On 08/15/2011 11:49 PM, David Kuehling wrote:
Hmm, this crashes so early on, I'd say it crashes*within* the dynamic linker (or did you just truncate the stace log so much?). It could also be one of the famous unaligned memory accesses, that use to work on all modern CPUs, except MIPS. Looking at the assembler instruction at the crash location + register dump that would be easy to diagnose (gdserver anyone? 'info regs', 'disas $pc-8,$pc+8').
Hi David thanks for the info.
In kernel 2.6.39 (probably earlier?) Linux now has an unaligned memory load/store emulation, that works similar to FPU emulation. And it is slow as hell. But at least it fixes these problems with misbehaving software. Maybe we should backport that to our kernel?
here is the plan: 1. I will release this one '08142011' as the next release since we have new gcc and kernel, all apps included in that image works fine. http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08142011-1719/ 2. after release I will rebase openwrt-xburst.git on top of last openwrt upstream 3. direct update our kernel to linux 3.0 or 3.1 maybe, no backport. :) 4. then find out what app not working. or all python stuff work again with new kernel 5. release one linux3.0/1 openwrt image as fast as possible. (have to include atBen stuff in this release) I already lose upstream and late on release openwrt images. so I want speed up a little what do you think David. any advice will be great. xiangfu _______________________________________________ Qi Hardware Discussion List Mail to list (members only): [email protected] Subscribe or Unsubscribe: http://lists.en.qi-hardware.com/mailman/listinfo/discussion
