Starting a subprocess in stopped state

Starting a subprocess in stopped state

[Eli Zaretskii]

My reading of process.c seems to indicate that the :stop attribute of
make-process only has effect on network or serial or pipe process
types; a process running a program cannot use that attribute, and can
only stopped by explicitly calling stop-process.  Is this correct, or
did I miss something?  This is not explicitly documented.

What I see in the code is that when make-process is called with the
:stop attribute non-nil, the file descriptor to be used for reading
the process output is not added to the list of descriptors watched by
pselect.  But that doesn't really suspend the process like SIGTSTP
would, right?  And I see no other code that specifically handles the
:stop attribute.  Am I missing something?

TIA

Re: Starting a subprocess in stopped state

[Eli Zaretskii]

> Date: Tue, 02 May 2017 22:04:27 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> 
> My reading of process.c seems to indicate that the :stop attribute of
> make-process only has effect on network or serial or pipe process
> types; a process running a program cannot use that attribute, and can
> only be stopped by explicitly calling stop-process.  Is this correct,
> or did I miss something?  This is not explicitly documented.
> 
> What I see in the code is that when make-process is called with the
> :stop attribute non-nil, the file descriptor to be used for reading
> the process output is not added to the list of descriptors watched by
> pselect.  But that doesn't really suspend the process like SIGTSTP
> would, right?  And I see no other code that specifically handles the
> :stop attribute.  Am I missing something?

Ping!  Can someone please confirm or refute my observations above?
Daiki?

Re: Starting a subprocess in stopped state

[Thien-Thi Nguyen]

[-- Attachment #1: Type: text/plain, Size: 1613 bytes --]


() Eli Zaretskii <eliz@gnu.org>
() Fri, 05 May 2017 10:29:52 +0300

   > What I see in the code is that when make-process is called
   > with the :stop attribute non-nil, the file descriptor to be
   > used for reading the process output is not added to the
   > list of descriptors watched by pselect.  But that doesn't
   > really suspend the process like SIGTSTP would, right?  And
   > I see no other code that specifically handles the :stop
   > attribute.  Am I missing something?

   confirm or refute

I see the docstring for ‘make-process’ says:

 :stop BOOL -- Start process in the ‘stopped’ state if BOOL
 non-nil.  In the stopped state, a process does not accept
 incoming data, but you can send outgoing data.  The stopped
 state is cleared by ‘continue-process’ and set by
 ‘stop-process’.

My understanding is that "stopped state", given that "you can
send outgoing data", is an Emacs-internal data routing concern
rather than an OS-level concern (SIGTSTP).  The latter would
completely prevent the asymmetric data feature (documented by
"you can send outgoing data"), i'd think.

So, omission from ‘pselect’ fd list is a correct implementation.
If anyone can refute this refutation, please do -- i'm all ears
(in a "stopped state", myself :-D)!

-- 
Thien-Thi Nguyen -----------------------------------------------
 (defun responsep (query)
   (pcase (context query)
     (`(technical ,ml) (correctp ml))
     ...))                              748E A0E8 1CB8 A748 9BFA
--------------------------------------- 6CE4 6703 2224 4C80 7502


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: Starting a subprocess in stopped state

[Eli Zaretskii]

> From: Thien-Thi Nguyen <ttn@gnu.org>
> Date: Sat, 06 May 2017 07:42:18 +0200
> 
> I see the docstring for ‘make-process’ says:
> 
>  :stop BOOL -- Start process in the ‘stopped’ state if BOOL
>  non-nil.  In the stopped state, a process does not accept
>  incoming data, but you can send outgoing data.  The stopped
>  state is cleared by ‘continue-process’ and set by
>  ‘stop-process’.
> 
> My understanding is that "stopped state", given that "you can
> send outgoing data", is an Emacs-internal data routing concern
> rather than an OS-level concern (SIGTSTP).  The latter would
> completely prevent the asymmetric data feature (documented by
> "you can send outgoing data"), i'd think.

However, stop-process does send SIGTSTP to the process, and
continue-process sends SIGCONT.  The above means that the process
started with :stop non-nil will indeed run, unlike a process which was
stopped with stop-process immediately after starting it.  So there are
two flavors of "stopped" state here, one with SIGTSTP delivered to the
process, the other without.

Also, not reading the Emacs end of the pipe, while the process runs is
probably going to get is to undefined behavior in some situations,
especially with multithreaded processes.

For network or serial or pipe "process", not reading from the
descriptor is all that counts.  But for a local subprocess, this is
not so, IMO.

Is there a way to create a process on Unix and GNU systems, but
prevent it from running until it gets SIGCONT or something similar?
(There is such a way on MS-Windows.)  Debuggers do that, of course,
but I don't mean starting a process as a debugger would, since that
would be gross, and probably requires special privileges at least on
some systems.

Thanks.

