Re: forking issue on Windows XP?

classic Classic list List threaded Threaded
42 messages Options
123
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
Hmmm, this is looking interesting.

re: msysinfo

msysinfo.txt attached for info.

Hugh

On 6/4/05, Michael S. Zick <[hidden email]> wrote:

> On Fri June 3 2005 15:31, Michael S. Zick wrote:
> > On Fri June 3 2005 03:44, Hugh Perkins wrote:
> >
> > I will see if I can work out a more informative test
> > than this uncontrolled one I just used.
> >
> Here we go, this one is well behaved.
> I will just inline it, it is small enough.
>
> execute it: ./rxtest <number>
> Where is "<number>" is some small number.
> The value of X in 2^X number of processes.
> Start testing with SMALL values of X.
>
> ksh.exe under uwin segfaults in posix.dll
> somewhere around 2^8 processes.
>
> #- - - - - test script - - - - -
> [ $# -eq 1 ] || exit 1
> [ ${1} -gt 0 ] || exit 1
>
> let lev=${1}-1
> echo 'hello '${lev}
>
> exec 3<>/dev/null
> exec 4<>/dev/null
> exec 5<>/dev/null
>
> ${0} ${lev} &
> ${0} ${lev} &
>
> sleep 10
> echo 'bye '${lev}
> exit 0
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> Mingw-msys mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/mingw-msys
>

msysinfo.txt (6K) Download Attachment
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
On Fri June 3 2005 21:07, Hugh Perkins wrote:
> Hmmm, this is looking interesting.
>
> re: msysinfo
>
> msysinfo.txt attached for info.
>
If we ever get your system past 'configure'...

If you will be building a modern C++ program...
Replace that gcc-2.95 with any recent 3.3.x set of
gcc-core/gcc-g++

If building under mingw/msys you should not need
any of the Microsoft tool kits, compilers, command line
manglers, etc.

If using a windows development IDE, it should be possible
to set it to use the external command line gcc/binutils toolchain.

Mike

> Hugh


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
re: threading, processor is P4 2.4G.  Not hyperthreaded, eg looking at
Device Manager, it says "ACPI Uniprocessor PC", and in taskmgr theres
just one processor graph.

Hugh

On 6/4/05, Michael S. Zick <[hidden email]> wrote:

> On Fri June 3 2005 21:07, Hugh Perkins wrote:
> > Hmmm, this is looking interesting.
> >
> > re: msysinfo
> >
> > msysinfo.txt attached for info.
> >
> If we ever get your system past 'configure'...
>
> If you will be building a modern C++ program...
> Replace that gcc-2.95 with any recent 3.3.x set of
> gcc-core/gcc-g++
>
> If building under mingw/msys you should not need
> any of the Microsoft tool kits, compilers, command line
> manglers, etc.
>
> If using a windows development IDE, it should be possible
> to set it to use the external command line gcc/binutils toolchain.
>
> Mike
>
> > Hugh
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> Mingw-msys mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/mingw-msys
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
In reply to this post by Hugh Perkins
An update of results to date;

Add the one line as shown below.
(Allows parent a chance to create the second job.)

Note:
Yes, I know that I left the file descriptors open.
They should 'close on exit'.

With swap file size set to a very low value -
force system to run out of physical AND virtual memory:

sleep.exe craters in msys dll - system makes a core dump.
at the same time the system presents the "virtual memory
size too small" message.

Expected behavior: "resource unavailable" and clean exit.

System recovers resources - and evidently dumps msys dll
since bash and its terminal go into the bit bucket also.

More when I know more.
Mike

> On Fri June 3 2005 15:31, Michael S. Zick wrote:
> > On Fri June 3 2005 03:44, Hugh Perkins wrote:
> >
> > I will see if I can work out a more informative test
> > than this uncontrolled one I just used.
> >
> Here we go, this one is well behaved.
> I will just inline it, it is small enough.
>
> execute it: ./rxtest <number>
> Where is "<number>" is some small number.
> The value of X in 2^X number of processes.
> Start testing with SMALL values of X.
>
> ksh.exe under uwin segfaults in posix.dll
> somewhere around 2^8 processes.
>
> #- - - - - test script - - - - -
> [ $# -eq 1 ] || exit 1
> [ ${1} -gt 0 ] || exit 1
>
> let lev=${1}-1
> echo 'hello '${lev}
>
# Add the following line
sleep 1

> exec 3<>/dev/null
> exec 4<>/dev/null
> exec 5<>/dev/null
>
> ${0} ${lev} &
> ${0} ${lev} &
>
> sleep 10
> echo 'bye '${lev}
> exit 0
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.  
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> Mingw-msys mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/mingw-msys
>
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
On Sat June 4 2005 13:08, Michael S. Zick wrote:
> An update of results to date;
>
> sleep.exe craters in msys dll - system makes a core dump.
> at the same time the system presents the "virtual memory
> size too small" message.
>
> Expected behavior: "resource unavailable" and clean exit.
>
Note 2: Don't even bother trying gnuwin32-sleep.exe

Note 3:
The PATH environment variable is not being searched
in a reliable manner in the background jobs.

