[PATCH] bitbake.conf: Remove replace call from MACHINE_ARCH

[PATCH] bitbake.conf: Remove replace call from MACHINE_ARCH

[Ioan-Adrian Ratiu]

This replace() hack was added to fix Yocto bug #946 [1] with commit id
69b3a11. It is time to fix this properly. It should not be an issue
anymore with libzypp after 4.5 years, but if the initial bug still
exists, then a more intelligent solution should be used to fix it.

If a package build fails because it expects _ instead of - inside
MACHINE_ARCH, then the problem is with the package build, not with the
global variable used by all machines.

*IF* the offending package build cannot be easily fixed, then there
are better workarounds than globally replacing. For example overwriting
MACHINE_ARCH in the <machine-name>.conf which is causing the offending
package to fail (thus not screwing all), or using conditional overrides
MACHINE_ARCH_qemux86-64 = "qemux86_64"

Mangling the variable globally for all machines causes other components
that expect the sane MACHINE_ARCH value to fail.

[1] https://bugzilla.yoctoproject.org/show_bug.cgi?id=946

Signed-off-by: Ioan-Adrian Ratiu <adrian.ratiu@ni.com>
---
 meta/conf/bitbake.conf | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
index 730ab80..19e7ddf 100644
--- a/meta/conf/bitbake.conf
+++ b/meta/conf/bitbake.conf
@@ -136,7 +136,7 @@ SDK_AS_ARCH = "${BUILD_AS_ARCH}"
 
 TUNE_PKGARCH ??= ""
 PACKAGE_ARCH ??= "${TUNE_PKGARCH}"
-MACHINE_ARCH = "${@[d.getVar('TUNE_PKGARCH', True), d.getVar('MACHINE', True)][bool(d.getVar('MACHINE', True))].replace('-', '_')}"
+MACHINE_ARCH = "${@[d.getVar('TUNE_PKGARCH', True), d.getVar('MACHINE', True)][bool(d.getVar('MACHINE', True))]}"
 PACKAGE_EXTRA_ARCHS ??= "${PACKAGE_EXTRA_ARCHS_tune-${DEFAULTTUNE}}"
 PACKAGE_ARCHS = "all any noarch ${PACKAGE_EXTRA_ARCHS} ${MACHINE_ARCH}"
 # MACHINE_ARCH shouldn't be included here as a variable dependency
-- 
2.1.4

Re: [PATCH] bitbake.conf: Remove replace call from MACHINE_ARCH

[Richard Purdie]

On Wed, 2015-07-29 at 13:49 +0300, Ioan-Adrian Ratiu wrote:
> This replace() hack was added to fix Yocto bug #946 [1] with commit id
> 69b3a11. It is time to fix this properly.

I'd love to see this fixed properly. I assume you're volunteering to do
this?

>  It should not be an issue
> anymore with libzypp after 4.5 years, but if the initial bug still
> exists, then a more intelligent solution should be used to fix it.

We no longer use libzypp. I believe however that this is a more deeply
rooted rpm issue.

So, what testing did you do of the rpm backend after this change?

Note that we also have to do mappings of TARGET_ARCH ->
TRANSLATED_TARGET_ARCH which is the same issue as this one.

Cheers,

Richard