yes, that works fine. I've tried the OE-uclibc rootfs with the ELDK kernel - that's how I got my first complete boot w/ uclibc, for instance (actually, when I do that I just enable USB-Storage in-kernel and boot the rootfs from a USB pen drive).
The OE toolchain that we use (for foonas-em) is a fork of OE-dev, to be honest, though_very_ close to OE-dev.
Two distinctions/details I should point out:
1. the build system of OE starts truly from scratch, pure source, even for the toolchain. There is no downloading of the toolchain, only downloading of the source packages that are used to build it. First, the native parts (for i686) are built, and then cross-tools, then the target packages are built. The build process/sequence goes : gcc-cross-initial -> uclibc-initial -> gcc-cross-intermediate -> uclibc -> gcc-cross So, if I understand this correctly, the gcc-cross that get built partly depends on uclibc.
SInce each "fresh" run of OE, is from the ground up, in effect I've tried rebuilding the entire toolchain perhaps a 15 or 20 times, and the problem is reproducible each time (if I use one of the reduced kernel configs. the hdhlad-config in /arch/mips/configs is, for example, "too small", as well. But since it works just fine in glibc, I am sure this is a toolchain+uclibc=mayhem issue, not a problem w/ your kernel or your defconfig for the board.
2. The way timtimred set the foonas OE toolchain
up for building foonas-em
requires two successive runs
A first build the foonas-em-image (the rootfs + a non-initramfs "bare" kernel)
B second rebuild the kernel and add in the cpio-ed rootfs to get the the initramfs image
When I test these kernels, I check both stages (A-the "bare" non-initramfs kernel, and B-the initramfs image) for bootability. I've also tried building a regular foonas (non-initramfs) kernel, and that results the same if I choose uclibc.
My thought is that I've found a difficult bug in the OE setup or maybe even in gcc 4.2.4. I'm just perplexed as where to start hunting. I did see this https://dev.openwrt.org/ticket/5351
, which make me wonder if I should try gcc 3.x.x, just for kicks.
Another thought is that this could be something particular to Ubuntu. Since the entire toolchain is built on our build boxes (and mine is running Jaunty Ubuntu), OE is sometimes sensitive to quirks/problems with what ever distro is installed.
EDIT: I'm trying w/ 4 3.4 now - maybe that will make some difference?
LS1 (foonas, nfs, Tranmission
print server, Firefly
for my Roku
Various LS-Pros v1,v2 (unbricked w/ serial & jtag)
KuroPro, LS2 & KuroHG (foonas)
Working on davysweather.dyndns.org
lately... => wooohooo!
Unknown command 'wooohooo!' - try 'help'