On Sat, Dec 04, 2004 at 12:31:49PM +0100, Thomas Klausner wrote:
> On Sat, Dec 04, 2004 at 10:47:31AM +0000, Alistair Crooks wrote:
> > pkgsrc bulk build results
> > NetBSD 2.99.10/i386
> > 
> > Summary:
> > 
> >         Build started:                  Sun Nov 28 11:21:35 2004 UTC
> >         Build ended:                    Fri Dec  3 16:53:54 2004 UTC
> 
> Much breakage in this one due to the header file change in
> current, where the isascii and tolower families of functions
> lost a cast. Two ways to fix: Either make the arguments
> unsigned chars (preferred), or remove -Werror from the compiler flags,
> e.g. with:
> BUILDLINK_TRANSFORM+=  rm:-Werror
> in the package Makefile.

Yes, -current's move to the more correct ctype macros seems to have
had a more far-reaching effect than the statvfs changes.  I agree that
the best way is to cast the arguments where necessary to unsigned
char.
 
> Also, the bulk build machine did seem to have had some network problems,
> and three internal compiler errors that were not there before (hardware
> problems?).

The comms with the big bad world went down for a while, and that accounted
for some of the fallout here. I am running another build now, taken from
sources earlier on today. That run is now at:

	2004/12/04 20:51:14  2698/5200=51.9% mail/php-imap @ i386> 

so I'd expect to have some results soon after the freeze starts.

I hope there aren't any hardware problems - the disks themselves are in
a RAID 1, and the memory has been stable since I fixed the duff DIMM
earlier on this year (and thanks to memtest for finding that).

Thanks once again for your analyses.

Regards,
Alistair