Re: ASAN:DEADLYSIGNAL

[Ben Abbott]

> On Jun 20, 2016, at 8:04 PM, Ben Abbott <address@hidden> wrote:
> 
>> On Jun 20, 2016, at 4:42 PM, Dmitri A. Sergatskov <address@hidden> wrote:
>> 
>> On Mon, Jun 20, 2016 at 1:44 PM, Ben Abbott <address@hidden> wrote:
>> On Jun 20, 2016, at 09:43, Dmitri A. Sergatskov <address@hidden> wrote:
>> 
>>> On Mon, Jun 20, 2016 at 7:21 AM, Ben Abbott <address@hidden> wrote:
>>>> 
>>>> I’m using Apple’s clang and did hot set -fsanitize. I do not know if other 
>>>> libraries did or if Apple uses -fsanitize by default.
>>>> 
>>>> Ben
>>> 
>>> Dmitri,
>>> 
>>> Are you suggesting that this is a essentially an ABI problem? Maybe a 
>>> problem between Apple’s clang and a library built with gcc?
>>> 
>>> ​No. I have no idea is what is going on. But in both cases the error 
>>> happens when you are building documentation. The octave binary should be 
>>> built but that time, so you should be able to run 
>>> it and experiment with it.
>> 
>> I'm unable to run either the gui or cli.
>> 
>> ./run-octave --norc
>> 
>> GEN      doc/interpreter/voronoi.txt
>> ASAN:DEADLYSIGNAL
>> =================================================================
>> ==68284==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000040 (pc 
>> 0x7fff8d3994cd bp 0x7fff51568360 sp 0x7fff515682e0 T0)
>> 
>> ​I seem to remember that mac used to have "dtrace" utility. If ti still 
>> does, you can do 
>> dtrace ./run-octave -norc 
>> (there should be some flags to control verbosity) 
>> and see where it breaks.​
> 
> Things have changed on MacOS X.  However, it turned out to be fairly simple 
> to modify ./run-octave to use the lldb debugger (“gdb —args” -> “lldb —“). 
> Then “./run-octave -g” put me into the clang debugger.
> 
> (lldb) run
> Process 66739 launched: 
> '/Users/bpabbott/Development/mercurial/default/sources/src/octave' (x86_64)
> ASAN:DEADLYSIGNAL
> =================================================================
> ==66749==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000040 (pc 
> 0x7fff8d3994cd bp 0x7fff5fbfcb40 sp 0x7fff5fbfcac0 T0)
>     #0 0x7fff8d3994cc in _pthread_mutex_lock (libsystem_pthread.dylib+0x14cc)
>     #1 0x7fff92277be5 in closedir (libsystem_c.dylib+0x25be5)
>     #2 0x1027ce418 in 
> octave::sys::dir_entry::open(std::__1::basic_string<char, 
> std::__1::char_traits<char>, std::__1::allocator<char> > const&) dir-ops.cc:92
>     #3 0x1019da729 in 
> octave::sys::dir_entry::dir_entry(std::__1::basic_string<char, 
> std::__1::char_traits<char>, std::__1::allocator<char> > const&) dir-ops.h:46
>     #4 0x1016f177c in genpath(std::__1::basic_string<char, 
> std::__1::char_traits<char>, std::__1::allocator<char> > const&, 
> string_vector const&) load-path.cc:2146
>     #5 0x101545021 in set_image_path(std::__1::basic_string<char, 
> std::__1::char_traits<char>, std::__1::allocator<char> > const&) 
> defaults.cc:301
>     #6 0x10154625f in install_defaults() defaults.cc:455
>     #7 0x10000171d in main main-gui.cc:102
>     #8 0x7fff875c95c8 in start (libdyld.dylib+0x35c8)
>     #9 0x7  (octave-gui)+0x7)
> 
> AddressSanitizer can not provide additional info.
> SUMMARY: AddressSanitizer: SEGV (libsystem_pthread.dylib+0x14cc) in 
> _pthread_mutex_lock
> ==66749==ABORTING
> Process 66739 exited with status = 0 (0x00000000) 
> 
> I’m not familiar with Octave’s startup process. Does anyone see something 
> indicative the the problem?
> 
> Ben

With “./run-octave -g —no-gui” I get …

(lldb) run
Process 67054 launched: 
'/Users/bpabbott/Development/mercurial/default/sources/src/octave' (x86_64)
Process 67054 stopped
* thread #1: tid = 0x8ffa1a, 0x00007fff5fc01000 dyld`_dyld_start, stop reason = 
exec
    frame #0: 0x00007fff5fc01000 dyld`_dyld_start
dyld`_dyld_start:
->  0x7fff5fc01000 <+0>: popq   %rdi
    0x7fff5fc01001 <+1>: pushq  $0x0
    0x7fff5fc01003 <+3>: movq   %rsp, %rbp
    0x7fff5fc01006 <+6>: andq   $-0x10, %rsp

Ben