No reason yet, just collecting observations.
Mike


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
On Sat June 4 2005 14:18, Michael S. Zick wrote:
> On Sat June 4 2005 13:08, Michael S. Zick wrote:
> > An update of results to date;
Hugh,

I have driven my system with msys-1.10 and bash-2.04
far beyond anything encountered in a 'configure' script.

Yes, I can get that combination to show problems, but I
do not think they are what you are seeing.

What can I say?

1) Make sure you do not have a broken sed(.exe)
get the newest one you can find - there is a note
somewhere on this subject.

2) Once inside the msys bash command shell...
echo $PATH
examine and set to the simplest possible path.

watch out for having directory paths to software
that might have been built with msys/cygwin.

Then set that simplest possible path by:
export PATH='/what/ever:/c/I decided:/to/use'
Try to avoid any paths with spaces in them;
if that is not possible, use the dos 8.3 format
for those names.

Group,

What did I find?
(I changed my rxtest to include PIDs in the messages.)

Probably the most interesting set is:

hello 0 PID: 10232
      0 [main] sh 5988 open_stackdumpfile: Dumping stack trace to sh.exe.stackdump
      0 [main] sh 8080 sync_with_child: child 2288(0x6D0) died before initialization with status code 0x80
    160 [main] sh 8080 sync_with_child: *** child state waiting for longjmp
./rxtest: fork: Resource temporarily unavailable
hello 0 PID: 6504
      0 [main] sh 5616 sync_with_child: child 2088(0x6D0) died before initialization with status code 0x80
    129 [main] sh 5616 sync_with_child: *** child state waiting for longjmp
./rxtest: fork: Resource temporarily unavailable
./rxtest: fork: Permission denied

Those errors are not recoverable - they should not be returning EAGAIN

The test run did eventually crater and zombie a half dozen copies of sh.exe
but the system did (after 5 minutes) clean up the mess.

On its way to eventually death, there where other errors reported.
I have attached the output of the full test run.

Mike

rxtest_9.txt.gz (12K) Download Attachment
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
On Sat June 4 2005 16:00, Michael S. Zick wrote:
> On Sat June 4 2005 14:18, Michael S. Zick wrote:
> > On Sat June 4 2005 13:08, Michael S. Zick wrote:
> > > An update of results to date;
> Hugh,
>
Hugh,

Check the properties of your menu entry and for any
desktop icon used to start msys...

You should expect to find it a shortcut to a batch file.
Drive:\msys\1.0\msys.bat is typical.

Open that batch file with your text editor.

Find the two places where bash-2.04 is started as sh.exe
The command options following should now be: --login -i
Add the following (without the quotes): ' -o posix'

Save the file and start msys

Enter: kill -l
That is the lower case letter ell.

If you get a single line listing signal names in uppercase
separated by a space...
You are in posix (sh) mode.

If you get a numbered listing of signal names...
You are in 'bash' mode.

Why:
On GNU/Linux, the native operating system of GNU/Bash,
executing the program by the name 'sh' puts bash into 'sh'
mode (including posix mode).

The msys-1.0.10/sh.exe (which is the bash executable) does
not mode shift when executed as 'sh'.

This makes a difference, about 32 differences that Chet
felt worth mentioning in the man page.

Scripts, including configure, that start with the splat line:
#! /bin/sh
Expect bash to be in posix mode as a result.
Some brain dead scripts actual check and believe the
value held in $SHELL (which is always /bin/sh in
msys-bash-2.04 regardless of what mode it is in).

Please let me know your experiences with this change.
If it improves your situation, I'll post a fixup patch for the
msys.bat file and the bash profile.
(There be other things that need fixups.)

Mike



-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Earnie Boyd
In reply to this post by Michael S. Zick
On 9:00:53 pm 2005-06-04 "Michael S. Zick" <[hidden email]> wrote:
> On Sat June 4 2005 14:18, Michael S. Zick wrote:
> >  On Sat June 4 2005 13:08, Michael S. Zick wrote:
> > >  An update of results to date;
> Hugh,
>
> I have driven my system with msys-1.10 and bash-2.04
> far beyond anything encountered in a 'configure' script.
>

It would be interesting to see what effects the CVS version show.

> Yes, I can get that combination to show problems, but I
> do not think they are what you are seeing.
>
> What can I say?
>

Yes exactly.  The msysinfo Hugh sent didn't show much interesting except
for the setting of CC and CXX to ccc1ming.  People are building some huge
projects ACE/TAO for one with MSYS and MinGW.

