This is the mail archive of the
mailing list for the GCC project.
Re: cpplib: what is a system header? [HEAD]
On Sat, Feb 17, 2001 at 08:16:43PM +0000, Neil Booth wrote:
> Russ Allbery wrote:-
> > Hm. This is going to cause a change in behavior of make depend in at
> > least some packages (INN being the example near and dear to my heart).
> > We'd end up listing Perl header files and the like in the dependencies in
> > the Makefile after this change without further changing our scripts, since
> > those directories are added to the include path with -I (we support
> > compilers other than gcc). Or am I missing something?
> No, I think this would cause a change of behaviour for you.
> Zack, what do you think we should do here? Should we insist people use
> -isystem when doing a "make depend", or abandon this idea?
It will cause a change of behavior, but I can't tell if it would be
wrong or not.
In 3.0, it's much too late to be changing semantics except to fix
bugs. I think the lack of transitive warning suppression counts as a
bug; we should change that. But people do depend on the existing -MM,
so let's not change that at all. Let's do the transitivity bit for
everything except dependencies, leave dependencies alone, and rename
system_include_depth to angle_bracket_depth or something to eliminate
the existing confusion.
On the mainline, we can experiment with making -MM use the same
definition of system headers as everything else. Then Russ and other
people can try it out and report if it really does break stuff. It
might be wisest to wait until after 3.0 is _released_.