Re: changing the default compilers for autoconf

[Thomas Dickey]

On Mon, 17 Sep 2007, Eric Blake wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> According to Thomas Dickey on 9/17/2007 6:23 AM:
> > It's more than that.  Setting the flags to -g is a policy decision made
> > in the mid-90's based on the claim that setting the optimization flag
> > in any compiler other than gcc would expose bugs in that compiler.
> >
> > (the claim of course is incorrect, but there's no point in discussing it
> > on this mailing list)
>
> I agree that the claim is probably not true, especially since compiler
> technology has improved in the meantime.  However, this IS the appropriate
> list for discussing the issue.  Please, feel free to post examples and/or
> patches of what you think can be improved, rather than assuming,
> pessimistically, that we don't care to hear what you have to say.
> Furthermore, I don't see this claim written in the manual, which means

It was what David MacKenzie told me (and even that long ago, my experience 
- comparing with vendor's compilers to Solaris, SunOS, AIX, HPUX - the 
most likely compiler to be a candidate for suppressing optimization was 
gcc(*)).

He produced no examples to support his statement at that time, nor have I 
seen any reliable reports on this mailing list.

(unsurprisingly enough, the same assertion has been repeatedly made on 
this mailing list - perhaps someone could find a few of those ;-)

(*) watching bug reports in FreeBSD regarding gcc 4.1, it appears to be 
still the case (ymmv)

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net