Re: Starting a subprocess in stopped state

[Philipp Stephani]

[-- Attachment #1: Type: text/plain, Size: 2045 bytes --]

Eli Zaretskii <eliz@gnu.org> schrieb am Sa., 6. Mai 2017 um 09:59 Uhr:

> > From: Thien-Thi Nguyen <ttn@gnu.org>
> > Date: Sat, 06 May 2017 07:42:18 +0200
> >
> > I see the docstring for ‘make-process’ says:
> >
> >  :stop BOOL -- Start process in the ‘stopped’ state if BOOL
> >  non-nil.  In the stopped state, a process does not accept
> >  incoming data, but you can send outgoing data.  The stopped
> >  state is cleared by ‘continue-process’ and set by
> >  ‘stop-process’.
> >
> > My understanding is that "stopped state", given that "you can
> > send outgoing data", is an Emacs-internal data routing concern
> > rather than an OS-level concern (SIGTSTP).  The latter would
> > completely prevent the asymmetric data feature (documented by
> > "you can send outgoing data"), i'd think.
>
> However, stop-process does send SIGTSTP to the process, and
> continue-process sends SIGCONT.  The above means that the process
> started with :stop non-nil will indeed run, unlike a process which was
> stopped with stop-process immediately after starting it.  So there are
> two flavors of "stopped" state here, one with SIGTSTP delivered to the
> process, the other without.
>
> Also, not reading the Emacs end of the pipe, while the process runs is
> probably going to get is to undefined behavior in some situations,
> especially with multithreaded processes.
>
> For network or serial or pipe "process", not reading from the
> descriptor is all that counts.  But for a local subprocess, this is
> not so, IMO.
>
> Is there a way to create a process on Unix and GNU systems, but
> prevent it from running until it gets SIGCONT or something similar?


It's at least theoretically possible on macOS using posix_spawn(3) and
POSIX_SPAWN_START_SUSPENDED (
https://developer.apple.com/legacy/library/documentation/Darwin/Reference/ManPages/man3/posix_spawnattr_setflags.3.html).
I don't think it's possible using fork/exec (but then, switching to
posix_spawn might be a good idea anyway).

[-- Attachment #2: Type: text/html, Size: 2681 bytes --]

Re: Starting a subprocess in stopped state

[Eli Zaretskii]

> From: Philipp Stephani <p.stephani2@gmail.com>
> Date: Sun, 07 May 2017 12:24:33 +0000
> Cc: Daiki Ueno <ueno@gnu.org>
> 
>  Is there a way to create a process on Unix and GNU systems, but
>  prevent it from running until it gets SIGCONT or something similar?
> 
> It's at least theoretically possible on macOS using posix_spawn(3) and
> POSIX_SPAWN_START_SUSPENDED
> (https://developer.apple.com/legacy/library/documentation/Darwin/Reference/ManPages/man3/posix_spawnattr_setflags.3.html).
> I don't think it's possible using fork/exec (but then, switching to posix_spawn might be a good idea anyway). 

Thanks.  However, since this an Apple extension, I don't think we can
use it unless there's an equivalent feature on other Unix and GNU
systems.

Re: Starting a subprocess in stopped state

[Paul Eggert]

Eli Zaretskii wrote:
> Is there a way to create a process on Unix and GNU systems, but
> prevent it from running until it gets SIGCONT or something similar?

Sure, the parent can fork and the child can send itself a SIGSTOP before doing 
the exec.

Re: Starting a subprocess in stopped state

[Eli Zaretskii]

> Cc: Daiki Ueno <ueno@gnu.org>
> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Sun, 7 May 2017 10:25:03 -0700
> 
> Eli Zaretskii wrote:
> > Is there a way to create a process on Unix and GNU systems, but
> > prevent it from running until it gets SIGCONT or something similar?
> 
> Sure, the parent can fork and the child can send itself a SIGSTOP before doing 
> the exec.

Thanks.  I think we should do this when make-process was called with
':stop' non-nil.

Re: Starting a subprocess in stopped state

[Andreas Schwab]

On Mai 07 2017, Paul Eggert <eggert@cs.ucla.edu> wrote:

> Eli Zaretskii wrote:
>> Is there a way to create a process on Unix and GNU systems, but
>> prevent it from running until it gets SIGCONT or something similar?
>
> Sure, the parent can fork and the child can send itself a SIGSTOP before
> doing the exec.

That will also delay all error reporting, though.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Starting a subprocess in stopped state

[Davis Herring]

> Also, not reading the Emacs end of the pipe, while the process runs is
> probably going to get is to undefined behavior in some situations,
> especially with multithreaded processes.

I don't think it's undefined behavior -- you may just hang (one or more 
threads of) the subprocess.  Of course, then it is stopped, much as if 
by SIGSTOP.

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or 
too sparse, it is because mass-energy conversion has occurred during 
shipping.

Re: Starting a subprocess in stopped state

[Eli Zaretskii]

> From: Davis Herring <herring@lanl.gov>
> Date: Mon, 8 May 2017 10:41:28 -0600
> Cc: Daiki Ueno <ueno@gnu.org>
> 
> > Also, not reading the Emacs end of the pipe, while the process runs is
> > probably going to get is to undefined behavior in some situations,
> > especially with multithreaded processes.
> 
> I don't think it's undefined behavior -- you may just hang (one or more 
> threads of) the subprocess.

Well, that's what I meant by "undefined" -- the exact result depends
on factors beyond our control, some of them depend on the program
being run, some on the OS and its configuration (like the size of the
pipe buffer).