This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Disabling fixproto: OS-less embedded targets?


On Tue, 2003-09-30 at 11:51, Nathanael Nerode wrote:
> >> or are using a C89-compliant library,
> >Hmm, this sounds overly simplistic to me. You should be aware that
> >embedded GCCs in most cases will be cross-toolchains:
> >
> >* People often regard "gcc" as secondary toolchain to vendor toolchains
> >and will be trying to use the vendor toolchain's libc.
> >
> >* Another kind of situation when people start using gcc for embedded
> >systems is a vendor not providing toolchains for a user's host
> >environment (Many vendors don't provide cross-toolchains for non-M$
> >systems).
> >
> >As vendor cross-toolchain target libc are host independent, people will
> >be trying to use these vendor libcs with gcc. 
> >
> >As gcc has no control about these vendor libcs, any assumption on C89
> >compliance would be a "wild guess".
> 
> It's been 14 *years* since ANSI C came out. 

Yes, but how long did it take to Sun to reach a point that Solaris
satisfies gcc-ANSI-fication expectation?

>  Fixproto is designed to
> convert K&R-style (prototype-free) headers to C89 (prototype-present)
> headers.

Hmm, I must be missing something. I thought fixproto does more than
KnR->ANSI prototype conversion? If adding ANSI-prototypes is all we are
discussing about, then I hereby withdraw my comments.

>   If vendors are still supplying headers without prototypes,
> nobody should be having anything to do with them....
Agreed, but ... the embedded scene is a bit different than the "big OS
support" scene, longer release cycles, small user base, people sticking
with older toolchains due to ridiculous pricing policies, lots of
hobbyists trying to resurrect ancient hardware, vendors selling "ancient
CPUs" targeting embedded applications.

All in all this results into the embedded scene to be "slow" and
"embarressingly conservative" wrt. standardization progress.

> >>  or can fend for themselves.
> >Are you trying to say that you are wanting to give up the amount
> >generality fixproto has tried to provide?
> Um, yes?  :-)
> 
> >> Is this reasonable?  (These are all -elf, -coff, or -aout targets.)
> >I don't think so.
> Well, I won't do it this release cycle, then.  :-)
Cf. above. If ANSI-prototypes is all this is about, ...

Ralf



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]