how to make sure subprocesses die?

how to make sure subprocesses die?

[Noel Grandin]

Hi

When running GNU make from the cygwin bash prompt and killing it with CTRL-C, the children don't seem to be dying properly.
I have to hit CTRL-C have a dozen times to kill everything off.

Could anyone give me some clues about where to look to solve this?

I'm actually hoping that fixing this simpler case will solve the similar problem we have with runaway jobs not dying
properly on our LibreOffice tinderboxes.

Thanks,
     Noel Grandin


--
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: how to make sure subprocesses die?

[Warren Young]

On Aug 18, 2015, at 8:11 AM, Noel Grandin <noelgrandin@gmail.com> wrote:
> 
> When running GNU make from the cygwin bash prompt and killing it with CTRL-C, the children don't seem to be dying properly.

  $ svn checkout svn://svn.code.sf.net/p/ctags/code/trunk ctags
  $ cd ctags
  $ autoconf && autoheader && ./configure
  $ make -j3
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c args.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c ant.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c asm.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c asp.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c awk.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c basic.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c beta.c
  gcc -I. -I. -DHAVE_CONFIG_H -g -O2 -c c.c
  <Ctrl-C>
  make: *** Deleting file 'basic.o'
  Makefile:220: recipe for target 'beta.o' failed
  make: *** [beta.o] Interrupt
  Makefile:220: recipe for target 'basic.o' failed
  make: *** [basic.o] Interrupt
  $

Works for me.

Disagree?  Provide a replicable, simple test case to show it.
--
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: how to make sure subprocesses die?

[Noel Grandin]

On 2015-08-18 04:31 PM, Warren Young wrote:
> Works for me.
>
> Disagree?  Provide a replicable, simple test case to show it.
> --

Ah, sorry, I didn't mean to imply I thought it was a bug, I'm sure it's something we're doing, I was just wondering if 
anyone had any hints about debugging.

This is the main LibreOffice build, and it's a multi-million line codebase, so shrinking the problem is my first 
priority :-)

--
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: how to make sure subprocesses die?

[Warren Young]

On Aug 19, 2015, at 3:09 AM, Noel Grandin <noelgrandin@gmail.com> wrote:
> 
> On 2015-08-18 04:31 PM, Warren Young wrote:
>> Works for me.
>> 
>> Disagree?  Provide a replicable, simple test case to show it.
> 
> This is the main LibreOffice build, and it's a multi-million line codebase, so shrinking the problem is my first priority :-)

So on the one hand, you have one of the simplest real-world build systems (Exuberant Ctags’) and it works fine, and you have one of the most complicated (LO’s) and it doesn’t work as expected.

Somewhere in that vast gap is your answer.

Instead of trying to shrink LO, what happens if you build LO to completion, then touch a common header file that causes another dozen or so files to need rebuilding?  Does it recur?  What happens if that header file spans multiple subsystems within LO, vs being confined in its effects to just one subsystem?

What I’m getting at is, maybe your problem is in the way one make(1) instance spawns another, something that doesn’t happen in non-nested build systems, as with ctags.  If it all works fine when you’re only rebuilding one subsystem with a single make(1) instance, but fails with nested make(1) calls...
--
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