On Saturday 15 January 2005 17:17, Rob Landley wrote: > Running ptrace under uml found the problem with gcc: UML's default > environment includes "." at the end of $PATH, and if you have that gcc > doesn't search the path for ld. > > So I took "." out of the path, and it started working. Yay! (Cool stuff, > guys.) > > Along the way, I found some interesting quirks: > > 1) I mentioned the gratuitous use of perl earlier. This is only for > --show-config, which could A) be replaced by a sed script, I've just answered with a patch. > B) should > presumably be wrapped in CONFIG_IKCONFIG anyway. No, no more possible - I planned doing that but then the plain-text version went away... in the kernel normally only the gzip'ed version is saved. > I ripped it out of the makefile and it built fine without it. I'll see > about patching the makefile with the sed thing and/or the CONFIG option > this weekend, although I'm not good at makefiles...
> 2) Building under uclibc dies with a conflicting definition error, due to > arch/um/include/user.h containing "extern int strlcpy(char *, const char *, > int);", when uclibc's string.h contains: > extern size_t strlcpy(char *__restrict dst, const char *__restrict src, > size_t n) __THROW; > I yanked the strlcpy line from arch/um/include/user.h and everything built > fine without it (didn't even see any new warnings), both under both uclibc > (more or less 0.9.27, a cvs snapshot a couple weeks before release) and > glibc (knoppix 3.6, comes with glibc 2.3.2), possibly because it's still in > include/linux/string.h... More likely because it's not used in UML "userspace" code - it cannot include include/linux/string.h (it could include the host's version, but I don't think it will do this for <linux/string.h> - it's unlikely). > The resulting uclibc binaries seemed to work fine in both environments. I > could even run the uclibc one from within the uclibc chroot environment via > "./vmlinux rootfs=hostfs rootflags=/ rw init=/bin/sh" and it behaved pretty > well. (Overmount /proc and /dev/pts and life is good...) > 3) Running my big build script, it hangs between ./configure and make of > binutils. No idea why yet. Just sits there spinning, with vmlinux eating > 100% of the CPU, and if you leave it for half an hour it still won't > advance, and this is something that should only take a second or so... Well, try doing things at hand, and using sh -x (or the new bash-debugging of 3.0, I've not yet learned it but it could be useful maybe). > 4) I can get all sorts of things to segfault under uclibc with ptrace -f. You mean strace, right? > For example, just try ptrace -f on a shellscript and watch the segfaults > fly. Is it just me? No, it's me too, but I never got to understanding what's going on there. > #4 makes #3 a bit harder to debug. > > I'll try upgrading to -bs5 and see if that makes anything better. Well, I don't think -bs5 it's worth the effort - the problems fixed from -bb4 likely don't affect you in this case (they cause early breakage in the boot). > Query: is a 2.6.10-bb likely in future, or should I just try -bk or -mm or > some such? Watch the site, there will be upgrades on this... raw 2.6.10 isn't good for UML... however the first 2.6.10-bb probably won't be better / stabler than 2.6.9-bb, at least until the issue you are experiencing won't be debugged... -- Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 http://www.user-mode-linux.org/~blaisorblade ------------------------------------------------------- This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting Tool for open source databases. Create drag-&-drop reports. Save time by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc. Download a FREE copy at http://www.intelliview.com/go/osdn_nl _______________________________________________ User-mode-linux-devel mailing list User-mode-linux-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