> 1) Make sure you do not have a broken sed(.exe)
> get the newest one you can find - there is a note
> somewhere on this subject.
>
> 2) Once inside the msys bash command shell...
> echo $PATH
> examine and set to the simplest possible path.
>
> watch out for having directory paths to software
> that might have been built with msys/cygwin.
>
> Then set that simplest possible path by:
> export PATH='/what/ever:/c/I decided:/to/use'
> Try to avoid any paths with spaces in them;
> if that is not possible, use the dos 8.3 format
> for those names.
>
> Group,
>
> What did I find?
> (I changed my rxtest to include PIDs in the messages.)
>
> Probably the most interesting set is:
>
> hello 0 PID: 10232
>       0 [main] sh 5988 open_stackdumpfile: Dumping stack trace to
> sh.exe.stackdump       0 [main] sh 8080 sync_with_child: child
> 2288(0x6D0) died before initialization with status code 0x80     160
> [main] sh 8080 sync_with_child: *** child state waiting for longjmp
> ./rxtest: fork: Resource temporarily unavailable hello 0 PID: 6504
>       0 [main] sh 5616 sync_with_child: child 2088(0x6D0) died before
> initialization with status code 0x80     129 [main] sh 5616
> sync_with_child: *** child state waiting for longjmp ./rxtest: fork:
> Resource temporarily unavailable ./rxtest: fork: Permission denied
>
> Those errors are not recoverable - they should not be returning EAGAIN
>
> The test run did eventually crater and zombie a half dozen copies of
> sh.exe but the system did (after 5 minutes) clean up the mess.
>
> On its way to eventually death, there where other errors reported.
> I have attached the output of the full test run.
>

Mike, since you've done this much work I hope you go the next step and
build a debugging version of MSYS.  Then execute your tests with strace.

Also, your tests are using children that are MSYS dependent.  The children
that are native binaries that call children that are MSYS dependent may
show different results.

Earnie

--
       MinGW - http://www.mingw.org/
        Wiki - http://www.mingw.org/MinGWiki/
  Bug Report - http://sourceforge.net/tracker/?group_id=2435&atid=102435
Submit Patch - http://sourceforge.net/tracker/?group_id=2435&atid=302435
  SF Project - http://sourceforge.net/projects/mingw
 Job Listing - http://sf.net/people/viewjob.php?group_id=2435&job_id=21643
 Job Listing - http://sf.net/people/viewjob.php?group_id=46778&job_id=22223



-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
In reply to this post by Michael S. Zick
Ok, so I did:
- added -o posix to start sh line in msys.bat
- did kill -l -> uppercase signal names (but actually I get this even
without -o posix, and even if I type sh first in msys)
- ./run doesnt use sed, so I dont suppose it matters for the purposes of run?
- simplied path to:
export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
(you need the system32 path, otherwise mingw crashes if you resize the window)
- dir c:\windows\*yg* -> nothing (from cmd)
- dir c:\windows\system32\*yg* -> nothing (from cmd)
- ls /usr/*yg* -> nothing  (from msys)

- opened taskmgr
- ran ./run
-> memory usage continued to climb at 2meg per second
- the memory was not reclaimed after stopping the script
- nor after exiting msys
- nor after waiting some minutes

Pretty sure there is a fairly serious memory leak here?  Obviously if
I want to migrate my own project (http://metaverse.sf.net) to use
mingw configure, I'm going to need to run configure and so on multiple
times to get it right, and having to reboot after each configure run
is obviously not something I'm really comfortable with!

Hugh

On 6/5/05, Michael S. Zick <[hidden email]> wrote:

> On Sat June 4 2005 16:00, Michael S. Zick wrote:
> > On Sat June 4 2005 14:18, Michael S. Zick wrote:
> > > On Sat June 4 2005 13:08, Michael S. Zick wrote:
> > > > An update of results to date;
> > Hugh,
> >
> Hugh,
>
> Check the properties of your menu entry and for any
> desktop icon used to start msys...
>
> You should expect to find it a shortcut to a batch file.
> Drive:\msys\1.0\msys.bat is typical.
>
> Open that batch file with your text editor.
>
> Find the two places where bash-2.04 is started as sh.exe
> The command options following should now be: --login -i
> Add the following (without the quotes): ' -o posix'
>
> Save the file and start msys
>
> Enter: kill -l
> That is the lower case letter ell.
>
> If you get a single line listing signal names in uppercase
> separated by a space...
> You are in posix (sh) mode.
>
> If you get a numbered listing of signal names...
> You are in 'bash' mode.
>
> Why:
> On GNU/Linux, the native operating system of GNU/Bash,
> executing the program by the name 'sh' puts bash into 'sh'
> mode (including posix mode).
>
> The msys-1.0.10/sh.exe (which is the bash executable) does
> not mode shift when executed as 'sh'.
>
> This makes a difference, about 32 differences that Chet
> felt worth mentioning in the man page.
>
> Scripts, including configure, that start with the splat line:
> #! /bin/sh
> Expect bash to be in posix mode as a result.
> Some brain dead scripts actual check and believe the
> value held in $SHELL (which is always /bin/sh in
> msys-bash-2.04 regardless of what mode it is in).
>
> Please let me know your experiences with this change.
> If it improves your situation, I'll post a fixup patch for the
> msys.bat file and the bash profile.
> (There be other things that need fixups.)
>
> Mike
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> Mingw-msys mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/mingw-msys
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
I suppose it could feasibly a kernel-mode I/O-type driver on my system
that is leaking, making the problem specific to my system.  That would
suck :-/

