Re: Buildbot failure, x86_gentoo_cfg1
Ar an dara lá déag de mí na Samhain, scríobh Mats Lidell:
> Aidan Kehoe writes:
> Is there any option for WONTFIX in the buildbot?
You mean like not flagging a build as failed if it fails due to an error
that will not be fixed or how should the WONTFIX option work?
Something like that. I suppose on thinking about it further it’s not very
practical.
> If not, could we try changing the buildbot to be
--without-optimizaiton?
The builds are configurable. Feel free to suggest what builds you find
most valuable and what they should contain. Both number of configurations
and what each configuration should contain can be changed.
We have now four different setups based on what once was use for
manually verifying builds. These are:
1. # Most useful debugging:
["--with-mule",
"--with-pdump",
"--with-bignum=gmp",
"--with-debug=yes",
"--with-error-checking=all",
"--with-kkcc=no",
"--with-newgc=no"]
2. # Maximum error catching:
["--with-xemacs-compiler=g++",
"--with-mule",
"--with-pdump",
"--with-error-checking=all",
"--with-debug=yes",
"--with-bignum=gmp",
"--with-xft=all",
"--with-kkcc",
"--with-newgc",
"--with-union-type",
"--with-optimization"]
OK, we should just remove --with-optimization from this. Full optimization
with a G++ build is pointless, it won’t give faster code than the equivalent
gcc for the end user, and will only flush out compiler bugs for us, it is
unlikely to pick up a bug with our code that one of the other builds doesn’t
provoke.
I’m happy enough with the others.
--
‘As I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after forty pints of stout’
(C. Moore)