rxvt loses output connection with non cygwin console processes

rxvt loses output connection with non cygwin console processes

[Pawel Jasinski]

hi,

i have discovered something peculiar.
I run my rxvt with the usual:

C:\cygwin\bin\rxvt.exe -bg wheat -fg black -sl 5000 -e /bin/bash -ls

now I try inside to run some *non* cygwin console program
rxvt shows nothing, and reports 30% cpu consumption

The actual program is executed, completes and control comes back to
bash, but the output is not there.
I can issue blindly other commands and see them in process list (e.g. ls -lR /)
I can also terminate the session with Ctrl-D or exit
As long as rxvt is running procexplorer reports 30% cpu usage.

This started to happen today after I did update (setup.exe)
The cygwin.dll version is 1.7.15-1
Now if I roll back cygwin.dll to 1.7.14-2 problem is gone.

The programs which triggers the problem are .net console applications
(2.0 or 4.0).
All cygwin binaries + java.exe work ok.

The trigger program runs ok in cmd.exe with bash.
And as a bonus, mintty has also similar problem. Cpu doesn't go as
high and input channel with bash is lost.



--pawel

--
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

RE: rxvt loses output connection with non cygwin console processes

[James Johnston]

I can reproduce this, although I did not try rolling back my Cygwin version.

The issue seems specific to .NET programs because they always do null
writes.  However the following straight Win32 program written in C will also
reproduce the problem (basically the same program I've sent earlier on this
mailing list).  So it is not specific to .NET.

#include <windows.h>
int main() {
	char * test = "Hello world!\n";
	DWORD written;
	HANDLE h = GetStdHandle(STD_OUTPUT_HANDLE);
	Sleep(1000);
	WriteFile(h, test, 0, &written, NULL); /* causes problems */
	Sleep(1000);
	WriteFile(h, test, lstrlenA(test), &written, NULL);
	return 0;
}

As the OP notes, I also only have this problem when running the test program
from a "GUI" terminal.  It seems that any program doing a null write to
standard output (and probably standard error?) will trigger the problem.

From Windows command prompt / standard Windows console window:  Running
Cygwin.bat to start Cygwin and then running ./Test.exe works fine.

From mintty.exe:  Running ./Test.exe causes bash to hang.  Test.exe runs to
completion and terminates, but bash doesn't get the hint.  CPU usage is at
0%.  If I use Windows Task Manager to kill bash.exe, then mintty.exe
correctly terminates. 

From rxvt.exe:  Running ./Test.exe causes bash to hang like with mintty.
Test.exe runs to completion and terminates but bash is hung.  For some
reason, a thread in rxvt spins in an infinite loop, sucking 100% of a single
CPU core.  The hung bash.exe doesn't use any CPU.  If I use Windows Task
Manager to kill bash.exe, then rxvt.exe correctly exits its loop and
terminates.

> i have discovered something peculiar.
> I run my rxvt with the usual:
> 
> C:\cygwin\bin\rxvt.exe -bg wheat -fg black -sl 5000 -e /bin/bash -ls
> 
> now I try inside to run some *non* cygwin console program rxvt shows
> nothing, and reports 30% cpu consumption
> 
> The actual program is executed, completes and control comes back to bash,
> but the output is not there.
> I can issue blindly other commands and see them in process list (e.g. ls
-lR /) I
> can also terminate the session with Ctrl-D or exit As long as rxvt is
running
> procexplorer reports 30% cpu usage.
> 
> This started to happen today after I did update (setup.exe) The cygwin.dll
> version is 1.7.15-1 Now if I roll back cygwin.dll to 1.7.14-2 problem is
gone.
> 
> The programs which triggers the problem are .net console applications
> (2.0 or 4.0).
> All cygwin binaries + java.exe work ok.
> 
> The trigger program runs ok in cmd.exe with bash.
> And as a bonus, mintty has also similar problem. Cpu doesn't go as high
and
> input channel with bash is lost.


--
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

RE: rxvt loses output connection with non cygwin console processes

[James Johnston]

I apologize, I forgot to mention that I tried both with and without the new
"pipe_byte" CYGWIN option.  It made no difference.

To OP: you'll definitely want to look into using the new "pipe_byte" option
since you are working with non-Cygwin processes.  See the thread I started
late last month titled "Cygwin passes through null writes to other software
when redirecting standard input/output".  If you redirect standard input for
a non-Cygwin program, you'll be affected.

