Thomas Fletcher
|
Re: Interpreting PPC backtrace
|
Thomas Fletcher
08/11/2008 9:51 PM
post11662
|
Re: Interpreting PPC backtrace
Ryan Mansfield wrote:
> Thomas Fletcher wrote:
>
>> I'm running into an issue with a largeish shared library on PPC where I'm getting a SIGILL when the library tries to
load.
>>
>> I've made a sample program that simply dlopen's the file and the problem occurs during the library initialization it
seems from the backtrace:
>>
>> The backtrace is:
>>
>> Backtrace from load:
>> #0 0xff213e04 in _SDA_BASE_ ()
>> from /home/thomas/mylib.so
>> #1 0xfe8df944 in frame_dummy ()
>> from /home/thomas/mylib.so
>> #2 0xff06de28 in _init ()
>> from /home/thomas/mylib.so
>> #3 0xfe33c9d4 in init () from /usr/qnx632/target/qnx6/ppcbe/lib/libc.so.2
>> #4 0xfe33e39c in dlopen () from /usr/qnx632/target/qnx6/ppcbe/lib/libc.so.2
>> #5 0x480405ac in main (argc=2, argv=0x4803fc24) at load.c:13
>>
>> This is all running on 6.3.2 and I was wondering if this triggered anyone's mind about a libc or complilation problem
I should be looking for since those init routines come from the OS side.
>>
>
> Would you be compiling the the shared oject with the 3.3.5 gcc driver?
> And is this C++ ? With or without exceptions?
>
Right on the money. This is C++, and I'm building with qcc (qcc
-V3.3.5,ntoppcbe_cpp) but looking at the
compile line (this is a port that likes GCC argument conventions) I'm
passing:
qcc -V3.3.5,ntoppcbe_cpp ... -f no-exceptions ... -f PIC
So, how bad is this?
It is going to be a bear to change things to -shared (to replace -f PIC)
and to remove -f no-exceptions
(I can switch to ntoppcbe_cpp-ne fairly easily though if the options
won't collide inside qcc). I'll run
a build to check this out further.
Thanks,
Thomas
|
|
|