[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: gcc sig 11

John Goerzen wrote:
> Hi,
> I have been attempting to compile perl 5.004 for Sparc just to try it out.
> However, gcc is sig 11'ing on me, at the exact same spot each time:
> cc -c  -Dbool=char -DHAS_BOOL -D_REENTRANT -I/usr/local/include
> -DSTRUCT_TM_HASZONE -O2    -DVERSION=\"1.02\" -DXS_VERSION=\"1.02\" -fpic
> -I../..  POSIX.c
> cc: Internal compiler error: program cc1 got fatal signal 11
> make[2]: *** [POSIX.o] Error 1
> make[2]: Leaving directory
> `/home/jgoerzen/packages/perl/perl-5.004.04/ext/POSIX'
> make[1]: *** [lib/auto/POSIX/POSIX.so] Error 2
> make[1]: Leaving directory `/home/jgoerzen/packages/perl/perl-5.004.04'
> make: *** [build] Error 2
> On an i386, I might think it was hardware.  However, this machine has now
> done this three times in a row at the exact same spot, so I am not leaning
> in that direction anymore at this point.  And ideas would be welcome.
> John

Yes, I've got the same error.  It's due to macro expansions in bits/string2.h.
The POSIX.c defines a very very huge expression (in the "constant" function)
based on a switch statement with lot of strcmp inside each case : each strcmp
is expanded in a rather huge expression which resolves to only one function
call at compile time (the best implementation depending on string types) and
gcc seems to merge all tests in one *big* expression.  It then has some
troubles to optimize it ;-(

BTW, to solve your problem, you have to disable string optimizations by adding


on top of the POSIX.xs file (not POSIX.c since the latter is built
automatically from the former).

Note: you cannot put this #define just before #include <string.h> in POSIX.xs
because it seems it is not the first access to this file.

Hope this help.

 Eric Delaunay                 | "La guerre justifie l'existence des militaires.
 delaunay@lix.polytechnique.fr | En les supprimant." Henri Jeanson (1900-1970)

To UNSUBSCRIBE, email to debian-sparc-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Reply to: