Thread overview
[bug(s)] More ICEs
Dec 29, 2005
Matthew
Dec 29, 2005
Walter Bright
Dec 29, 2005
Matthew
December 29, 2005
I'm getting lots more ICEs with DMC++ in a number of test programs for XSTL, most of which give no information, beyond the "helpful" Windows process crash dialog (which helpfully holds up the builds until someone clicks the button). It seems that DMC++ can handle most of the techniques that the other "modern" compilers can also handle when such techniques are presented individually, but dies obtusely when presented with them in concert.

Is there no way to have DMC++ provide some kind of useful information when it crashes, so as to assist us to assist you in fixing things? Otherwise, all one is left with is the sense that DMC++ simply can't handle complex techniques in combination, which is a bit of a shame.

Cheers

The Whinging Harbinger. ;-)


December 29, 2005
"Matthew" <matthew@hat.stlsoft.dot.org> wrote in message news:dov94o$1cjm$1@digitaldaemon.com...
> I'm getting lots more ICEs with DMC++ in a number of test programs for
> XSTL,
> most of which give no information, beyond the "helpful" Windows process
> crash dialog (which helpfully holds up the builds until someone clicks the
> button). It seems that DMC++ can handle most of the techniques that the
> other "modern" compilers can also handle when such techniques are
> presented
> individually, but dies obtusely when presented with them in concert.
>
> Is there no way to have DMC++ provide some kind of useful information when it crashes, so as to assist us to assist you in fixing things? Otherwise, all one is left with is the sense that DMC++ simply can't handle complex techniques in combination, which is a bit of a shame.

What I do with an ICE is build the compiler with debug on, then run it under the debugger and see where it is dying, and proceed from there. This is why I need a reproducible example.


December 29, 2005
> > I'm getting lots more ICEs with DMC++ in a number of test programs for
> > XSTL,
> > most of which give no information, beyond the "helpful" Windows process
> > crash dialog (which helpfully holds up the builds until someone clicks
the
> > button). It seems that DMC++ can handle most of the techniques that the
> > other "modern" compilers can also handle when such techniques are
> > presented
> > individually, but dies obtusely when presented with them in concert.
> >
> > Is there no way to have DMC++ provide some kind of useful information
when
> > it crashes, so as to assist us to assist you in fixing things?
Otherwise,
> > all one is left with is the sense that DMC++ simply can't handle complex techniques in combination, which is a bit of a shame.
>
> What I do with an ICE is build the compiler with debug on, then run it
under
> the debugger and see where it is dying, and proceed from there. This is
why
> I need a reproducible example.

Understood. I'll try and note which things cause it, and then when I've got this beastie complete I'll try and get you a set of examples where it happens.

Cheers

Matthew