crash on latest cygwin snapshot

[marco atzeri]

On 6/26/2012 5:31 PM, Corinna Vinschen wrote:
> On Jun 26 11:04, Christopher Faylor wrote:
>> On Tue, Jun 26, 2012 at 10:10:28AM -0400, Christopher Faylor wrote:
>>> On Tue, Jun 26, 2012 at 07:53:05AM -0400, Ken Brown wrote:
>>>> On 6/24/2012 5:30 PM, marco atzeri wrote:
>>>>> while building latest atlas I hit a bug that is crashing all the running
>>>>> cygwin process.
>>>>> It crashes on 20120619 snapshot and also latest cvs source.
>>>>> No issue on 20120611 snapshot.
>>>>
>>>> I'm also seeing this crash, but it does not kill all running cygwin
>>>> processes.  The only processes that die are the mintty processes and the
>>>> associated bash processes.  For example, it just happened a few minutes
>>>> ago, and XWin survived, along with emacs and xterm.
>>>>
>>>> It happens intermittently, maybe once every 3 days, and I haven't found
>>>> a simple sequence of steps for reproducing it.  All I can say for sure
>>>> is that I always have more than one mintty running, and they all die.
>>>>
>>>> Marco, are you sure all cygwin processes die when you experience this
>>>> problem?  Also, since you can reliably reproduce the crash, have you
>>>> tried reverting to the previous version of mintty?  Note that mintty was
>>>> updated just a few days before the 20120619 snapshot was released.
>>>
>>> He indicated that some processes kept running.
>>>
>>> But, hmm.  I haven't updated mintty lately.  I will try that.
>>
>> This still just hangs my system by creating many /bin/sh and gcc jobs.
>> What am I supposed to be seeing for output?  If I set the PATH to just
>> /usr/bin, I see:
>>
>> User Override Compilers:
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>     'none' : 'none' '-fno-common'
>>
>> and then no more output until I have to restart the machine due to
>> excessive process creation.
>
> This might be a result of the problem, too.  When I run the command
> on W7 and W2008R2, it hangs a couple of seconds at this point, apparently
> running a find(1) command, and then the output starts like this:
>
> --- SNIP ---
> ierr=256 in command='find $HOME/local /home/corinna/bin /usr/local/bin /usr/bin /bin /mnt/c/Windows/system32 /mnt/c/Windows /mnt/c/Windows/System32/Wbem /mnt/c/Windows/System32/WindowsPowerShell/v1.0 -name '*gcc*' -exec ./xisgcc '{}' \;'!
>
> OUTPUT:
> =======
> find: `/home/corinna/local': No such file or directory
> /usr/bin/gcc-4.exe
> /usr/bin/gcc.exe
> /usr/bin/i686-pc-cygwin-gcc-4.5.3.exe
> /usr/bin/i686-pc-cygwin-gcc-4.exe
> /usr/bin/i686-pc-cygwin-gcc.exe
> /usr/bin/i686-pc-mingw32-gcc-4.5.2.exe
> /usr/bin/i686-pc-mingw32-gcc.exe
> /usr/bin/x86_64-w64-mingw32-gcc-4.5.3.exe
> /usr/bin/x86_64-w64-mingw32-gcc.exe
> /bin/gcc-4.exe
> /bin/gcc.exe
> /bin/i686-pc-cygwin-gcc-4.5.3.exe
> /bin/i686-pc-cygwin-gcc-4.exe
> /bin/i686-pc-cygwin-gcc.exe
> /bin/i686-pc-mingw32-gcc-4.5.2.exe
> /bin/i686-pc-mingw32-gcc.exe
> /bin/x86_64-w64-mingw32-gcc-4.5.3.exe
> /bin/x86_64-w64-mingw32-gcc.exe
> [...]
> --- SNAP ---
>
> and then a couple of hundreds line of output with various make commands
> or something like that.
>
>
> Corinna
>

This is the expected behaviour.
The program is part of a "weird" configure scripts used by ATLAS
http://math-atlas.sourceforge.net/
to detect the best compiler options for optimizing the math
perfomance on the given machine.

It stresses a bit the computer, but on snapshot 20110619 and later CVS
the side effect is to crash all mintty sessions on my W7/64 and
to block cgf's one.

So some of the change between 20110611 and 20110619 are responsable of
the mintty crash.

Marco


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple