Re: Bug in GCC 2.6.2

[Robert.Wilhelm@Physik.TU-Muenchen.DE]

> 
> Hi all!
> 
> I have found a real bug in gcc 2.6.2 that may or may not be the cause
> for the problems with MiNT 1.12 and MultiTOS.  Here is a test case:
> [...]

This may not be the cause.
Gcc 2.5.8 compiles this right and we have got Multitos problems with
it, too.


> P.S.: addq #4,sp definitely does NOT change any condition codes.
> 

The problem might be some uninitialised variables somewhere.
Consider following buggy example:

honi:~/misc$ cat t.c
void b(int dummy)
{
  int m;
  printf("%d\n",m);
  m=1;
}

main()
{
  b(0);
  b(0);
}

honi:~/misc$ gcc t.c
honi:~/misc$ a.out
-1073742700
1
honi:~/misc$ gcc -fdefer-pop t.c
honi:~/misc$ a.out
-1073742700
-1073742700


Did anybody yet go through proc.c compiling only one function per try 
with -fno-defer-pop ? This might help us to narrow the problem.

ciao,
 Robert