On Fri, Mar 25, 2005 at 03:45:58AM -0800, Steve Langasek wrote: > On Fri, Mar 25, 2005 at 01:08:30PM +0200, Lars Wirzenius wrote: > > Attached is a patch to fix most of the compilation problems for > > util-vserver. Steve Langasek tested this on sparc and it worked. Builds > > still fail on at least hppa, powerpc, and mipsel, since the dietlibc > > linking fails. See #300926 for some discussion about this. > > > I didn't try giving the -Os option to the diet command at the linking > > stage, since I didn't see an easy way to do it from util-vserver's > > debian/rules. > > > I have so far failed to debug the linking problem to any successful > > degree. Gerrit, perhaps you could give Ola a helping hand with the > > linking issue? > > <vorlon> liw: if you invoke it as "diet -v gcc -Wall -g [...]" instead of > "diet -v hppa-linux-gcc [...]", what happens? > <liw> vorlon, it fails by not recognizing some link file without libtool; if > I add the libtool stuff in front, then it does link > <vorlon> liw: basically, by calling ./configure --host=... --build=..., > util-vserver is incorrectly causing autoconf to think this is a > cross-compile invocation, and therefore calls hppa-linux-gcc > instead of just calling gcc; if you do in fact get different errors, > this may be the root of the problem > <vorlon> liw: in that case, switching util-vserver to use the confflags > sample stanza from /usr/share/doc/autotools-dev/README.Debian.gz > may be enough to cure the remaining ills
I prepared a new dietlibc upload to fix #300917 and some other things. I'll now also include the missing defines for O_NOFOLLOW and O_LARGEFILE on hppa, which is already fixed in upstream cvs. So these two workarounds can be omitted when build-depending on dietlibc-dev (>= 0.28-2). The workaround for sparc wouldn't be necessary if the diet wrapper is called with the -Os option (this adds -mcpu=supersparc), a good idea anyway for all architectures. With the additional change suggested by Steve, things should look much better then. Regards, Gerrit. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]