Re: help with reporting bug compiling proprietary code

[Mike Stump <mrs at apple dot com>]

On Tuesday, September 24, 2002, at 06:20 AM, Rob Taylor wrote:
> > I'm hitting a pretty serious optimisation bug in some proprietary 
> > code, and I
> > don't seem to be able to provide a simple testcase (if I remove a
> > single line of
> > code, it goes away!). I can't post all the prepossessed source up on 
> > gnats for
> > all the world to see, but I could wrangle allowing certain gcc 
> > maintainers
> > access - what's the procedure in these cases?
> I should point out that I'm not asking for free support here and can 
> work around
> the bug, but I would rather the bug was fixed in gcc, as I'm sure 
> would all.
> Which is the reason for my query.
Three thoughts come to mind.

1  You can reduce the testcase and sanitize all interesting aspects out 
of your code.
2  You can binary search for when the bug went in (or came out)
3  You can debug the problem yourself, and see if you can't figure out 
enough to be
   able to describe what is wrong in the logic of the compiler.
4  Commercial support.

4 may be the easiest for you, after that, 2 may be the next easiest, 
though, you'll have to slog through unrelated things.  Also, you need 
to be dead sure that it is a bug and not just a misunderstanding of the 
rules on your part.  For example, if you discover that the bug went in, 
when alias analysis was turned on by default, we'll all probably just 
laugh, and say try -fno-strict-aliasing.

For 3, you can start with -da, and see where the compiler first goes 
wrong.  You can then run the compiler under gdb and investigate the 
where and why.

Good luck.