Hugh

On 6/5/05, Hugh Perkins <[hidden email]> wrote:

> Ok, so I did:
> - added -o posix to start sh line in msys.bat
> - did kill -l -> uppercase signal names (but actually I get this even
> without -o posix, and even if I type sh first in msys)
> - ./run doesnt use sed, so I dont suppose it matters for the purposes of run?
> - simplied path to:
> export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
> (you need the system32 path, otherwise mingw crashes if you resize the window)
> - dir c:\windows\*yg* -> nothing (from cmd)
> - dir c:\windows\system32\*yg* -> nothing (from cmd)
> - ls /usr/*yg* -> nothing  (from msys)
>
> - opened taskmgr
> - ran ./run
> -> memory usage continued to climb at 2meg per second
> - the memory was not reclaimed after stopping the script
> - nor after exiting msys
> - nor after waiting some minutes
>
> Pretty sure there is a fairly serious memory leak here?  Obviously if
> I want to migrate my own project (http://metaverse.sf.net) to use
> mingw configure, I'm going to need to run configure and so on multiple
> times to get it right, and having to reboot after each configure run
> is obviously not something I'm really comfortable with!
>
> Hugh
>
> On 6/5/05, Michael S. Zick <[hidden email]> wrote:
> > On Sat June 4 2005 16:00, Michael S. Zick wrote:
> > > On Sat June 4 2005 14:18, Michael S. Zick wrote:
> > > > On Sat June 4 2005 13:08, Michael S. Zick wrote:
> > > > > An update of results to date;
> > > Hugh,
> > >
> > Hugh,
> >
> > Check the properties of your menu entry and for any
> > desktop icon used to start msys...
> >
> > You should expect to find it a shortcut to a batch file.
> > Drive:\msys\1.0\msys.bat is typical.
> >
> > Open that batch file with your text editor.
> >
> > Find the two places where bash-2.04 is started as sh.exe
> > The command options following should now be: --login -i
> > Add the following (without the quotes): ' -o posix'
> >
> > Save the file and start msys
> >
> > Enter: kill -l
> > That is the lower case letter ell.
> >
> > If you get a single line listing signal names in uppercase
> > separated by a space...
> > You are in posix (sh) mode.
> >
> > If you get a numbered listing of signal names...
> > You are in 'bash' mode.
> >
> > Why:
> > On GNU/Linux, the native operating system of GNU/Bash,
> > executing the program by the name 'sh' puts bash into 'sh'
> > mode (including posix mode).
> >
> > The msys-1.0.10/sh.exe (which is the bash executable) does
> > not mode shift when executed as 'sh'.
> >
> > This makes a difference, about 32 differences that Chet
> > felt worth mentioning in the man page.
> >
> > Scripts, including configure, that start with the splat line:
> > #! /bin/sh
> > Expect bash to be in posix mode as a result.
> > Some brain dead scripts actual check and believe the
> > value held in $SHELL (which is always /bin/sh in
> > msys-bash-2.04 regardless of what mode it is in).
> >
> > Please let me know your experiences with this change.
> > If it improves your situation, I'll post a fixup patch for the
> > msys.bat file and the bash profile.
> > (There be other things that need fixups.)
> >
> > Mike
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> > a projector? How fast can you ride your desk chair down the office luge track?
> > If you want to score the big prize, get to know the little guy.
> > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> > _______________________________________________
> > Mingw-msys mailing list
> > [hidden email]
> > https://lists.sourceforge.net/lists/listinfo/mingw-msys
> >
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
In reply to this post by Earnie Boyd
On Sat June 4 2005 21:27, Earnie Boyd wrote:

> On 9:00:53 pm 2005-06-04 "Michael S. Zick" <[hidden email]> wrote:
> > On Sat June 4 2005 14:18, Michael S. Zick wrote:
> > >  On Sat June 4 2005 13:08, Michael S. Zick wrote:
> > > >  An update of results to date;
> > Hugh,
> >
> > I have driven my system with msys-1.10 and bash-2.04
> > far beyond anything encountered in a 'configure' script.
> >
>
> It would be interesting to see what effects the CVS version show.
>
I am following in Max's footsteps here, I am still learning to
install the build environment.
Long time before I know enough to compile anything.

