Comments on: A much faster popen() and system() implementation for Linux http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/ Enthusiasm never stops Wed, 01 Feb 2012 12:33:24 +0000 hourly 1 http://wordpress.com/ By: Ivan Zahariev http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-2095 Fri, 18 Mar 2011 09:43:00 +0000 http://blog.famzah.net/?p=179#comment-2095 This problem has now been fixed. Thank you.

]]> By: david coz http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-2084 Sun, 13 Mar 2011 09:10:02 +0000 http://blog.famzah.net/?p=179#comment-2084 Ok!
I just updated your ticked with a crash example when you compile everything in C++.

]]> By: Ivan Zahariev http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-2082 Sat, 12 Mar 2011 11:40:44 +0000 http://blog.famzah.net/?p=179#comment-2082 The exit() vs. _exit() was a good point. I’ve created an issue ticket about it:
http://code.google.com/p/popen-noshell/issues/detail?id=1

Thanks for your comments.

]]> By: david coz http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-2078 Wed, 09 Mar 2011 20:59:49 +0000 http://blog.famzah.net/?p=179#comment-2078 Nice articles!
I mentioned your article in this stackoverflow topic

Another drawback of fork():
Even if it doesn’t duplicate memory in the end thanks to COW, fork() still fails if there is not enough memory to duplicate the memory used by your parent process.
It can be quite annoying …

I looked at your code, and it looks good in general!
I’ll still stick with posix_spawn() for production code, but your code was inspiring!

You could try to write shorter lines (less than 80 chars is ideal :) ,
and IMHO you should replace exit() by _exit() in void _popen_noshell_child_process().
Otherwise you may run into troubles esp. with c++ code, see http://www.unixguide.net/unix/programming/1.1.3.shtml

]]> By: Ivan Zahariev http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-2060 Tue, 01 Mar 2011 07:22:00 +0000 http://blog.famzah.net/?p=179#comment-2060 I have looked into the posix_spawn() method, but it didn’t seem very well documented at that time, nor stable. In the previous user comment, Xavier Roche reports that the API is also a bit broken.

]]> By: Ryan Tomayko http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-2058 Sun, 27 Feb 2011 13:22:02 +0000 http://blog.famzah.net/?p=179#comment-2058 This is great stuff. I wonder, have you ever considered implementing a subset of posix_spawn() instead of, or in addition to, popen() and system()? You should be able to implement popen() and system() using the equivalent of only posix_spawn_file_actions_addclose(), posix_spawn_file_actions_addopen(), and posix_spawn_file_actions_adddup2(), and would have a significantly more versatile interface to boot.

]]> By: Xavier Roche http://blog.famzah.net/2009/11/20/a-much-faster-popen-and-system-implementation-for-linux/#comment-1925 Fri, 10 Dec 2010 08:31:57 +0000 http://blog.famzah.net/?p=179#comment-1925 Thanks for this interesting article. One additional interesting technic: the posix_spawn() (http://www.opengroup.org/onlinepubs/009695399/functions/posix_spawn.html) function is supposed to be a modern replacement to the old fork()/exec(). The performances, as far as I tested, are equivalent to the clone()/exec() one. [Unfortunately the API behind is a bit broken, and is not really allowing to close all opened process handles, which is just a shame.]

]]>