On Thu, May 31, 2001 at 12:06:15PM -0400, Uri Guttman wrote:
> DN> Watchers are stopped implicitly if their new
> DN> configuration deprives them of the ability to generate
> DN> events.
>
> events are not callbacks. theoretically you could have an event with an
> implied action and no callback. the bug is as i said, it should not call
> the null callback. cancelling the watcher because of no callback makes
> no sense. you just cancel the watcher.
I have a watcher with a callback of [ $obj, "method" ]. $obj contains
a reference to the watcher, creating a circular reference chain. I
want $obj to be GCd, so I need to either remove the watcher's reference
to $obj, or vice-versa.
Destroying the watcher is insufficient, as this does not free the
contents of its cb attribute. I need to explicitly undefine the
callback.
The above quote from the manpage indicates, by my reading, that
removing the callback implicitly stops the watcher. To support this
reading, I quote from elsewhere in the manpage:
$watcher->start
Activate the watcher. Watchers refuse to start()
without sufficient configuration information to
generate events. Constructors always invoke start()
unless the parked=1> option is requested. You will
need to set the parked option if you preallocate
unconfigured watchers.
If I create a watcher with no cb attribute and without the parked
option, I receive an exception: "Event: can't start '?? boom:11'
without callback". I interpret this as occurring because the watcher
does not have sufficient configuration information to generate
events: events must have an associated callback.
The alternative is to say that it is acceptable for an io watcher
to spin, generating a continuous series of callbackless events,
if it is configured without a callback.
> and it is not a perl bug in any case. it is trivial to get perl to core
> dump from an XS module.
I did not say that it is a bug in perl.
I am currently feeling quite frustrated. I spent some time tracking
down the bug (not a perl bug; a bug in Event) which was causing my
program to core dump. I found it, and submitted a description and
patch to the appropriate people. In return, I am being told that
I am the one at fault.
- Damien