But you are correct - it would be nice to track changes to the
CVS version.
That little script can be modified to consume whatever resources
you might like to test.
Could be much different with the bash-2.05b that is in CVS also.

- - - - -
>
> Mike, since you've done this much work I hope you go the next step and
> build a debugging version of MSYS.  Then execute your tests with strace.
>
Any idea if valgrind has been ported win-native?
Or if it could track into msys.dll?

> Also, your tests are using children that are MSYS dependent.  The children
> that are native binaries that call children that are MSYS dependent may
> show different results.
>
Only tried GnuWin32-sleep.exe (by accident of path setting).
You really don't want to know what happened.  Made a horrible mess.

The one try running it under ksh / uwin just showed horrible problems
with path handling and environment passing.

As a base-line, that script, run under GNU/Bash on GNU/Linux can take
the kernel into its 'geometric system overhead growth mode' - but Bash
doesn't break.  (I also once wrote a sched.c that stops that problem)

Mike


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
In reply to this post by Hugh Perkins
On Sat June 4 2005 21:31, Hugh Perkins wrote:
> Ok, so I did:
> - added -o posix to start sh line in msys.bat
> - did kill -l -> uppercase signal names (but actually I get this even
> without -o posix, and even if I type sh first in msys)
>
Maybe I wasn't clear - the names are always upper case;
posix (sh): single output line, single space between each name.
bash: multiple output lines, multiple columns, format: #) NAME #) NAME...

> - ./run doesnt use sed, so I dont suppose it matters for the purposes of run?
>
Is the goal to get your test script to work or to get your './configure' to run?

Note: I had to uninstall my LyX, which included QT built against msys.
There are a few yards of C++ in QT.

> - simplied path to:
> export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
> (you need the system32 path, otherwise mingw crashes if you resize the window)
> - dir c:\windows\*yg* -> nothing (from cmd)
> - dir c:\windows\system32\*yg* -> nothing (from cmd)
> - ls /usr/*yg* -> nothing  (from msys)
>
> - opened taskmgr
>
Try the system/performance monitor graphic thingy.

> - ran ./run
> -> memory usage continued to climb at 2meg per second
> - the memory was not reclaimed after stopping the script
> - nor after exiting msys
> - nor after waiting some minutes
>
> Pretty sure there is a fairly serious memory leak here?
>
Your test is calling the g++ wrapper to the gcc frontend.
Lots of room for problems there.

And if your msysinfo is to be believed - your calling v-2.95
which was retired about 4 years ago.

Try calling something other than g++ in your test script.
I was calling msys/sleep.exe

Mike


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
> Your test is calling the g++ wrapper to the gcc frontend.
> Lots of room for problems there.

Well, in reality it will be calling out to cl.exe, but the same issue
is apparent with g++.  I'm using g++ in the test script because most
people with msys have g++, but not everyone has cl.exe.

> Is the goal to get your test script to work or to get your './configure' to run?

My goal is to feel comfortable that I can develop in msys all day
without having to reboot my computer.

Hugh

On 6/5/05, Michael S. Zick <[hidden email]> wrote:

> On Sat June 4 2005 21:31, Hugh Perkins wrote:
> > Ok, so I did:
> > - added -o posix to start sh line in msys.bat
> > - did kill -l -> uppercase signal names (but actually I get this even
> > without -o posix, and even if I type sh first in msys)
> >
> Maybe I wasn't clear - the names are always upper case;
> posix (sh): single output line, single space between each name.
> bash: multiple output lines, multiple columns, format: #) NAME #) NAME...
>
> > - ./run doesnt use sed, so I dont suppose it matters for the purposes of run?
> >
> Is the goal to get your test script to work or to get your './configure' to run?
>
> Note: I had to uninstall my LyX, which included QT built against msys.
> There are a few yards of C++ in QT.
>
> > - simplied path to:
> > export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
> > (you need the system32 path, otherwise mingw crashes if you resize the window)
> > - dir c:\windows\*yg* -> nothing (from cmd)
> > - dir c:\windows\system32\*yg* -> nothing (from cmd)
> > - ls /usr/*yg* -> nothing  (from msys)
> >
> > - opened taskmgr
> >
> Try the system/performance monitor graphic thingy.
>
> > - ran ./run
> > -> memory usage continued to climb at 2meg per second
> > - the memory was not reclaimed after stopping the script
> > - nor after exiting msys
> > - nor after waiting some minutes
> >
> > Pretty sure there is a fairly serious memory leak here?
> >
> Your test is calling the g++ wrapper to the gcc frontend.
> Lots of room for problems there.
>
> And if your msysinfo is to be believed - your calling v-2.95
> which was retired about 4 years ago.
>
> Try calling something other than g++ in your test script.
> I was calling msys/sleep.exe
>
> Mike
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> Mingw-msys mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/mingw-msys
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
Just to confirm, the issue is also apparent using sleep.exe, but its
not quite so marked.  Possibly sleep has a smaller memory footprint
and that is influencing how much memory is being consumed at each
call?

