Re: Fortran candidate patches for gcc-2.95.3.

[Bernd Schmidt <bernds at redhat dot com>]

On Sun, 10 Dec 2000, Toon Moene wrote:

> Bernd Schmidt wrote:
> > To be honest, I don't know yet.  What kind of failures are these?  Do they
> > (or are they likely to) occur in real-life code?
> 
> Well, surely they do - we wouldn't know of them otherwise :-)

Well, it does happen that someone spots a bug in the source and constructs a
matching testcase ;)
Do you know which programs are affected?

> The FFE patch enables code to use the FPUTC intrinsic (in libU77); this
> is the least important of the two, because the bug doesn't lead to
> miscompiled, but running code.

So this is an ICE on valid code?

> The libf2c patch fixes reading of some NAMELISTs, which is of the form
> "runs, but produces wrong results".

Thanks; but it's still not quite a detailed enough description.  What are
NAMELISTs (sorry, I don't know Fortran), what exactly are we doing
incorrectly, and how does your patch fix the problem?

Who is the Fortran maintainer these days?  I don't fell I know this code well
enough to judge these patches properly.  I'm especially a bit nervous about
the second patch - it doesn't seem to be installed in current CVS, so it
probably didn't get sufficient testing yet.


Bernd