diff options
author | Christopher Vollick <0@psycoti.ca> | 2023-09-14 15:12:58 -0400 |
---|---|---|
committer | Marvin W <git@larma.de> | 2024-04-28 21:36:27 +0200 |
commit | d473efcbfe3603fac62594bd9d8fee8b23ffcbe5 (patch) | |
tree | ec7d09a8ac92b3ab434b383ab02215aa2632055b /xmpp-vala/src/module/tls.vala | |
parent | ba83a4ba3d6886b711278b8804566dcbe8ad7621 (diff) | |
download | dino-d473efcbfe3603fac62594bd9d8fee8b23ffcbe5.tar.gz dino-d473efcbfe3603fac62594bd9d8fee8b23ffcbe5.zip |
Add WeakTimeout Pattern to Prevent Leaks
While doing testing I noticed that skeletons were being leaked, and
eventually tracked it down to the timer that updates the time label
closing over "this" and then keeping the reference alive, potentially
for 24 hours.
I noticed a few other places in the code doing some version of this, and
one of them had the "static and weak pointer" approach, which I pulled
out into a util class. Now, we still have to make sure we're passing it
a static method instead of a lambda, as that would also close over
"this" and render the whole thing useless, but at least most of the
annoying parts live in the util class now.
Also the call_widget version was doing a weird thing where it was
removing itself, but then returning "true"? I'm not sure what that
accomplishes, because returning "false" means to not run this again. So
I think my new version is the same in practice, but simpler...
There are other timeouts in the code that I briefly looked over, but all
of them seemed to be relatively short hard-coded durations, so I left
them alone.
But if any of them are long-lived, it's possible they could also benefit
from this class in the future.
Closes #1480
Co-Authored-By: Marvin W <git@larma.de>
Diffstat (limited to 'xmpp-vala/src/module/tls.vala')
0 files changed, 0 insertions, 0 deletions