Re: 21.5 build time segmentation fault
Aidan Kehoe writes:
No, nothing convincingly like this. The log file doesn’t show the usual
XEmacs “Fatal error” message.
That’s a *very* new GCC and those are some *very* aggressive optimisation
flags. Were your own crashes with a similar level of optimisation?
Yes. It was built on the machine used with the xemacs build bot "amd64 gentoo"
using the gentoo package build. But unfortunately(!?) I have only been able to
reproduce the crash once so far.
Interestingly the "cfg2" configuration (optimized build) used by the xemacs
build bot does actually use -03 as it picks the value provided by configure in
the "checking for preferred optimization flags.." check. So that is aggressive
but it seems to work.
In the gentoo build case "only" -O2 is used since the value that configure
wants is overridden by my default setting -O2. That is the same setting that
is used by the reporter. So the crash is for a -O2 compiled binary (The log
might look differently at a first glance but according to the gcc manual the
last optimization setting on the command line is the one that will be in
effect. So further down the command line you will find -O2 present.)
Would you consider -O2 to be aggressive?
NEWGC currently does give segmentation violations on OS X, because
(as far
as I can see) the Mach exception handler code is broken, and some of my
recent changes prompted configure to treat OS X as non-POSIX again.
OK. That would not be or solve the problem for the Gentoo build then.
Yours
--
%% Mats