The memory is not reclaimed after stopping the script or exiting msys
or waiting a while, so the effect is cumulative.  There's no way to
cleanup short of rebooting AFAIK.

Hugh

On 6/5/05, Hugh Perkins <[hidden email]> wrote:

> > Your test is calling the g++ wrapper to the gcc frontend.
> > Lots of room for problems there.
>
> Well, in reality it will be calling out to cl.exe, but the same issue
> is apparent with g++.  I'm using g++ in the test script because most
> people with msys have g++, but not everyone has cl.exe.
>
> > Is the goal to get your test script to work or to get your './configure' to run?
>
> My goal is to feel comfortable that I can develop in msys all day
> without having to reboot my computer.
>
> Hugh
>
> On 6/5/05, Michael S. Zick <[hidden email]> wrote:
> > On Sat June 4 2005 21:31, Hugh Perkins wrote:
> > > Ok, so I did:
> > > - added -o posix to start sh line in msys.bat
> > > - did kill -l -> uppercase signal names (but actually I get this even
> > > without -o posix, and even if I type sh first in msys)
> > >
> > Maybe I wasn't clear - the names are always upper case;
> > posix (sh): single output line, single space between each name.
> > bash: multiple output lines, multiple columns, format: #) NAME #) NAME...
> >
> > > - ./run doesnt use sed, so I dont suppose it matters for the purposes of run?
> > >
> > Is the goal to get your test script to work or to get your './configure' to run?
> >
> > Note: I had to uninstall my LyX, which included QT built against msys.
> > There are a few yards of C++ in QT.
> >
> > > - simplied path to:
> > > export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
> > > (you need the system32 path, otherwise mingw crashes if you resize the window)
> > > - dir c:\windows\*yg* -> nothing (from cmd)
> > > - dir c:\windows\system32\*yg* -> nothing (from cmd)
> > > - ls /usr/*yg* -> nothing  (from msys)
> > >
> > > - opened taskmgr
> > >
> > Try the system/performance monitor graphic thingy.
> >
> > > - ran ./run
> > > -> memory usage continued to climb at 2meg per second
> > > - the memory was not reclaimed after stopping the script
> > > - nor after exiting msys
> > > - nor after waiting some minutes
> > >
> > > Pretty sure there is a fairly serious memory leak here?
> > >
> > Your test is calling the g++ wrapper to the gcc frontend.
> > Lots of room for problems there.
> >
> > And if your msysinfo is to be believed - your calling v-2.95
> > which was retired about 4 years ago.
> >
> > Try calling something other than g++ in your test script.
> > I was calling msys/sleep.exe
> >
> > Mike
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> > a projector? How fast can you ride your desk chair down the office luge track?
> > If you want to score the big prize, get to know the little guy.
> > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> > _______________________________________________
> > Mingw-msys mailing list
> > [hidden email]
> > https://lists.sourceforge.net/lists/listinfo/mingw-msys
> >
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
Hmmm, this is interesting.  Running taskmgr in parallel with the
original run from the bug report that was submitted in March.

The permission denied error occurs *before* memory is saturated, ie I
still have 200meg of free meory.  This is evidence that there is some
kernel resource (file handles? object handles?) that is being consumed
and not released?

Hugh

On 6/5/05, Hugh Perkins <[hidden email]> wrote:

