Re: [Buildroot] [PATCH 1/4] arch/Config.in.xtensa: adjust endianness logic to avoid bogus configurations

[Thomas Petazzoni <thomas.petazzoni@bootlin.com>]

Hello,

On Sun, 25 Sep 2022 09:25:02 +0200
"Yann E. MORIN" <yann.morin.1998@free.fr> wrote:

> >  (1) Showing the overlay file option only when building an internal
> >      toolchain. For external toolchain configurations, it does not
> >      make sense to show it as it will have no effect.  
> 
> This should be a separate patch.

ACK.

> >  (2) Showing the endianness choice either for internal toolchain
> >      configurations with non-empty BR2_XTENSA_OVERLAY_FILE, or for
> >      external toolchain configurations.  
> 
> As an empty BR2_XTENSA_OVERLAY_FILE is not accepted anymore (see below),

See my patch 2/4. Not accepting an empty BR2_XTENSA_OVERLAY_FILE is
totally wrong, as it breaks the build with external toolchains.
External toolchains can target a custom Xtensa core, but specifying a
overlay file with an external toolchain does not make any sense.

My patch 2/4 therefore reverts 4cbf7336914f, which is wrong.

> this can be simplified to "sharing the endianness chice for internal and
> external toolchains".
> 
> >  (2) Defaulting to BE when BR2_XTENSA_OVERLAY_FILE is empty.  
> 
> An empty BR2_XTENSA_OVERLAY_FILE is now no longer accepted, after commit
> 4cbf7336914f (arch/xtensa: custom configuration requires an overlay).

Right, but this commit is incorrect, and needs to be reverted.

> > Fixes:
> >   http://autobuild.buildroot.net/results/702e2886156f291466375dfcf412c20f1aa5857d/  
> 
> This commit if from August the 15th; I could not find any such failure
> since 4cbf7336914f was applied; the last occured 2022-09-08, on commit
> d2141f65e449, which is the parent of 4cbf7336914f (there are build
> failures on older branches, of course, but not on master).

See above: 4cbf7336914f is incorrect. It breaks for example the test
case for the Bootlin Xtensa LX60 external toolchain. An empty overlay
file is perfectly valid for a custom Xtensa core, when using an
external toolchain.

And I am precisely doing the revert *after* this patch 1/4 so that the
series is bisectable. I.e introduce the correct fix, and then remove
the incorrect fix.

Best regards,

Thomas
-- 
Thomas Petazzoni, co-owner and CEO, Bootlin
Embedded Linux and Kernel engineering and training
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot