Re: This is not a proposal that can be implemented in make
"Roberto C. Sanchez" <email@example.com> writes:
> On Sat, Jan 06, 2007 at 08:59:47PM +0100, Josselin Mouette wrote:
>> Le samedi 06 janvier 2007 à 13:32 -0600, Manoj Srivastava a écrit :
>> > Something like this needs to be designed from a system
>> > perspective, and needs, probably, to cooperation the package build
>> > system.
>> > Since this is something that cant be done by a single pacage,
>> > I am reassigning this bug. (For example, my employer would be annoyed
>> > if all out product builds fail since make suddenly injects random
>> > flags for which our software is unprepared for).
>> Isn't that what pentium-builder does?
> If you are talking about the injection of random flags, then does it
> matter? Using pentium-builder is optional for building software. Using
> make generally is not. A misbehaved wrapper is easy to work around
> (don't use it). A misbehaved hard dependency is a bit tougher.
And why should make do that?
Gcc already provides this feature. You have to setup a specs file and
include the options you want there. This is also so flexible that you
can omit options if conflicting options are given already. E.g. add
-m64 only if no -m32 is given.