Re: [PATCH v4 1/4] image.bbclass: fix dependency calculation when using conversion chaining

[Ed Bartosh <ed.bartosh@linux.intel.com>]

On Thu, Jul 28, 2016 at 09:42:44PM +0100, Richard Purdie wrote:
> On Wed, 2016-07-27 at 15:51 +0300, Ed Bartosh wrote:
> > From: Patrick Ohly <patrick.ohly@intel.com>
> > 
> > When using conversion chaining (for example example: .dsk.vdi.xz),
> > the imagetypes_getdepends() did not properly detect all compression
> > commands (thus skipping the corresponding COMPRESS_DEPENDS) and
> > the base type, leading to missing dependencies of the image's
> > do_rootfs
> > task.
> > 
> > This is not a big problem in practice because in those cases where
> > conversion chaining is useful (as in the example above), the missing
> > dependency is qemu-native, which typically gets built anyway.
> > 
> > imagetypes_getdepends() had hard-coded special treatment for certain
> > image types. This gets replaced with setting the normal IMAGE_DEPENDS
> > variables for these types.
> > 
> > [YOCTO #9076]
> > 
> > Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
> > Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com>
> 
> This patch has come in multiple times and each time I look at it I ask
> that it be checked against changes that happened in parallel since I
> believe there is something "not right" in the merge. Its not an easy
> one to review as it makes several different kinds of changes.
> 
> I just spent a while comparing the changes and it removes the reference
> to IMAGE_FSTYPES_DEBUGFS so I do believe this patch *does* cause at
> least one regression. Whether there are other issues its really hard to
> say.
> 
> Can someone try and perhaps make the diff more obvious and split it up?
> 
Thank you for the review.

I'll try to split this patch later. For now I'm going to remove it from
the patchset.

--
Regards,
Ed