Hi Waldemar,
2016-02-09 21:46 GMT+02:00 Waldemar Brodkorb wbx@uclibc-ng.org:
I rather would like to avoid supporting such very old kernels. The oldest supported kernel is 2.6.32.70. Could you try using this for the toolchain? What runtime problems occur, when you use 2.6.32.70 for toolchain and then run the code on 2.6.22.19?
I verified that 2.6.32.70 works with uClibc-ng for mipsel toolchain.
I haven't tried 2.6.32.70 -- the issues I had happened with 2.6.36.4 toolchain on 2.6.22.19 system. One issue I'm absolutely sure is due to older kernel is this one: https://sourceware.org/ml/gdb-prs/2015-q4/msg00337.html I later found out that ARMv5 gdb has the same issue if run on 2.6.12, but works OK on my test 2.6.36.4 machine (2.6.36.4 toolchain as well). And there're others that are very hard to debug, e.g., I found out that php crashes when configuring owncloud8 if pdo_mysql module is loaded (this doesn't happen on other test targets where systems run kernels not older than toolchains').
I also see that Entware-ng project is using 2.6.22.19 kernel with uClibc-ng 1.0.9. They're also using quite a lot of uClibc-ng patches: https://github.com/Entware-ng/Entware-ng/tree/master/toolchain/uClibc/patche... The first two are safe to ignore (they're just hardcoding '/opt' into prefix definitions, I don't need that -- I configure UCLIBC_RUNTIME_PREFIX as '/opt' instead; well, I'd need some parts of the patches to get working utils on the target system, but I don't package them as of now), but I wonder if some of the others would make uClibc-ng compile with 2.6.22.19 headers? It'd be good if you could take a look.
Thanks, Alex
P.S. I'll try 2.6.32.70 headers as soon as I can