[OE-core] Different behavior of same IMAGE_FSTYPES definition in edison branch and denzil branch

[McClintock Matthew-B29882]

On Wed, Jul 18, 2012 at 8:05 AM, Luo Zhenhua-B19537
<B19537 at freescale.com> wrote:
> I meet a different behavior of IMAGE_FSTYPES in edison branch and denzil
> branch.
>
>
>
> In edison branch:
>
> IMAGE_FSTYPES is defined as following in poky/meta/conf/bitbake.conf.
>
> IMAGE_FSTYPES ?= "tar.gz"
>
> I defined IMAGE_FSTYPES as following in my customized image bb file(in my
> layer)
>
> IMAGE_FSTYPES ?= "tar.gz ext2.gz.u-boot jffs2"
>
> The actual value of IMAGE_FSTYPES is "tar.gz ext2.gz.u-boot jffs2" which is
> the expectation.
>
>
>
> In denzil branch:
>
> IMAGE_FSTYPES is defined as following in poky/meta/conf/bitbake.conf.
>
> IMAGE_FSTYPES ?= "tar.gz"
>
> I defined IMAGE_FSTYPES as following in my customized image bb file(in my
> layer)
>
> IMAGE_FSTYPES ?= "tar.gz ext2.gz.u-boot jffs2"
>
> The actual value of IMAGE_FSTYPES is "tar.gz"
>
>
>
> Is this the expected behavior in denzil branch? Should I use ‘IMAGE_FSTYPES
> = "tar.gz ext2.gz.u-boot jffs2"’ to override previous definition?

This has to do with the order the files are processed I believe. Which
ever ?= processed first takes precendence. I for one think we should
probably make everything in oe-core be ??= and layers can defined ?=
and local.conf/users can use =.

Thoughts?

-M