Date: Thu, 8 Apr 2004 15:53:49 -0700 (PDT) From: Doug White <dwhite@gumbysoft.com> To: Ruslan Ermilov <ru@freebsd.org> Cc: current@freebsd.org Subject: Re: can't build world from 5.2.1-REL Message-ID: <20040408145857.F13329@carver.gumbysoft.com> In-Reply-To: <20040408194746.GD1919@ip.net.ua> References: <20040408122547.E13329@carver.gumbysoft.com> <20040408194746.GD1919@ip.net.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 8 Apr 2004, Ruslan Ermilov wrote: > On Thu, Apr 08, 2004 at 12:32:19PM -0700, Doug White wrote: > > Hey folks, > > > > I've got a dual Xeon machine here that can't seem to build -current from > > 5.2.1-RELEASE. It appears something is badly broken in the gcc build or > > yacc or both. The victim machine was just freshly installed with > > 5.2.1-REL, and I've only built the kernel to enable SMP and remove the > > drivers I don't use. > > > > The first time I tried to buildworld, it croaked on trying to link > > gengtype complaining that a bunch of yy* symbols weren't defined. I > > toasted /usr/obj and built again and got "attempt to use poisoned > > "strdup"" messages during a make depend step in cc_tools. > > > > The last time the "poisioned" messages appeared, which was last year, it > > was due to broken sed. Looking at the diffs between the foo.c and > > foo.c%DIKED files, the correct changes are being made. > > > > I'm working on generating buildworld typescripts once I get a fresh src > > tree downloaded to the machine again. Since my other -current builders, > > which update near daily, are building OK, I'm guessing there is a > > non-backwards-compatible change in there somewhere. > > > > I'll post the URLs once I get the logs uploaded. > > > This looks very familiar, though I don't remember the details, > but you may try searching the mailing list archives. Or for a > brute force attack of the problem, please try upgrading to the > -current version of yacc(1) before doing buildworld. You're thinking of the sed breakage from last June. It croaks with the same errors, but the suggested test of diffing the %DIKED files comes back with the expected changes (malloc -> xmalloc, etc.). freebsd-native.h matches the one on my other -current builders. I even tried reinstalling sed to be safe -- no joy. If the comment in freebsd-native.h is representative, its some screwup in libiberty. But those files haven't changed in eons. -- Doug White | FreeBSD: The Power to Serve dwhite@gumbysoft.com | www.FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040408145857.F13329>