[Fwd: Re: regression in char_transpose_1.f90]

[Jerry DeLisle <jvdelisle at verizon dot net>]

Steve Kargl wrote:
> On Mon, Jan 30, 2006 at 04:45:24PM -0800, Steve Kargl wrote:
>
> > Is anyone else seeing a regression in char_transpose_1.f90?
> >
> > troutmask:sgk[203] gfc4x -o z char_transpose_1.f90
> > char_transpose_1.f90: In function 'MAIN__':
> > char_transpose_1.f90:9: internal compiler error: in gfc_conv_array_transpose, at fortran/trans-array.c:717
> > Please submit a full bug report,
> > with preprocessed source if appropriate.
> >
> > I have actually traced this all the way back to the original commit.
> > troutmask:sgk[204] gfc4x -o z char_transpose_1.f90 --version
> > GNU Fortran 95 (GCC) 4.2.0 20051213 (experimental)
>
>
> Well, I'm just completely baffled.  The first 5 executable lines
> of gfc_conv_array_transpose are 
>
>   loop = se->loop;
>
>   src_ss = gfc_walk_expr (expr);
>   dest_ss = se->ss;
>
>   src_info = &src_ss->data.info;
>   dest_info = &dest_ss->data.info;
>
> In a gdb session, it appears as if "src_info = &src_ss->data.info;"
> is never executed!(?)  So, when src_info is referenced in a gcc_assert
> some 10 lines down, I get an ICE.
You might be seeing a gdb anomaly where it looks like its not being executed
because the code has been gyrated or otherwise optimized.

BTW I am not getting this regression on i686.

Jerry