On Tue, Jan 22, 2008 at 07:41:04PM +0100, Wouter Verhelst wrote: > On Thu, Jan 17, 2008 at 11:15:49AM -0600, Stephen R Marenka wrote: > > In current sid executing 'find /usr/bin -type d -o -ls' results in a > > segfault. This works fine in etch-m68k. > > > > Anyone care to followup? > > Having a look at it now.
Looks like some fine register or stack corruption somewhere. ------------ Core was generated by `find /usr/bin/ -type d -o -ls'. Program terminated with signal 11, Segmentation fault. #0 0x80007cba in ?? () (gdb) x/i $pc 0x80007cba <[EMAIL PROTECTED]>: unlk %fp (gdb) p $fp $1 = (void *) 0x82f888 (gdb) p $sp $2 = (void *) 0xef82f870 (gdb) ------------ Unh. For some reaason, the leading byte is stripped off the frame pointer. This is reproducable; in concurrent runs, every time the frame pointer loses the 0xef in the beginning. Of course dropping a stack frame is very problematic if your frame pointer is broken. The entire subroutine looks like this: ------------ (gdb) disass 0x80007c8e 0x80007cbd Dump of assembler code from 0x80007c8e to 0x80007cbd: 0x80007c8e <[EMAIL PROTECTED]>: linkw %fp,#0 0x80007c92 <[EMAIL PROTECTED]>: movel 0x8001e58c <stdout>,[EMAIL PROTECTED] 0x80007c98 <[EMAIL PROTECTED]>: movel 0x8001e9bc,[EMAIL PROTECTED] 0x80007c9e <[EMAIL PROTECTED]>: movel 0x8001e9b2,[EMAIL PROTECTED] 0x80007ca4 <[EMAIL PROTECTED]>: movel %fp@(12),[EMAIL PROTECTED] 0x80007ca8 <[EMAIL PROTECTED]>: movel 0x8001e9d8,[EMAIL PROTECTED] 0x80007cae <[EMAIL PROTECTED]>: movel %fp@(8),[EMAIL PROTECTED] 0x80007cb2 <[EMAIL PROTECTED]>: bsrl 0x800093ce <[EMAIL PROTECTED]> 0x80007cb8 <[EMAIL PROTECTED]>: moveq #1,%d0 0x80007cba <[EMAIL PROTECTED]>: unlk %fp 0x80007cbc <[EMAIL PROTECTED]>: rts ------------ If I set a breakpoint somewhere before that bsrl and inspect the registers, everything is fine. At 800093ce, a somewhat longish subroutine starts, with the rts at 800097ea, and many jumps in between; and somewhere in there, that byte is lost. Debugging this is going to be fun. -- <Lo-lan-do> Home is where you have to wash the dishes. -- #debian-devel, Freenode, 2004-09-22 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
