Re: [MiNT] gcc4 vs gcc2 crash "test case"

[Vincent Rivière <vincent.riviere@freesbee.fr>]

Frank Naumann wrote:

> Is the PC inside the quake address room (e.g. what's the basepage address
> if the crash happen).

Unfortunately no, it is completely bogus. I had to put a lot of traces 
in the code in order to find where the crash came from. There is a Trace 
History function in ARAnyM, I don't know it it could be usable to find 
that kind of bug, in order to know where we come from.

> If yes there are good chances to use the gdb to
> lookup at which source code line the crash happen.

I spent a lot of time in gdb these days (I never used it before). And 
with optimization, and without the frame pointer, I had a lot of 
trouble. I didn't manage to step on one line of disassembly (as opposed 
to one source line)... I searched a bit on the web, and I didn't find 
how to do. However, gdb is a very powerful tool, it would be very useful 
to learn all its functionalities.

> Beside from that, that
> it work without optimization might be a advice for an optimization bug
> (maybe related to the mixed assembler <-> C source code).

It is usually the case... but not this time. The assembler sources of 
Quake seems very clean. This is an optimization bug... but in GCC this 
time :-(

I didn't try X-Window on for MiNT on ARAnyM yet.
Maybe it would be possible to compile DDD for MiNT in order to enable 
visual debugging ? A cross debugger would be fine, too.

--
Vincent Rivière