We got a report that MemChaser leaks a lot of cuddlefish.js compartments. Those are still around even if the extension has been disabled or removed. It sounds similar to bug 725603 but we mainly require different modules. Also the patch on bug 724433 doesn't actually solve the problem for us. So I have disabled the require widgets line in main.js because of bug 724404, but that also made no difference.
That's why I file this as a new bug for now. Whenever an improvement lands we will make sure to test our extension again.
So I will make this more general because it's not only cuddlefish.js which is leaking. With bug 724433 the situation has already been improved massively. But there are still compartments lingering around when disabling MemChaser.
│ ├───2,407,696 B (03.59%) -- compartment([System Principal], resource://memchaser-at-quality-dot-mozilla-dot-org/api-utils/lib/environment.js, 0x1133a9000)
│ ├─────822,384 B (01.23%) -- compartment([System Principal], resource://memchaser-at-quality-dot-mozilla-dot-org/api-utils/lib/environment.js, 0x1080b9000)
│ ├─────192,304 B (00.24%) -- compartment([System Principal], jar:file:///var/folders/wd/zmy4z7xn7wd7sjq90z1y52f80000gn/T/tmp08VsmH.firstname.lastname@example.org!/bootstrap.js, 0x107dc7000)
Alex, shall I file separate bugs for those issues or can we take care of here?
No, we can provide specific patch here, if bugs 724433 and 724404 doesn't address all leaks.
Any compartment of MemChaser is now removed when disabling the extension. So everything is fine now. Thanks for those quick fixes!
nope, still occurs
memchaser-at-quality-dot-mozilla-dot-org/api-utils/lib/environment.js compartment still exists after disabling the extension.
Daniel, which version of the SDK are you using?
Do you confirm that all leaks are now gone in 1.6 ?
I just tested memchaser with master head against SDK master head.
And I clicked on all possible actions of memchaser, then disabled it and haven't seen any leftover compartments.
(btw, I now clearly understand your request about widget on panel position on pointer ;))
As I have mentioned earlier I'm not able to reproduce any of the leaks as mentioned in this bug anymore with the SDK 1.6. So not sure which version of the SDK Danial was using when testing it. Doesn't look like the 1.6 one.
Closing the bug again given that it works and we never got more information from Danial.