After battling with this error for several hours, I was able to resolve this
error simply by putting my nop instruction pattern as the first pattern.
Initially, I had movqi as my first instruction pattern. Is this some
restriction or a bug in gcc?
While debugging the source I found that the assertion "gcc_assert
(cached_reg_save_code[reg][mode]);" gets failed in reg_save_code(), which
implies that value returned by "cached_reg_save_code[reg][mode]" has to be
non-zero.
Now, in reg_save_code(), cached_reg_save_code[reg][mode] gets its value from
"recog_memoized (saveinsn)" as:
---snip---
cached_reg_save_code[reg][mode] = recog_memoized (saveinsn);
---snip---
Now, in file insn-recog.c generated during the build, I see that movqi has
got a number 0 returned by recog function:
---snip---
L2: ATTRIBUTE_UNUSED_LABEL
if (general_operand (x1, QImode))
{
operands[1] = x1;
return 0; /* movqi */
}
x1 = XEXP (x0, 0);
goto L4;
---snip---
If I place any other pattern before movqi, so that movqi gets a non-zero
value, I do not get the internal compiler error.
Could you please elaborate on this behaviour? I could not find any such
restriction in the gcc documentation.
Thanks,
Deepti
> -----Original Message-----
> From: gcc-help-owner@xxxxxxxxxxx [mailto:gcc-help-owner@xxxxxxxxxxx] On
> Behalf Of Ian Lance Taylor
> Sent: 13 June 2012 23:01
> To: damodar.sonone
> Cc: gcc-help@xxxxxxxxxxx
> Subject: Re: upgrading gcc-4.4.0 to gcc-4.7.0:configure: error: cannot
> compute suffix of object files: cannot compile
>
> "damodar.sonone" <dsonone@xxxxxxxxx> writes:
>
> > conftest.c: In function 'main':
> > conftest.c:16:1: internal compiler error: in reg_save_code, at
> > caller-save.c:158
> > Please submit a full bug report,
> > with preprocessed source if appropriate.
> > See <http://gcc.gnu.org/bugs.html> for instructions.
>
> This one is the problem.
>
> It means that your newly built compiler has a bug--it gets an internal
> compiler error. Unfortunately I don't have a simple solution here. I
> don't know why such an error would occur for you, when it clearly
> doesn't happen for other people.
>
> Ian
