I like the idea of an optional "sanity check" flag. I can imagine that
if you had lots of connections open, scanning the connections for
existing result structures could be relatively costly. Its useful when
debugging: much nicer that random corruption, but I'm afraid it isn't
a "cheap" check that should always be enabled?
On Dec 13, 2009, at 12:22 , Eric Day wrote:
As far as assertions, I try to avoid them and instead report an error
in a sane way if possible.
Fair enough. Unfortunately, one way to "catch" this error happens
pretty late: when trying to free result structures, you can look for
the single circular list item that was causing a loop (->next == -
>prev == current). However, at this point some corruption may have
already occurred, and it might be confusing to debug this error. Thus,
I'm not sure if adding a check for this is particularly useful.
Thanks,
Evan
--
Evan Jones
http://evanjones.ca/
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
