Re: [Savannah-hackers-public] git accumulations on vcs

[James Cloos]

>>>>> "KB" == Karl Berry <address@hidden> writes:

KB> I would speculate that it's the net,

KB> Process 9935 attached - interrupt to quit
KB> read(0,  <unfinished ...>

KB> They are all invoked ... from inetd

That confirms your hypothesis, since inetd(8) provides networking via
stdin and stdout.  The process is waiting on input from the other side
of the socket.

Is 9935's parent inetd process still running; what does lsof(8) say
about said parent process?  Ie, are the tcp socket(s) still ESTABLISHED?

If not, the bug likely is in inetd.

If the inetd process is gone (which likely would mean that the git
process' PPID is now 1) then the timeout is killing the inetd child
w/o also killing that PID's children.

(Above I presume that sv's version of inetd uses a child process for
each invocation, to manage the network socket <-> file descriptor
mapping for the invoked process.  If the git's have a PPID which
matches the master inetd process, things obviously would work quite
a bit differently.)

-JimC
-- 
James Cloos <address@hidden>         OpenPGP: 1024D/ED7DAEA6