Re: [Gm2] gm2-20040213 internal error

[Gaius Mulley]

Paul Whittington <address@hidden> writes:

> Hello all,
> 
> I've been lurking for quite a long while now, and I've finally had
> some time to play with the compiler.
> 
> During my testing, which involved the compilation of several DEF
> files, I've encountered the following internal error.
> 
> What's the best way to proceed?
> 
> Are there some docs on how to go about running gm2 in gdb and finding
> out what's going on?
> 

Hi Paul,

no docs for this, but I'll add a section to gm2.texi as this is likely
to be useful. I run gm2 under gdb using the following method.

Firstly I add the following lines to the file
gcc-3.3.2/gcc/gdbinit.in

---- cut here---- cut here---- cut here---- cut here---- cut here
cd ../../gcc-3.3.2/gcc
break exit
run -Wq -Igm2/gm2-libs:gm2/gm2-compiler -g -gstabs 
testsuite/gm2/extensions/pass/optparam.mod -o -
---- cut here---- cut here---- cut here---- cut here---- cut here

now 
cd <whateverbuilddir>/gcc
emacs .gdbinit
M-x gdb
[gdb prompts for an executable, tell it to use cc1gm2]

and debug using emacs/gdb.

I also find that running cc1gm2 under valgrind to be tremendously
useful. I think I've seen that gcc can tell its subcomponents to run
inside valgrind.. but I've not yet found out how to do this.  If
anyone else knows the rhunes I'd love to know.  A number of times I've
run cc1gm2 under valgrind, and many times thought valgrind has given
me incorrect information - always valgrind was right and it found some
really useful memory bugs.. a great tool

Gaius