> As the OP notes, I also only have this problem when running the test
> program from a "GUI" terminal.  It seems that any program doing a null
write
> to standard output (and probably standard error?) will trigger the
problem.
> 
> From Windows command prompt / standard Windows console window:
> Running Cygwin.bat to start Cygwin and then running ./Test.exe works fine.
> 
> From mintty.exe:  Running ./Test.exe causes bash to hang.  Test.exe runs
to
> completion and terminates, but bash doesn't get the hint.  CPU usage is at
> 0%.  If I use Windows Task Manager to kill bash.exe, then mintty.exe
> correctly terminates.
> 
> From rxvt.exe:  Running ./Test.exe causes bash to hang like with mintty.
> Test.exe runs to completion and terminates but bash is hung.  For some
> reason, a thread in rxvt spins in an infinite loop, sucking 100% of a
single CPU
> core.  The hung bash.exe doesn't use any CPU.  If I use Windows Task
> Manager to kill bash.exe, then rxvt.exe correctly exits its loop and
> terminates. 


--
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

Re: rxvt loses output connection with non cygwin console processes

[Christopher Faylor]

On Tue, May 15, 2012 at 06:59:14PM +0200, Pawel Jasinski wrote:
>i have discovered something peculiar.
>I run my rxvt with the usual:
>
>C:\cygwin\bin\rxvt.exe -bg wheat -fg black -sl 5000 -e /bin/bash -ls
>
>now I try inside to run some *non* cygwin console program
>rxvt shows nothing, and reports 30% cpu consumption

This problem should be fixed in the next snapshot.

Ken Brown, if you're reading this, would you mind confirming that the
problem with emacs and gdb -mi that I recently fixed is has not been
resurrected?

(If you're not reading this then I'd still appreciate confirmation)

cgf

--
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

Re: rxvt loses output connection with non cygwin console processes

[Ken Brown]

On 5/15/2012 9:59 PM, Christopher Faylor wrote:
> On Tue, May 15, 2012 at 06:59:14PM +0200, Pawel Jasinski wrote:
>> i have discovered something peculiar.
>> I run my rxvt with the usual:
>>
>> C:\cygwin\bin\rxvt.exe -bg wheat -fg black -sl 5000 -e /bin/bash -ls
>>
>> now I try inside to run some *non* cygwin console program
>> rxvt shows nothing, and reports 30% cpu consumption
>
> This problem should be fixed in the next snapshot.
>
> Ken Brown, if you're reading this, would you mind confirming that the
> problem with emacs and gdb -mi that I recently fixed is has not been
> resurrected?

It's still fixed.

Ken


--
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

Re: rxvt loses output connection with non cygwin console processes

[Christopher Faylor]

On Wed, May 16, 2012 at 06:57:55AM -0400, Ken Brown wrote:
>On 5/15/2012 9:59 PM, Christopher Faylor wrote:
>> On Tue, May 15, 2012 at 06:59:14PM +0200, Pawel Jasinski wrote:
>>> i have discovered something peculiar.
>>> I run my rxvt with the usual:
>>>
>>> C:\cygwin\bin\rxvt.exe -bg wheat -fg black -sl 5000 -e /bin/bash -ls
>>>
>>> now I try inside to run some *non* cygwin console program
>>> rxvt shows nothing, and reports 30% cpu consumption
>>
>> This problem should be fixed in the next snapshot.
>>
>> Ken Brown, if you're reading this, would you mind confirming that the
>> problem with emacs and gdb -mi that I recently fixed has not been
>> resurrected?
>
>It's still fixed.

Thank you for confirming.

cgf

--
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

RE: rxvt loses output connection with non cygwin console processes

[James Johnston]

> On Tue, May 15, 2012 at 06:59:14PM +0200, Pawel Jasinski wrote:
> >i have discovered something peculiar.
> >I run my rxvt with the usual:
> >
> >C:\cygwin\bin\rxvt.exe -bg wheat -fg black -sl 5000 -e /bin/bash -ls
> >
> >now I try inside to run some *non* cygwin console program rxvt shows
> >nothing, and reports 30% cpu consumption
> 
> This problem should be fixed in the next snapshot.

Confirmed fixed with both my C test program and a .NET program too.


--
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