Re: [PATCH] depcomp and gcc3 still not safe enough

[Tom Tromey]

>>>>> "Hari" == Raja R Harinath <address@hidden> writes:

>> Dependency files are different from .o files though.  It would make
>> sense to write a temporary dependency file and then `mv' it on
>> success (or `rm' on failure).

Hari> But, does 'gcc' have to do it?  There are valid usage modes (use
Hari> sinclude/-include) which don't require the "save to temp and
Hari> move" method.

My goal for automake is to eventually have configure detect some
really winning version of gcc and then avoid invoking depcomp at all
in that case.  That way gcc users will be rewarded by faster compiles
(leaving aside the issue of gcc performance, of course).  We can't do
this if gcc doesn't cooperate.  Of course my point of view is biased.

What we have now is still better than the other compilers, so I'm not
going to push too hard on this issue for gcc.  There are plenty of
more important things.

Tom