[OE-core] [oe] gcc-cross-initial patching fails

[Andreas Müller]

On Thu, Feb 23, 2012 at 8:18 PM, Koen Kooi <koen at dominion.thruhere.net> wrote:
>
> Op 22 feb. 2012, om 22:39 heeft Andreas Müller het volgende geschreven:
>
>> On Wed, Feb 22, 2012 at 10:18 PM, Khem Raj <raj.khem at gmail.com> wrote:
>>> On Wed, Feb 22, 2012 at 12:15 PM, McClintock Matthew-B29882
>>> <B29882 at freescale.com> wrote:
>>>>> Applying patch gcc45-no-add-needed.patch
>>>>> patching file gcc/config/alpha/elf.h
>>>>> patching file gcc/config/ia64/linux.h
>>>>> patching file gcc/config/linux.h
>>>>> Hunk #1 succeeded at 89 (offset -12 lines).
>>>>> patching file gcc/config/rs6000/sysv4.h
>>>>> Hunk #1 FAILED at 908.
>>>>> 1 out of 1 hunk FAILED -- rejects in file gcc/config/rs6000/sysv4.h
>>>>> Patch gcc45-no-add-needed.patch does not apply (enforce with -f)
>>>>> ERROR: Function failed: patch_do_patch
>>>>> ERROR: Logfile of failure stored in:
>>>>> /home/Superandy/tmp/oe-core-eglibc/work-shared/gcc-4.5-r45+svnr181733/temp/log.do_patch.5367
>>>>> NOTE: package gcc-cross-initial-4.5-r45+svnr181733: task do_patch: Failed
>>>>> ERROR: Task 2829
>>>>> (/home/Superandy/data/oe-core/sources/meta-openembedded/meta-oe/recipes-devtools/gcc/gcc-cross-initial_4.5.bb,
>>>>> do_patch) failed with exit code '1'
>>>>>
>>>>> git log does not help me much -  what goes wrong?
>>>>
>>>> Which actual patch is it failing on? Is there a conflict between
>>>> oe-core and meta-oe patch set against gcc?
>>>
>>> gcc45-no-add-needed.patch is the patch in question. I dont see
>>> anything wrong with it.
>>> Can you do bitbake -ccleanall gcc-cross-initial and redo bitbake
>>> gcc-cross-initial -cpatch ?
>>>
>> FYI: I removed the line which came in by the commits I mentioned above
>>
>> should_apply[vardepsexclude] += "PN"
>>
>> in openembedded-core/meta/recipes-devtools/gcc/gcc-common.inc
>>
>> and build from scratch works without issues.
>
> I received reports of the same problem and I handed your fix to them, I'm waiting to hear back.
>
Having some experience to this problem now, I have the feeling that
there is some race condition on patching. I think gcc-cross-initial /
gcc-cross-intermediate / gcc-runtime share same source directory -
correct me if I am wrong. In case all start patching about the same
time, the patches don't match for those being late. This theory is
encouraged by the fact that the following sequence helps when building
from scratch:

1. bitbake gcc-cross-initial
2. bitbake gcc-cross-intermediate
3. bitbake gcc-runtime
4. bitbake <my-image>

Cheers

Andreas