Re: OpenBSD hangs / crashes with Suduku

[Sebastian Reitenbach]

On Wednesday, June 22, 2011 11:04 CEST, Riccardo Mottola 
<riccardo.mottola@libero.it> wrote:

> Hi,
>
> I am able to reproduce the problems Sebastian has on my own OpenBSD/x86 box. 
> I get this problem only on OpenBSD. I use gcc and its standard runtime. The 
> same combination appears to work fine on NetBSD or Linux.

>
> The problem happens with "New" inside Sudoku (latest version from CVS of GAP).

The thing is, it works well with gcc-2.95, using "old" libobjc1 runtime. The 
problem only shows up for me on i386.
I also tried to use libbacktrace* on i386, which is a replacement for 
libexecinfo. Then, interestingly it crashes, instead of hanging. I think with 
the hand rolled backtrace generation in gnustep, iit hangs because of fiddling 
around with the signals around it.
Also the test program that comes with libbacktrace crashes for me. I created a 
port for libbacktrace and sent it to the author, who is also an OpenBSD 
developer, but did not got an answer back yet. Since the example program also 
crashes, I suspect nothing to do with gnustep in general, but with the 
__builtin_... function used there, doing sth. strange on i386.

Sebastian

* http://opensource.conformal.com/wiki/Backtrace

>
> The happening is:
>
> 1) Sudoku is run  normally (not in GDB). Sudokup.apps remains eternally at 
> 99% of CPU usage and hangs
> 2) If run inside GDB, I get the same broken crash of Sebastian:
> #0  0x0d8ea287 in NSCountFrames () at NSDebug.m:1009
> #1  0x0d8edf19 in GSPrivateStackAddresses () at NSDebug.m:1112
> #2  0x0d902f5f in -[GSStackTrace init] (self=0x7d17a2e8, _cmd=0x2d8e89e8)
>      at NSException.m:605
> #3  0x0d951942 in +[NSObject new] (self=0x2d8cfac0, _cmd=0x2d8cfb40)
>      at NSObject.m:1281
> #4  0x0d90387c in -[NSException raise] (self=0x7d17a168, _cmd=0x2d8cfba8)
>      at NSException.m:960
> #5  0x0d902e04 in +[NSException raise:format:arguments:] (self=0x2d8cf9c0,
>      _cmd=0x2d8cfb90, name=0x3c00615c, format=0x3c006150,
>      argList=0xcfbf0a80 "ðR÷\213h¾O'\027\003") at NSException.m:849
> #6  0x0d903cf3 in +[NSException raise:format:] (self=0x2d8cf9c0,
>      _cmd=0x3c0060f8, name=0x3c00615c, format=0x3c006150) at NSException.m:835
> #7  0x1c0070ad in -[Sudoku doFind:] (self=0x7e509008, _cmd=0x3c0060d8,
>      seen=0x87e36fc8) at Sudoku.m:457
> #8  0x1c006fbe in -[Sudoku doFind:] (self=0x7e509008, _cmd=0x3c0060d8,
>      seen=0x87e36fc8) at Sudoku.m:524
> #9  0x1c006fbe in -[Sudoku doFind:] (self=0x7e509008, _cmd=0x3c0060d8,
>      seen=0x87e36fc8) at Sudoku.m:524
> #10 0x1c006fbe in -[Sudoku doFind:] (self=0x7e509008, _cmd=0x3c0060d8,
>      seen=0x87e36fc8) at Sudoku.m:524
> #11 0x1c006fbe in -[Sudoku doFind:] (self=0x7e509008, _cmd=0x3c0060d8,
>      seen=0x87e36fc8) at Sudoku.m:524
>
>
> I can also put a breakpoint in NSException raise and get the similar output:
>
> #0  -[NSException raise] (self=0x7c139488, _cmd=0x2835eba8)
>      at NSException.m:956
> #1  0x08391e04 in +[NSException raise:format:arguments:] (self=0x2835e9c0,
>      _cmd=0x2835eb90, name=0x3c006174, format=0x3c006168,
>      argList=0xcfbf01f0 "0÷\234\201h®\210&\027\003") at NSException.m:849
> #2  0x08392cf3 in +[NSException raise:format:] (self=0x2835e9c0,
>      _cmd=0x3c0060f8, name=0x3c006174, format=0x3c006168) at NSException.m:835
> #3  0x1c006f80 in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:452
> #4  0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
> #5  0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
> #6  0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
> #7  0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
> #8  0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
> #9  0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
> #10 0x1c006fbe in -[Sudoku doFind:] (self=0x84f0a008, _cmd=0x3c0060d8,
>      seen=0x87651648) at Sudoku.m:524
>
> the exception is raised from line 452;
>        [NSException raise:EX_COMPLETE format:EX_COMPLETE_FMT];

>
> line 524:
>            [self doFind:seen];
>
> (which means indeed recursion)
>
>
> Seen as this it looks like a code problem, but if it happens only on  OpenBSD 
> I think of some kind of corruption. "placed" gets initialized before...
>
> Thanks,
>    Riccardo
>
>
> and seq is an object instance variable, which makes me think of corruption in 
> the object.
>
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep