Bug#190818: [m68k] Another Internal compiler error in extract_insn, at recog.c:2148

[Matthias Klose <doko@cs.tu-berlin.de>]

J.H.M. Dassen (Ray) writes:
> Package: gcc-3.2
> Version: unclear, perhaps 1:3.2.3-0pre9 ?
> Severity: important
> Tags: upstream
> 
> Severity motivation: ice-on-legal-code when building a package that's not
> supposed to percolate to testing (#180052).
> 
> Buildd builds of gnumeric 1.1.x on m68k have been failing with an ICE for
> some time now; see
> 	http://buildd.debian.org/build.php?&pkg=gnumeric&arch=m68k
> 
> This is from http://buildd.debian.org/fetch.php?&pkg=gnumeric&ver=1.1.16-6&arch=m68k&stamp=1049328812&file=log&as=raw
[...]
> ../../../plugins/fn-financial/functions.c: In function `gnumeric_db':
> ../../../plugins/fn-financial/functions.c:950: unrecognizable insn:
> (insn 266 84 85 (set (reg:DF 17 %fp1)
>         (const_double:DF 0 [0x0] 0 [0x0] -2147467265 [0x80003fff])) -1 (nil)
>     (nil))
> ../../../plugins/fn-financial/functions.c:950: Internal compiler error in extract_insn, at recog.c:2148
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
> make[4]: *** [functions.lo] Error 1

we do have a patch for this in the debian-gcc CVS (m68k-const,
submitted by Roman).  Unfortunately with this patch applied, the g77
testsuite freezes at least two m68k machines, that's the reason it is
not applied.

Maybe the m68k build crew could rebuild the final gcc-3.2.3 upload
with this patch enabled (but not upload) and upload the packages that
don't build otherwise?

	Matthias