This one has the patch given in my previous note; as a consequence,
it takes restarts gracefully, rather than going into a useless loop in
which it madly forks children which immediately die (when compiled with
the locking code enabled).
rst
PS --- so there's no confusion about why I chose the fix I chose to
this problem, tying the lock-token file to the pool machinery really
is a good idea --- it ensures that the file will be closed before
exec-ing untrusted child processes (CGI, etc.), which means that they
can't drag the server to a halt by locking it themselves. You just
have to be aware that when the pool gets cleared, the thing is gone...
it takes restarts gracefully, rather than going into a useless loop in
which it madly forks children which immediately die (when compiled with
the locking code enabled).
rst
PS --- so there's no confusion about why I chose the fix I chose to
this problem, tying the lock-token file to the pool machinery really
is a good idea --- it ensures that the file will be closed before
exec-ing untrusted child processes (CGI, etc.), which means that they
can't drag the server to a halt by locking it themselves. You just
have to be aware that when the pool gets cleared, the thing is gone...