Re: -std=gnu99

[David Holland <dholland-tech%netbsd.org@localhost>]

On Fri, May 23, 2014 at 10:14:41AM +0100, Iain Hibbert wrote:
 > > Well, there's pcc, but I suppose if it didn't understand this option
 > > ragge would have said so upthread.
 > 
 > Well, I specifically mentioned PCC in my last message as supporting this
 > option..

Yes, after the part I was replying to. :-p

 > > However, what I meant was: any others we might want to try or tinker
 > > with.
 > 
 > ..so my question remains, which other compilers?

Any others we might want to try or tinker with. Are you really going
to argue that there are none and none will ever be written?

 > > All you're doing is increasing the amount of effort it takes to
 > > try a different compiler.
 > 
 > Not really. I have been trying a different compiler (PCC) and the *lack*
 > of -std=gnu99 has made it more difficult.  In fact, I would probably have
 > picked up on it earlier, but it turned out that currently unless you
 > specifically disable all other compiler support you get -std=gnu99 added,
 > which is the bug I am wanting to fix.

Yes, please fix that.

 > > By all means fix the conditional to test the right thing, and add pcc to
 > > it if appropriate. But there's no benefit to *removing* the conditional.
 > 
 > I maintain that there is, for the moment. The *source code* is C99 with
 > some GNU dialect features. This means that this is not conditional, as no
 > compiler will be able to build it correctly without GCC compat active.

...and what makes you think that the frobozz compiler's option for
selecting this is "-std=gnu99" as opposed to, say,
"--language-version=99 --enable-language-extensions=gnu"?

I mean, it's likely that any new compiler will accept a lot of gcc's
arguments, for the same reason clang does; but there are still a bunch
of old compilers floating around out there and one of them might come
back from the dead at any time. After all, pcc did.

 > >  > and you just can't build
 > >  > NetBSD without GCC compat active.
 > >
 > > That's a bug. Don't add to it / make it worse.
 > 
 > and as I said, that is not a bug I am prepared to tackle at this time.

You don't have to tackle it at this time. If you want to help, file
PRs; otherwise, ignore it. Just don't make it worse. Ok?

Removing conditionals that prevent passing compiler X's flags to
compiler Y is not "cleanup" or "progress", it's vandalism.

 > ** 
 > http://nxr.netbsd.org/source/search?q=UNSUPPORTED_COMPILER&project=src&defs=&refs=&path=&hist=

Given that most of that is gcc or gcc-related sources, it's thoroughly
unconvincing.

-- 
David A. Holland
dholland%netbsd.org@localhost