On Thu, Feb 24, 2005 at 01:57:28PM -0700, Liberty Young wrote: > On Thu, 2005-02-24 at 11:20, Ray Lehtiniemi wrote: > > On Thu, Feb 24, 2005 at 10:50:36AM -0700, Liberty Young wrote: > > > > > ftp://embeddedARM.com/downloads/bare_bones-debian.tar.gz > > > > thanks, i'll grab that and give it a try too > > > > > > > Another interesting thing to note is that booting into a busybox OS > > > doesn't exhibit these problems. > > > > hmmm... busybox compiled with uclibc instead of glibc, perhaps? i've > > been using the buildroot compiler to date, which targets uclibc, and > > i've not had problems... i've just built up a crosstool compiler > > targetting glibc. i wonder if i'll start seeing problems too? there > > are certainly lots of glibc-related user-space faulting problems in > > the mailing list archives... > > Nope. busybox compiled with glibc (kegel's crosstool, > 0.28rc-something-or-rather) well, i guess that rules glibc out... i tried both the debian root fs and the denx 3.1 rootfs using the latest bk tree on my edb9302 eval board. i didn't really observe any segfault problems, although both had their share of issues... the debian system seemed to have nothing really configured in the inittab and /etc/rc1.d directories... i had to create a bunch of symlinks by hand, but once i did that, it seemed stable enough. the denx system has some weird userid permissions issue... it seems that everything is running as root during bootup, but early in the bootup process, mounting /proc fails because "only root can do that". go figure... i was able to login as root without seeing the "malloc botched assertion" messages that llandre reported earlier, although root was still not allowed to mount filesystems :-/ so, no luck, i can't seem to reproduce either problem here at my end. all i can suggest at this point is to pull the latest bk tree and cross your fingers... -- ---------------------------------------------------------------------- Ray L <rayl@xxxxxxxx>