Re: jobserver fds remain open in subprocess

[Keith Rarick]

Thanks for the reply! I just tried with make 3.81 on two different
machines, and saw the error on only one.

Good uname -a:
Linux 3054d0d3-73b7-4f7d-bd02-dc3ab70449dc 3.8.11-ec2 #1 SMP Fri May 3
09:11:15 UTC 2013 x86_64 GNU/Linux

Bad uname -a:
Darwin plum.local 12.4.0 Darwin Kernel Version 12.4.0: Wed May  1
17:57:12 PDT 2013; root:xnu-2050.24.15~1/RELEASE_X86_64 x86_64


Is there anything else I can try or do to help figure this out?
It's not really affecting me negatively, I'm mostly just curious
at this point.



Here's a typescript of the failure:

plum:tmp kr$ mkdir t
plum:tmp kr$ cd t
plum:t kr$ curl -s -O
https://gist.github.com/kr/b4852e149f45a43fb17a/raw/f4dc5316cadcb51875a52cb2967f986f2f4031f5/x.c
plum:t kr$ printf 'foo:\n\t./x\n' >Makefile
plum:t kr$ make x
cc     x.c   -o x
plum:t kr$ make -j4 foo
./x
rjob is 3
wjob is 4
read make job token 43 +
read make job token 43 +
read make job token 43 +
number of jobs is 4
make: *** [foo] Error 1

plum:t kr$ uname -a
Darwin plum.local 12.4.0 Darwin Kernel Version 12.4.0: Wed May  1
17:57:12 PDT 2013; root:xnu-2050.24.15~1/RELEASE_X86_64 x86_64
plum:t kr$ make -v
GNU Make 3.81
Copyright (C) 2006  Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.

This program built for i386-apple-darwin11.3.0
plum:t kr$ gcc --version
i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 (Based on Apple Inc.
build 5658) (LLVM build 2336.11.00)
Copyright (C) 2007 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.





On Sun, Jul 21, 2013 at 7:56 PM, Paul Smith <address@hidden> wrote:
> On Wed, 2012-12-12 at 17:11 -0800, Keith Rarick wrote:
>> I'm seeing some behavior that surprises me:
>> https://gist.github.com/b4852e149f45a43fb17a
>>
>> I'd expect that putting a + in front of the command
>> would cause make to leave the jobserver fds open,
>> and leaving off the + would cause make to close
>> those fds. This is the behavior described in
>> http://mad-scientist.net/make/jobserver.html and
>> it matches my reading of the make source code
>> (retrieved from http://ftp.gnu.org/gnu/make/make-3.82.tar.gz).
>>
>> Can anyone help me understand what's going on?
>
> I've tried to reproduce your example with 3.81, 3.82, and with the
> latest Git version and in all cases make works as expected (the program
> reports that number of jobs is 1).
>
> If you still see this, please provide more details such as the operating
> system you're using, etc.
>