Re: luajit-2.0.0-beta6 crash

[Henk Boom <henk@...>]

Thanks for the quick reply Mike!

On 2 March 2011 14:05, Mike Pall <mikelu-1103@mike.de> wrote:
> This is the exact same configuration as my developer box. I would
> be _very_ surprised if it crashes on something basic as Fibonacci.

Me too, I made this simple example because I was having the same
problem running any non-trivial lua code I had lying around.

> I just tried your variant of the code with both beta6 and git HEAD.
> But I cannot reproduce a crash, no matter how hard I try.
>
>> (gdb) bt
>> #0  0x00000000394cffed in ?? ()
>> #1  0x0000000040000f18 in ?? ()
>> #2  0x000000000041f9e0 in lj_vm_hotcall ()
>> #3  0x000000000040c2aa in lua_pcall (L=0x40000378,
>
> Something is wrong with the debug info or the call frames. This is
> not a plausible stack trace (for various reasons). You could try
> to enable -DLUAJIT_USE_GDBJIT, to see if the unknown frames are
> part of the generated code (but then it can't originate from
> lj_vm_hotcall ...).
>
> Some random idea: maybe you forgot to run 'make clean' after
> changing the flags in the Makefile?

I just did a `make clean` before adding the option you mentioned. Same
results, with a hopefully more useful backtrace:


(gdb) bt
#0  0x00000000394dffed in TRACE_1 () at badfib.lua:1
#1  0x000000000040c32a in lua_pcall (L=0x40000378,
    nargs=<value optimized out>, nresults=-1, errfunc=<value optimized out>)
    at lj_api.c:1027
#2  0x0000000000403f46 in docall (L=0x40000378, narg=1, clear=0)
    at luajit.c:120
#3  0x0000000000404f34 in handle_script (L=0x40000378) at luajit.c:284
#4  pmain (L=0x40000378) at luajit.c:488
#5  0x000000000041ec89 in lj_BC_FUNCC ()
#6  0x000000000040b752 in lua_cpcall (L=0xfffffffb40007c40, func=0x400082e8,
    ud=0x400082c8) at lj_api.c:1049
#7  0x0000000000403e7f in main (argc=3, argv=0x7fffffffe2c8) at luajit.c:516

I also dumped a core in case you find it useful, you can get it here:

http://dl.dropbox.com/u/6410967/luajit-core

    henk