[oe] [RFC] Badness with -ibad, a replacement for zecke-no-host-includes.patch.

[Graham Gower]

On 9 July 2010 20:14, Richard Purdie <rpurdie at rpsys.net> wrote:
> On Thu, 2010-07-08 at 23:36 +0800, Holger Freyther wrote:
>> On 07/08/2010 01:39 PM, Graham Gower wrote:
>> > This patch introduces a new command line parameter to gcc (at
>> > Zecke's suggestion), -ibad, which may be used to specify
>> > blacklisted include prefixes. E.g.
>> >
>> > grg at eye7:/tmp$ ~/oe2/tmp/cross/mipsel/bin/mipsel-angstrom-linux-gcc -c test.c -ibad /usr/include -I /usr/include
>> > CROSS COMPILE Badness: /usr/include in INCLUDEPATH: /usr/include
>>
>> nice! The only thing that is worrying is the question how we make sure
>> these bad path's are enforced.
>>
>> E.g. a recipe not honoring our CPPFLAGS might silently pass this check
>> (while before it was hardcoded into the compiler). My proposal would be
>> to extend the sanity check and grep the compile log for gcc and -ibad
>> and fail if gcc exists but -ibad does not occur?

I haven't tried this, but I suspect it wouldn't be too difficult to
put them into gcc's default spec file.
That would fix the problem of breaking things for an external toolchain too.

>
> How about making this option an environmental variable rather than a
> flag? I've seen that technique used to great effect for this exact
> reason before...

That would work too.

-Graham