Fix c++/9393

Geoff Keating geoffk@geoffk.org
Sat Apr 12 02:33:00 GMT 2003


Zack Weinberg <zack@codesourcery.com> writes:

> Geoffrey Keating <gkeating@apple.com> writes:
> 
> > This patch lets you have reproducibility if you want it, determinism
> > if you want it, and correct behaviour all the time.
> 
> Suggest that with -v, you print out a suitable -frandom-seed= switch
> to reproduce the behavior, like we do for the GGC memory-size
> heuristics? 

I don't see an need for this.  In all the time that GCC's had this
behaviour, I've never heard a request to be able to control this value
for debugging GCC; it's mostly so that people can build the same
sources and get the same .o files.

I don't want to needlessly expand the -v output.  It's already pretty big.

You *do* get the value when you use -dA, in the output .s file.  I'm
not sure about the benefit of that; it might just be needlessly
preventing people from comparing .s files.

> And maybe this should be a --param instead of an -f switch?

It doesn't seem like a parameter.  It doesn't actually change GCC's
behaviour, other than a few 6-character strings in the output.

-- 
- Geoffrey Keating <geoffk@geoffk.org>



More information about the Gcc-patches mailing list