Re: Bugs in g++ on Sarge?
On Fri, 2004-10-22 at 18:33, James Wiggs wrote:
> On Fri, 2004-10-22 at 11:22, Nikita V. Youshchenko wrote:
> > > On Fri, 2004-10-22 at 08:59, Nikita V. Youshchenko wrote:
> > > > > The bugs are *not* repeatable. I can type make, have the make
> > > > > choke on something like the above, type make again, and it will get
> > > > > through the compile.
> > > >
> > > > Are you sure you don't have hardware problems?
> > > > Most likely such gcc failures are caused by bad ram and/or processor
> > > > overheat.
> > >
> > > I'd actually be relieved if it was a hardware problem, then I could
> > > just by some replacement memory or whatever and be done with it. Any
> > > suggestions as to tests I could run to track that down? I can run a
> > > copy of Memtest86 on the box tonight. I just don't see any other signs
> > > of instability on the box, and it gets quite heavily used, compiling a
> > > lot of code every day as a development server. Is it just that g++ is
> > > working the memory a lot harder than gcc/php/mysql/xfree86/everything
> > > else?
> >
> > I'm not absolutely sure that it's hearware-related.
> > However, non-reproducable ICE on normal hardware is something really
> > strange. G++ itself is completely deterministic.
> > G++ on large files with hard usage of templates is something that exercises
> > CPU and memory seriously. So it *may* show problems that other load
> > doesn't show. On the other case, if the box is in heavy daily use, and no
> > hardware failures happen, maybe hardware is OK.
> >
> > Strange. Very strange.
> >
> > Can't you test compilation of library in question on another box?
>
> Well, I just successfully compiled the code on a box I have
> access to which is installed with RH 8.0. It's a P4 1.5 GHz with
> 512 MB RAM. The version of g++ on it is 3.2-7 (RedHat 8.0). OK,
> so this begins to look like it *could* be the hardware, but I hate
> comparing apples to oranges on something like this. I really wish
> I could do the test on a machine that's running Sarge. Has anyone
> else compiled this software (QuickFIX 1.9.2) successfully on a
> Sarge box?
>
> Thanks for the feedback on this. I will post more as I go.
>
> regards,
> Jim
Well, there have been multiple follow-ups at this point from
people who have successfully compiled this code on their Sarge
boxes. It's really looking like there's a problem with my box.
I booted the thing off a Memtest86v3.1a boot floppy and ran it
for 6 hours solid, going through two full passes with 0 errors
of any kind. I don't think there's anything wrong with the RAM
in this machine. Can anyone suggest what else might be wrong
on the box that could lead to problems like this? Could a CPU
be flaky in some way to cause this? Hard drives? I'm really
stumped, here...
thanks,
Jim
Reply to: