[LLVMdev] Wrong Dwarf info with -O2? [PowerPC]

[Eric Christopher]

On May 10, 2012, at 7:27 AM, Joerg Sonnenberger <joerg at britannica.bec.de> wrote:

> On Thu, May 10, 2012 at 03:24:01PM +0200, Gabor Greif wrote:
>> Hi all,
>> 
>> today I was debugging some code (compiled by a 2 weeks old clang) on a
>> PowerPC embedded system using the newest gdb (v7.4).
>> 
>> I got a backtrace like this:
>> 
>> (gdb) bt
>> #0  0x0fe96314 in poll () from /lib/tls/libc.so.6
>> #1  0x1002c8ac in mli_poll (
>>    fds=<error reading variable: Unhandled dwarf expression opcode 0x1>,
>>    nfds=<optimized out>, timeout=<optimized out>) at msglib.c:7560
>> #2  0x10027684 in mli_do_work (peer_index=<optimized out>,
>>    creply=<error reading variable: Unhandled dwarf expression opcode 0xff>,
>>    flags=<optimized out>) at msglib.c:6323
>> #3  0x1002cce8 in ml_wait_for_next_event () at msglib.c:6821
>> #4  0x1000fdd0 in sys_mainloop () at config.c:1221
>> #5  0x1000bda8 in main (argc=<optimized out>, argv=<optimized out>)
>>    at sysmon.c:437
>> 
>> 
>> Please note the "Unhandled dwarf expression"  errors. Is this a known
>> problem? How can I help debugging this? Btw., this is dead-simple C
>> code.
> 
> Please add what you have to:
> 
> http://llvm.org/bugs/show_bug.cgi?id=12733

Yes please. I could use a testcase that's a little easier to look at than an optimized clang :)

(Not that Joerg's repro instructions aren't nice mind you...)

-eric