epa problem has returned

epa problem has returned

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

The epa problem is not gone.  It was never gone.

I tested for it with a few messages, and they worked -- but they
were all short.  Now it appears that sufficiently short messages work
but longer ones cause the problem.

I tried the idea of loading the latest versions
of the preloaded Lisp files into the old dumped Emacs.
I loaded all except loaddefs.el, macroexp.el, leim-list.el
and the language/* files.

The problem does NOT occur when I do that.

So it appears the problem is in the C code.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use Ekiga or an ordinary phone call.

epa problem has returned

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

I just tried decryption again with a call to message in the loop
before accept-process-output.  I got this output

    Decrypting...
    1 (BEGIN_DECRYPTION)
    Decrypting......0%
    2 (BEGIN_DECRYPTION)
    3 (BEGIN_DECRYPTION)
    4 (BEGIN_DECRYPTION)
    5 (BEGIN_DECRYPTION)
    6 (BEGIN_DECRYPTION)
    7 (BEGIN_DECRYPTION)
    Decrypting......done

after which it hung.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use Ekiga or an ordinary phone call.

Re: epa problem has returned

[T.V Raman]

From what I've seen, this problem is likely caused by the failure to
recognize process exit correctly -- I reported that a few weeks ago
after seeing it with both gpg and mplayer.

from memory somewhere around when this happened(src/Changelog)

2014-05-03  Paul Eggert  <eggert@cs.ucla.edu>

	Handle systems without WCONTINUED consistently.  (Bug#15110, 17339)
	* process.c (handle_child_signal): Remove WCONTINUED ifdef,
	because WCONTINUED is always defined now.
	* syswait.h (WCONTINUED): Move here from ../lib-src/emacsclient.c.

-- 

Re: epa problem has returned

[Eli Zaretskii]

> Date: Tue, 20 May 2014 06:23:02 -0400
> From: Richard Stallman <rms@gnu.org>
> 
> I just tried decryption again with a call to message in the loop
> before accept-process-output.  I got this output
> 
>     Decrypting...
>     1 (BEGIN_DECRYPTION)
>     Decrypting......0%
>     2 (BEGIN_DECRYPTION)
>     3 (BEGIN_DECRYPTION)
>     4 (BEGIN_DECRYPTION)
>     5 (BEGIN_DECRYPTION)
>     6 (BEGIN_DECRYPTION)
>     7 (BEGIN_DECRYPTION)
>     Decrypting......done
> 
> after which it hung.

When it hangs, can you attach GDB and show where is it looping?  I
don't understand how come accept-process-output could hang when it is
called with a timeout of 1 sec.

Is the process object with which accept-process-output is called alive
at the time of the hang?

Re: epa problem has returned

[raman]

I believe what actually happens is that the process exits successfully
but Emacs is somehow failing to recognize that the process is done.

Re: epa problem has returned

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

    When it hangs, can you attach GDB and show where is it looping?  I
    don't understand how come accept-process-output could hang when it is
    called with a timeout of 1 sec.

I am running all the time under GDB.
I stopped it with a signal during that hang
and it was hung in `read', called from `emacs_read', called
from `read_process_output', called from `wait_running_process_output'.
I put a breakpoint on the line in `emacs_read' that calls read, 
and it gets called over and over.  `read' returns 0.

Another time it was in pselect called from xg_select called from
wait_running_process_output.

I tried stepping and saw that it went around the big loop
in wait_running_process_output (starting at line 4333)
and did not exit it.

I don't know where and how it is supposed to exit,
so I could not tell why it was wrong.

I can't recall, and can't find, the GDB command to direct
GDB output to a file.

    Is the process object with which accept-process-output is called alive
    at the time of the hang?

Yes.  The process status is `run'.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use Ekiga or an ordinary phone call.

Re: epa problem has returned

[Jorgen Schaefer]

On Sun, 25 May 2014 12:01:03 -0400
Richard Stallman <rms@gnu.org> wrote:

> I am running all the time under GDB.
> I stopped it with a signal during that hang
> and it was hung in `read', called from `emacs_read', called
> from `read_process_output', called from `wait_running_process_output'.
> I put a breakpoint on the line in `emacs_read' that calls read, 
> and it gets called over and over.  `read' returns 0.
>
> Another time it was in pselect called from xg_select called from
> wait_running_process_output.

This sounds like the bug I am experiencing:

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=17561

Regards,
Jorgen

Re: epa problem has returned

[Eli Zaretskii]

> Date: Sun, 25 May 2014 12:01:03 -0400
> From: Richard Stallman <rms@gnu.org>
> CC: emacs-devel@gnu.org
> 
> I stopped it with a signal during that hang
> and it was hung in `read', called from `emacs_read', called
> from `read_process_output', called from `wait_running_process_output'.
> I put a breakpoint on the line in `emacs_read' that calls read, 
> and it gets called over and over.  `read' returns 0.
> 
> Another time it was in pselect called from xg_select called from
> wait_running_process_output.
> 
> I tried stepping and saw that it went around the big loop
> in wait_running_process_output (starting at line 4333)
> and did not exit it.

Paul, could this be another manifestation of bug #17561?  The symptoms
sound similar to me.

> I can't recall, and can't find, the GDB command to direct
> GDB output to a file.

 (gdb) set logging on

This puts all output from GDB on a file named gdb.txt in the directory
where you run GDB.

Re: epa problem has returned

[Paul Eggert]

Eli Zaretskii wrote:
> Paul, could this be another manifestation of bug #17561?  The symptoms
> sound similar to me.

It could be, yes.

Re: epa problem has returned

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

     (gdb) set logging on

    This puts all output from GDB on a file named gdb.txt in the directory
    where you run GDB.

Would you like me to log the result of stepping around the big loop
in wait_running_process_output?


-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use Ekiga or an ordinary phone call.

Re: epa problem has returned

[Eli Zaretskii]

> Date: Tue, 27 May 2014 02:11:27 -0400
> From: Richard Stallman <rms@gnu.org>
> CC: eggert@cs.ucla.edu, emacs-devel@gnu.org
> 
>      (gdb) set logging on
> 
>     This puts all output from GDB on a file named gdb.txt in the directory
>     where you run GDB.
> 
> Would you like me to log the result of stepping around the big loop
> in wait_running_process_output?

Paul, could you please answer that?  Based on my understanding of bug
#17561, I'd recommend to wait until it's resolved, and see if that
also fixes Richard's problem.  But I think your opinion on this counts
more than mine.

Re: epa problem has returned

[Paul Eggert]

Richard Stallman wrote:
> Would you like me to log the result of stepping around the big loop
> in wait_running_process_output?

It'd probably be more useful to see whether it's a vfork problem; see 
Bug#17561.

Does the problem go away if you use fork instead of vfork?  You can do 
that by removing sysdep.o, callproc.o, and process.o and compiling with 
-Dvfork=fork.

Also, what platform are you running on?  What is the output of the 
following shell commands?

uname -a
gcc --version | head -1
grep 'model name' /proc/cpuinfo
cat /etc/issue