This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Fatal error buildling libobjc/linking.m
- To: law at cygnus dot com
- Subject: Re: Fatal error buildling libobjc/linking.m
- From: Marc Espie <espie at quatramaran dot ens dot fr>
- Date: Mon, 31 May 1999 03:31:01 +0200
- Cc: egcs at egcs dot cygnus dot com
- Organization: Ecole Normale Superieure de Paris
In article <11713.928106581@upchuck.cygnus.com> you write:
>OK. I've figured out what is causing the segfault when building
>libobjc/linking.m on some systems.
>It occurs when one has c-parse.{c,h} generated with bison-1.25 and
>objc-parse.c generated with bison-1.27 or vice-versa.
>bison-1.25 and bison-1.27 start numbering tokens at different values, thus
>(for example) IDENTIFIER may have the value 258 for one version of bison and
>257 for a different version of bison.
Ouch, I had forgotten about this one...
I did reach the same conclusions independently some weeks ago, and forgot
to tell about it. I'm very sorry about that.
Another nicety of the current state of affairs:
bison is not strictly required for building those parsers.
OpenBSD Berkeley Yacc does not have any trouble with them at all.
Test-results are perfectly identical with the bison version, and there
does not seem to be any problem with them so far... after building roughly
300 MB of source code.