[bitbake-devel] [PATCH v2] bb/fetch2: fixes copying of file://dir; subdir=foo, bug 6128 and bug 6129

[Sasha Shashkevich]

>> Changes were tested on 1.8 and 2.0 branches for core-image-sato and
>> meta-sdk images.
>> 
>> Additional changes since last sent patch:
>> 
>> - New test cases were added to cover absolute paths in SRC_URI for
>> local files.
>> - Fixed recipe nativesdk-qemu-helper_1.0.bb: it was based on
>> incorrect copying behavior which had been changed by this patch.
>> 
>> Please review my changes.
> 
> Could you have a look at master please? I already took one of your
> earlier patches which I followed with a changes to correctly handle
> absolute paths:
> 
> http://git.yoctoproject.org/cgit.cgi/poky/commit/bitbake/lib?id=2a73181
> 33bbd64f5dcb2ddf400229d4627925580
> 
> I wasn't convinced the behaviour in the qemu-helper recipe was
> incorrect. If you still believe it is, I'm happy to consider that.
> 
> I'm also very interested in getting some updated test cases in.
> 

Documentation have no mentions how bitbake handles absolute paths, so I propose to discuss it here. I see it (and have implemented) as this: only last file/dir from whole path should be taken:

	file:///some/abs/path/to/file → ${WORKDIR}/file

Behavior prior to this patch and how it works in current master:

	file:///some/abs/path/to/file → ${WORKDIR}/some/abs/path/to/file

qemu_helper recipe relies as well on last one:

	file://${COREBASE}/scripts/runqemu → ${WORKDIR}/home/user/projects/yocto/scripts/runqemu

But I believe that recipe authors wouldn't have 'home/user/projects/yocto/scripts' dir in ${WORKDIR} as it completely host specific and useless. For such reason I have implemented behavior that copies only last file/dir entity from absolute path to ${WORKDIR} and changed qemu_helper recipe accordingly. New test cases are covering this behavior too.

If you agree with such behavior, then you can integrate patch v2 and remove your fix for qemu_helper in master since it conflicts with my changes.

BTW, I want to add one more test case in upcoming patch: absolute paths with wildcards. This should be covered also.

Regards,
Alexander
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/bitbake-devel/attachments/20160310/4f584943/attachment-0002.html>