Re: OpenSolaris core dump ?

[Jim Langston]

I have another question, this fix was talked about, but
apparently it has not made it into the source tree, is there
a reason it did not ? Also, without initializing in liboctave/file-ops.cc
with:

   static std::string xdir_sep_chars ("/");


Octave will core dump on OpenSolaris, but with the initialization,
all the tests are passing (except for known reasons).

Jim

///////////////////////

Jim Langston wrote:
> Hi all,
>
> I'm working with OpenSolaris and applied this to Octave 3.0.5 as noted by
> John last August.
>
> bool
> file_ops::is_dir_sep (char c)
> {
> #if defined (OCTAVE_HAVE_WINDOWS_FILESYSTEM)
>     static std::string xdir_sep_chars ("/\\");
>   #else
>     // static std::string xdir_sep_chars (file_ops::dir_sep_str);
>     static std::string xdir_sep_chars ("/");
>   #endif
>   return xdir_sep_chars.find (c) != NPOS;
> }
>
>
> If I reverse the comment and depend on file_ops::dir_sep_str,
> I will get a core dump similar to what was in the bug report, ie:
>
> (gdb) bt
> #0  basic_string (this=0xfe2a9528, address@hidden) at basic_string.h:221
> #1  0xfdd4ae08 in file_ops::is_dir_sep (c=110 'n') at file-ops.cc:865
> #2  0xfdd4af03 in file_ops::concat (address@hidden, address@hidden)
>     at file-ops.cc:878
> #3  0xfe7e4e93 in default_history_file () at oct-hist.cc:91
> #4  0xfe7e5005 in __static_initialization_and_destruction_0 
> (__initialize_p=1,
>     __priority=65535) at oct-hist.cc:97
> #5  0xfe7e5098 in global constructors keyed to oct_hist.cc ()
>     at oct-hist.cc:739
> #6  0xfecb6c2d in __do_global_ctors_aux () from ./src/liboctinterp.so
> #7  0xfecb6c5a in _init () from ./src/liboctinterp.so
>
> setting the separator, no problems running the entire test harness
>
> Did I miss another static to set ?
>
> Thanks,
>
> Jim
>
>   


--
/////////////////////////////////////////////

Jim Langston
Sun Microsystems, Inc.

(513) 702-4741 (Cell)
(877) 854-5583 (AccessLine)
AIM: jl9594
address@hidden