Обсуждение: Re: regression failure/UnixWare7.1.1/current sources

* Peter Eisentraut <peter_e@gmx.net> [001029 14:32]:
> Larry Rosenman writes:
> 
> > Would the timezone change last night be causing this? 
> 
> The "timestamp" failure, yes.  The "geometry", no.  Geometry simply needs
> a new expected file, but unfortunately they're not the same for "cc" and
> "gcc"...
Hmm. I wonder why cc and gcc are doing different math.  Wierd. 

I suspect it might have to do with what gcc was compiled on (7.0.x of
UW).   Can we make 2 expected files and have the map file figure it
out? 

As to the timestamp, can we make a note that on timechange sundays
don't run the regression test? :-) 

LER

> 
> -- 
> Peter Eisentraut      peter_e@gmx.net       http://yi.org/peter-e/

-- 
Larry Rosenman                      http://www.lerctr.org/~ler
Phone: +1 972-414-9812 (voice) Internet: ler@lerctr.org
US Mail: 1905 Steamboat Springs Drive, Garland, TX 75044-6749

Larry Rosenman writes:

> Hmm. I wonder why cc and gcc are doing different math.  Wierd. 

Not only that, but you get different results with the same compiler
depending on different optimization settings.  The joys of binary floating
point...

> I suspect it might have to do with what gcc was compiled on (7.0.x of
> UW).  Can we make 2 expected files and have the map file figure it
> out?

The resultmap mechanism isn't really prepared for this yet, but it is
doable I'd say.

> As to the timestamp, can we make a note that on timechange sundays
> don't run the regression test? :-) 

We've had a note in there since the last change (April was it?), which was
when I ran into this. :-)

-- 
Peter Eisentraut      peter_e@gmx.net       http://yi.org/peter-e/