> Just to confirm, the issue is also apparent using sleep.exe, but its
> not quite so marked.  Possibly sleep has a smaller memory footprint
> and that is influencing how much memory is being consumed at each
> call?
>
> The memory is not reclaimed after stopping the script or exiting msys
> or waiting a while, so the effect is cumulative.  There's no way to
> cleanup short of rebooting AFAIK.
>
> Hugh
>
> On 6/5/05, Hugh Perkins <[hidden email]> wrote:
> > > Your test is calling the g++ wrapper to the gcc frontend.
> > > Lots of room for problems there.
> >
> > Well, in reality it will be calling out to cl.exe, but the same issue
> > is apparent with g++.  I'm using g++ in the test script because most
> > people with msys have g++, but not everyone has cl.exe.
> >
> > > Is the goal to get your test script to work or to get your './configure' to run?
> >
> > My goal is to feel comfortable that I can develop in msys all day
> > without having to reboot my computer.
> >
> > Hugh
> >
> > On 6/5/05, Michael S. Zick <[hidden email]> wrote:
> > > On Sat June 4 2005 21:31, Hugh Perkins wrote:
> > > > Ok, so I did:
> > > > - added -o posix to start sh line in msys.bat
> > > > - did kill -l -> uppercase signal names (but actually I get this even
> > > > without -o posix, and even if I type sh first in msys)
> > > >
> > > Maybe I wasn't clear - the names are always upper case;
> > > posix (sh): single output line, single space between each name.
> > > bash: multiple output lines, multiple columns, format: #) NAME #) NAME...
> > >
> > > > - ./run doesnt use sed, so I dont suppose it matters for the purposes of run?
> > > >
> > > Is the goal to get your test script to work or to get your './configure' to run?
> > >
> > > Note: I had to uninstall my LyX, which included QT built against msys.
> > > There are a few yards of C++ in QT.
> > >
> > > > - simplied path to:
> > > > export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
> > > > (you need the system32 path, otherwise mingw crashes if you resize the window)
> > > > - dir c:\windows\*yg* -> nothing (from cmd)
> > > > - dir c:\windows\system32\*yg* -> nothing (from cmd)
> > > > - ls /usr/*yg* -> nothing  (from msys)
> > > >
> > > > - opened taskmgr
> > > >
> > > Try the system/performance monitor graphic thingy.
> > >
> > > > - ran ./run
> > > > -> memory usage continued to climb at 2meg per second
> > > > - the memory was not reclaimed after stopping the script
> > > > - nor after exiting msys
> > > > - nor after waiting some minutes
> > > >
> > > > Pretty sure there is a fairly serious memory leak here?
> > > >
> > > Your test is calling the g++ wrapper to the gcc frontend.
> > > Lots of room for problems there.
> > >
> > > And if your msysinfo is to be believed - your calling v-2.95
> > > which was retired about 4 years ago.
> > >
> > > Try calling something other than g++ in your test script.
> > > I was calling msys/sleep.exe
> > >
> > > Mike
> > >
> > >
> > > -------------------------------------------------------
> > > This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> > > a projector? How fast can you ride your desk chair down the office luge track?
> > > If you want to score the big prize, get to know the little guy.
> > > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> > > _______________________________________________
> > > Mingw-msys mailing list
> > > [hidden email]
> > > https://lists.sourceforge.net/lists/listinfo/mingw-msys
> > >
> >
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Earnie Boyd
In reply to this post by Hugh Perkins
On 2:31:24 am 2005-06-05 Hugh Perkins <[hidden email]> wrote:

> Ok, so I did:
> - added -o posix to start sh line in msys.bat
> - did kill -l -> uppercase signal names (but actually I get this even
> without -o posix, and even if I type sh first in msys)
> - ./run doesnt use sed, so I dont suppose it matters for the purposes
> of run? - simplied path to:
> export PATH=/usr:/f/PROGRA~1/MICROS~2/bin:/c/WINDOWS/system32
> (you need the system32 path, otherwise mingw crashes if you resize
> the window) - dir c:\windows\*yg* -> nothing (from cmd)
> - dir c:\windows\system32\*yg* -> nothing (from cmd)
> - ls /usr/*yg* -> nothing  (from msys)
>
> - opened taskmgr
> - ran ./run
> -> memory usage continued to climb at 2meg per second

I didn't see this.  Really I tried your test case.  So the purpose of
ccclming is to convert the cl command line to gcc/g++?

> - the memory was not reclaimed after stopping the script
> - nor after exiting msys
> - nor after waiting some minutes
>

I suspect something else is wrong if your memory doesn't reclaim on its own
after exiting the programs using it.  When ran your test what I saw was a
fluctuation of memory usage of about 4 meg.  Memory was reclaimed on exit.

> Pretty sure there is a fairly serious memory leak here?  Obviously if
> I want to migrate my own project (http://metaverse.sf.net) to use
> mingw configure, I'm going to need to run configure and so on multiple
> times to get it right, and having to reboot after each configure run
> is obviously not something I'm really comfortable with!
>

I'm not convinced the problem is with MSYS.  ACE/TAO uses MSYS and MinGW to
build its win32 binaries.  That is one huge project that takes hours to
build from scratch.  I use MSYS and MinGW to build other projects as well.

As a test, can you configure binutils from source?

As a test, do you have another computer from which to try your configure
script?  Is the configure script available yet; I didn't see it in CVS?

