Joy. I just discovered the contents of /dev/fd on different systems are unpredictable.
AIX - doesn't have it Solaris - an ls of it returns all *possible* file descriptors Looks like the test should check if /proc/$$/fd exists, then fallback to /dev/fd ----- Original Message ----- From: [email protected] To: Tom Tanner (BLOOMBERG/ LONDON), [email protected] At: Mar 3 2014 17:10:12 On Mac I see the three usual handles plus a couple of other numeric entries in /dev/fd as directories; I don't understand those but will take a look. On Mon, Mar 3, 2014 at 11:57 AM, Tom Tanner (BLOOMBERG/ LONDON) <[email protected]> wrote: Just been hunting around and apparently it's /dev/fd (rather than /proc/xxx/fd) on MacOS, and also apparently /dev/fd will work equally well for linux (although presumably ls -l /dev/fd will actually produce the handles ls has passed to it) If someone who has MacOS could test that and see if it works and do a pull request. ----- Original Message ----- From: [email protected] To: Tom Tanner (BLOOMBERG/ LONDON), [email protected] At: Mar 3 2014 16:43:43 Hi, On OSX 10.7.5 I get this: ls /proc/$$/fd | wc -l ls: /proc/97956/fd: No such file or directory 0 So it appears that item 2 below is the culprit. *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*- Rob Managan email managan at llnl.gov LLNL phone: 925-423-0903 P.O. Box 808, L-095 FAX: 925-422-3389 Livermore, CA 94551-0808 On 3/3/14 1:19 AM, "Tom Tanner (BLOOMBERG/ LONDON)" <[email protected]> wrote: On the OSX one, it looks like you don't have SWIG and RANLIB installed and it's not recognising that it hasn't. I seem to remember having to install a lot of software on my linux (Ubuntu) box in order to get the tests to run clean. If that's the case, I'd imagine it's a bug really. The leaky-handles test is possibly an issue with OSX not behaving quite like other linuxes. In order to detect how many handles are open in a forked subshell, it runs ls /proc/$$/fd | wc -l and expects that to return 3 (stdin, stdout, stderr). If it doesn't, then either 1) python isn't closing files in a child process properly 2) OSX doesn't have a proc/<pid>/fd directory 3) OSX has other standard handles. 4) I've written the test wrong and it doesn't gracefully exit for non posix systems. I don't have access to an OSX system so I can't really tell, though if it (os.name) returns 'posix' that should work. Cheers TT _______________________________________________ Scons-dev mailing list [email protected] http://two.pairlist.net/mailman/listinfo/scons-dev -- Gary
_______________________________________________ Scons-dev mailing list [email protected] http://two.pairlist.net/mailman/listinfo/scons-dev
