Warnings from autotools_copy_aclocals()

Warnings from autotools_copy_aclocals()

[Andre McCurdy]

Does anyone else see these type of warnings from
autotools_copy_aclocals()? I don't remember exactly when they started
(a few months ago at least, maybe more).

  WARNING: gmp-4.2.1-r0 do_configure:
/.../build-dcx960-debug/tmp/sstate-control/manifest-dcx960-debug-glibc-initial.populate_sysroot
not found
  WARNING: gmp-4.2.1-r0 do_configure:
/.../build-dcx960-debug/tmp/sstate-control/manifest-dcx960-debug-libgcc-initial.populate_sysroot
not found

Re: Warnings from autotools_copy_aclocals()

[Richard Purdie]

On Wed, 2016-04-13 at 16:58 -0700, Andre McCurdy wrote:
> Does anyone else see these type of warnings from
> autotools_copy_aclocals()? I don't remember exactly when they started
> (a few months ago at least, maybe more).
> 
>   WARNING: gmp-4.2.1-r0 do_configure:
> /.../build-dcx960-debug/tmp/sstate-control/manifest-dcx960-debug
> -glibc-initial.populate_sysroot
> not found
>   WARNING: gmp-4.2.1-r0 do_configure:
> /.../build-dcx960-debug/tmp/sstate-control/manifest-dcx960-debug
> -libgcc-initial.populate_sysroot
> not found

Yes, I also do have a rough idea of the cause.

They happen in builds from sstate where the system has decided that for
whatever reason, a given dependency isn't needed. The aclocal
dependency traversal code looks down the full dependency list and warns
when these things aren't present.

Examples of sstate's decisions would be skipping libgcc-initial when
libgcc is available, same for glibc-initial and glibc.

Somehow we need to reconcile the two views. sstate is clearly correct
or there would be failures. The question is how to make the aclocal
code have the same understanding (which will then limit the number of
aclocal macros copied in).

Cheers,

Richard