At Sun, 12 Aug 2012 21:50:29 -0600, Michael Wilber wrote: > Will such custodian shutdown callbacks be run when racket receives a > SIGTERM?
I think tat would make sense, but Racket doesn't currently try to handle SIGTERM specially, so it defaults to abort()-like behavior. Probably Racket should catch SIGTERM and SIGHUP and trigger calls to new handlers whose default actions call the exit handler. I'll try making that change. > On Sun, 12 Aug 2012 20:36:36 -0600, Matthew Flatt <mfl...@cs.utah.edu> wrote: > > At Sun, 5 Aug 2012 15:02:10 -0600, Matthew Flatt wrote: > > > Beware that FFI callback memory management is slightly tricky with > > > scheme_add_managed(): > > > > I've added an `ffi/unsafe/custodian' library with a > > `register-custodian-shutdown' function to make this easier. > > > > The `register-custodian-shutdown' function also helps with a related > > problem: calling a shutdown callback when Racket exits. Most resources > > are released when the Racket process exits, but if an explicit release > > is needed, supply `#:at-exit? #t'. > > > > Of course, while `register-custodian-shutdown' makes custodian > > registration easier, there are still many low-level issues to consider > > --- such as the fact that the callback is invoked atomically in an > > unspecified Racket thread --- which is why it's under `ffi/unsafe'. > > > > ____________________ > > Racket Users list: > > http://lists.racket-lang.org/users > > ____________________ Racket Users list: http://lists.racket-lang.org/users