Earnie

--
       MinGW - http://www.mingw.org/
        Wiki - http://www.mingw.org/MinGWiki/
  Bug Report - http://sourceforge.net/tracker/?group_id=2435&atid=102435
Submit Patch - http://sourceforge.net/tracker/?group_id=2435&atid=302435
  SF Project - http://sourceforge.net/projects/mingw
 Job Listing - http://sf.net/people/viewjob.php?group_id=2435&job_id=21643
 Job Listing - http://sf.net/people/viewjob.php?group_id=46778&job_id=22223



-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Earnie Boyd
In reply to this post by Hugh Perkins
On 3:19:41 am 2005-06-05 Hugh Perkins <[hidden email]> wrote:
>
> My goal is to feel comfortable that I can develop in msys all day
> without having to reboot my computer.
>

LOL, I use multiple session of MSYS rxvt daily and I mean daily.  It is a
tool I use in my job.

I also configure and make packages often, I have yet to have the problems
you experience.

Earnie



-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Michael S. Zick
In reply to this post by Hugh Perkins
On Sat June 4 2005 22:27, Hugh Perkins wrote:
> Hmmm, this is interesting.  Running taskmgr in parallel with the
> original run from the bug report that was submitted in March.
>
> The permission denied error occurs *before* memory is saturated, ie I
> still have 200meg of free meory.  This is evidence that there is some
> kernel resource (file handles? object handles?) that is being consumed
> and not released?
>
The 'permission denied' error is a red herring.  Sort of a
general purpose "I couldn't do that - can't say why".

It is the not getting the 'resource temporarily unavailable' VS
'resource unavailable' returns properly handled.

Use the performance monitor graph to track "free page table entries" -

I think you will find that windows can not recover from that while programs
continue to need pages - for whatever reason.
But finding why the return is EAGAIN rather than ENOMEM will lead to
avoiding this windows gotcha.

Also, you can expect a near deadlock condition when it tries to reclaim
swap space when short of free page table entries.

Note1 : I am running with 384Mbyte physical.

Note2 : I have tried to execute your run script - had to pull the
ac power off of the machine to regain control.

That is why I posted something with behavior easy to control.

Mike
Oh: start/control panel/administrative ???/performance monitor
or something like that.

> Hugh


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Earnie Boyd
In reply to this post by Michael S. Zick
On 2:58:44 am 2005-06-05 "Michael S. Zick" <[hidden email]> wrote:

> Any idea if valgrind has been ported win-native?
> Or if it could track into msys.dll?
>

No, I don't know.  There are macros to turn on malloc debugging in MSYS.
I've never tried it.  I have traced the function mess (and it is a mess) to
try to track down earlier issues I was having with memory allocation.

> >  Also, your tests are using children that are MSYS dependent.  The
> >  children that are native binaries that call children that are MSYS
> >  dependent may show different results.
> >
> Only tried GnuWin32-sleep.exe (by accident of path setting).
> You really don't want to know what happened.  Made a horrible mess.
>

I can imagine, one of the problems is controlling children not using the
msys dll.  For instance the pty emulation issue for native programs.

Earnie

--
       MinGW - http://www.mingw.org/
        Wiki - http://www.mingw.org/MinGWiki/
  Bug Report - http://sourceforge.net/tracker/?group_id=2435&atid=102435
Submit Patch - http://sourceforge.net/tracker/?group_id=2435&atid=302435
  SF Project - http://sourceforge.net/projects/mingw
 Job Listing - http://sf.net/people/viewjob.php?group_id=2435&job_id=21643
 Job Listing - http://sf.net/people/viewjob.php?group_id=46778&job_id=22223



-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: forking issue on Windows XP?

Hugh Perkins
In reply to this post by Earnie Boyd
Well, thanks for trying anyway.  I guess as more people start to use
mingw other people will see the same issue.  In the meantime
gnumake.exe is ok, but not perfect.

> I didn't see this.  Really I tried your test case.  So the purpose of
ccclming is to convert the cl command line to gcc/g++?

ccclming converts g++ command to cl

> Is the configure script available yet; I didn't see it in CVS?

Well, obviously I'm not going to consider migrating my project to a
build configuration that crashes my computer :-O

Hugh

On 6/5/05, Earnie Boyd <[hidden email]> wrote:

> On 3:19:41 am 2005-06-05 Hugh Perkins <[hidden email]> wrote:
> >
> > My goal is to feel comfortable that I can develop in msys all day
> > without having to reboot my computer.
> >
>
> LOL, I use multiple session of MSYS rxvt daily and I mean daily.  It is a
> tool I use in my job.
>
> I also configure and make packages often, I have yet to have the problems
> you experience.
>
> Earnie
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> Mingw-msys mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/mingw-msys
>


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r 
_______________________________________________
Mingw-msys mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/mingw-msys
123
Loading...