Re: amd64(gcc4): aptitude dumps core
On Thu, Jan 13, 2005 at 01:46:41PM -0600, Pete Harlan wrote:
> On Thu, Jan 13, 2005 at 08:09:00PM +0100, Kurt Roeckx wrote:
> >
> > I'm assuming you're talking about mysql?
>
> Yes.
>
> > I've just build mysql-dfsg 4.0.23-1, ran make test, and got:
> > All 211 tests were successful.
>
> We got that too, until we ran "make test" in a loop. Every twenty
> times or so through the loop (running the 211 test each time) a test
> would fail, usually one of the tests having to do with replication,
> but which one varied from run to run. i386 was fine, gcc-3.4 was
> fine, pure64 wasn't. If you can't reproduce this, then perhaps the
> relevant thing in pure64 got fixed, which would be great.
I've let Frederik Schueler run run that test on an dual CPU box
in a loop. He says that all 120 runs of it were succesful. I
also tried it many times myself and they all were succesful too.
If there was a problem, it might have been fixed, or it might not
have been. If a problem only shows once in a while, it's ussualy
a timing issue. Building with a different compiler might
generate different results because of this. I've seen it too
many times that when rebuild without -O or with gdb attached the
problems seem to go away. Compilation with a different compiler
can have the same effect. That does not mean the problem isn't
there anymore, it just means it's now showing up in the tests you
do. I doubt in this case gcc-3.4 solved anything.
Kurt
Reply to: