Re: [Gm2] Mac OS X 10.4/ppc Compiler Bug?

[Gaius Mulley]

Michael Lambert <address@hidden> writes:

> Hi Gaius,
> 
> Under Mac OS X 10.4 on a PowerPC (G4) I'm getting a few failures on
> check-gm2.  One set is from longtypes9.mod.  Checking it by hand:
> 
> -----
> [lassa:/tmp] lambert% gm2 -save-temps -c longtypes9.mod
> [lassa:/tmp] lambert% gm2 longtypes9.mod
> [lassa:/tmp] lambert% ./a.out
> [lassa:/tmp] lambert% echo $status
> 1
> -----
> 
> If I put in some print statements for debugging, the y+1+1 test works,
> but apparently 1+1+y equals 0 for y=5 (as does 1+y+1).
> 
> I'm sending the assembler output to Gaius separately.
> 
> Michael

Hi Michael,

thanks!  Yes this is odd, I'm in the middle of fixing the
REAL/LONGREAL bug reported a few days ago which exhibited very similar
characteristics - when the expression is being passed as an actual
parameter.  Looks like it needs converting inside CodeParam just like
the REAL/LONGREAL bug

regards,
Gaius