[OE-core] [PATCH 3/3] uclibc.inc: avoid immediate expansion for UCLIBC_EXTRA_CFLAGS

[Richard Purdie]

On Thu, 2015-08-06 at 18:14 -0700, Andre McCurdy wrote:
> On Thu, Aug 6, 2015 at 4:02 PM, Richard Purdie
> <richard.purdie at linuxfoundation.org> wrote:
> > On Thu, 2015-08-06 at 12:41 -0700, Andre McCurdy wrote:
> >> If immediate variable expansion is used, then UCLIBC_EXTRA_CFLAGS can
> >> sometimes be derived from a partially expanded value of CFLAGS.
> >>
> >> One specific problem case occurs when security_flags.inc is used: CFLAGS
> >> is evaluated before the SECURITY_CFLAGS_pn-uclibc over-ride has been
> >> applied and therefore UCLIBC_EXTRA_CFLAGS ends up being based on the
> >> default SECURITY_CFLAGS instead of the uclibc specific value.
> >>
> >> Signed-off-by: Andre McCurdy <armccurdy at gmail.com>
> >> ---
> >>  meta/recipes-core/uclibc/uclibc.inc | 4 ++--
> >>  1 file changed, 2 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/meta/recipes-core/uclibc/uclibc.inc b/meta/recipes-core/uclibc/uclibc.inc
> >> index c483228..3ed8b60 100644
> >> --- a/meta/recipes-core/uclibc/uclibc.inc
> >> +++ b/meta/recipes-core/uclibc/uclibc.inc
> >> @@ -57,8 +57,8 @@ export V="2"
> >>  # -O<n> -fno-omit-frame-pointer ends up with GCC ICE on thumb as reported
> >>  # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44860
> >>  #
> >> -CFLAGS_arm := "${@oe_filter_out('-fno-omit-frame-pointer', '${CFLAGS}', d)}"
> >> -UCLIBC_EXTRA_CFLAGS  := "${@oe_filter_out('(-I\S+|-i\S+)', '${CFLAGS}', d)}"
> >> +CFLAGS_arm = "${@oe_filter_out('-fno-omit-frame-pointer', '${CFLAGS}', d)}"
> >
> > I'm rather surprised we don't get circular references without the :=
> > operator with this. I suspect we can see build failures if you change
> > the code like this. That said, I've been meaning to go through and
> > replace all the uses of oe_filter_out with the remove operator. Can we
> > do that here please instead?
> >
> > E.g:
> > CFLAGS_remove_arm = "-fno-omit-frame-pointer"
> 
> OK, new patch on the way.
> 
> >> +UCLIBC_EXTRA_CFLAGS = "${@oe_filter_out('(-I\S+|-i\S+)', '${CFLAGS}', d)}"
> 
> Is it OK to keep using oe_filter_out() here to support the regex?

Yes, the regex is obviously trickier with _remove so I think that one is
a fair use of oe_filter_out(). Thanks for the updated patches, just need
to answer Khem's question, it sounds like LDFLAGS may be used somehow
but I've not looked at the code as yet.

Cheers,

Richard