uses prefunc/postfunc at correct place
>> - Original Tarball / Patches
>> - Post Configuration
>>
> We can define 8 classes to complete 8 kinds of archiving methods. every
> class will inherit archiver.bbclass.

I am not sure we need 8 classes anymore, that was the point of this 
email to simplify the needs.

2 Classes:

source_patches.bbclass
  - Original Tarball & Patches & Other files
  - do_patch[postfunc] = <archive ${S}/patches with Original tarball>
  - Included temp dir if enabled via SOURCE_ARCHIVE_LOG

configured_source.bbclass
  - Get the ${S} dir (or $S & any build dir) after do_configure
  - do_configure[postfunc] = <archive configured source>
  - Include temp dir if enabled via SOURCE_ARCHIVE_LOG


> and then realize do_stage[prefunc/postfunc] in this class.
> stage include unpack, patch, configure, build package and so on.
> for example. we want only to get original tarball, then we can define a
> bbclass named "source.bbclass"
> source.bbclass:
> inherit archiver.bbclass
> do_unpack[postfunc] = " do_get_source"
>
> if we want to get both original and patches tarball, then we can define
> bbclass names "source_patches.bbclass"
> source_patches.bbclass
> inherit archiver.bbclass
> do_unpack[postfunc] = " do_get_source_patches"
>
>> SOURCE_ARCHIVE_PACKAGE_TYPE = {tar, srpm}
>> SOURCE_ARCHIVE_LOG = {True, False}
>>
>> Please let us know if you have any questions.
>>
>> Sau!
>>
> If my understanding don't meet your ideas, Please correct me.
>
I should be available via IRC in your afternoon, I will be away during 
your morning.

Sau!

> Thanks for your help very much.
>
> Thanks
> Yan
>> On 02/15/2012 05:19 PM, Xiaofeng Yan wrote:
>>
>




More information about the Openembedded-core mailing list