2077257 - Improve ThreadSafeWeakPtrHashKey behaviour during an object's destructor

Status: RESOLVED FIXED

(Core :: XPCOM, enhancement)

Description

[Nika Layzell [:nika] (ni? for response)]

Currently ThreadSafeWeakPtrHashKey is defined fully in terms of the concrete referenced type, meaning that it cannot work late during an object's lifecycle, after the object's refcount has reached 0 and during the destructor. I expect that removing entries from a table during the destructor is a clear use-case for this type, meaning that it would be better if we supported this.

This patch does this by simplifying the handling of ThreadSafeWeakPtr such that it's used directly, and slightly improving the ergonomics of the underlying type so that it can be used directly in this way.

Comment 1

[Nika Layzell [:nika] (ni? for response)]

Attachment: Bug 2077257 - Improve ThreadSafeWeakPtrHashKey behaviour during an object's destructor, r=#xpcom-reviewers!

Currently ThreadSafeWeakPtrHashKey is defined fully in terms of the
concrete referenced type, meaning that it cannot work late during an
object's lifecycle, after the object's refcount has reached 0 and during
the destructor. I expect that removing entries from a table during the
destructor is a clear use-case for this type, meaning that it would be
better if we supported this.

This patch does this by simplifying the handling of ThreadSafeWeakPtr
such that it's used directly, and slightly improving the ergonomics of
the underlying type so that it can be used directly in this way.

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/b7b9f961af7b
https://hg.mozilla.org/integration/autoland/rev/e8a242ab68d5
Improve ThreadSafeWeakPtrHashKey behaviour during an object's destructor, r=xpcom-reviewers,dom-storage-reviewers,emilio,janv

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/d136f271c6d4
https://hg.mozilla.org/integration/autoland/rev/3f1f2902caa3
Revert "Bug 2077257 - Improve ThreadSafeWeakPtrHashKey behaviour during an object's destructor, r=xpcom-reviewers,dom-storage-reviewers,emilio,janv" for causing android build bustages

Comment 4

[Sandor Molnar[:smolnar]]

Backed out for causing android build bustages

Backout link: https://hg.mozilla.org/integration/autoland/rev/3f1f2902caa3bd66cbf1ead79899a0af5ac9dcfc

Push with failures

Failure log

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/41c39394fafb
https://hg.mozilla.org/integration/autoland/rev/da9927b0fec6
Improve ThreadSafeWeakPtrHashKey behaviour during an object's destructor, r=xpcom-reviewers,dom-storage-reviewers,emilio,janv

Comment 6

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/da9927b0fec6

Comment 8

[Nika Layzell [:nika] (ni? for response)]

Attachment: Bug 2077257 - Improve ThreadSafeWeakPtrHashKey behaviour during an object's destructor,

Currently ThreadSafeWeakPtrHashKey is defined fully in terms of the
concrete referenced type, meaning that it cannot work late during an
object's lifecycle, after the object's refcount has reached 0 and during
the destructor. I expect that removing entries from a table during the
destructor is a clear use-case for this type, meaning that it would be
better if we supported this.

This patch does this by simplifying the handling of ThreadSafeWeakPtr
such that it's used directly, and slightly improving the ergonomics of
the underlying type so that it can be used directly in this way.

Original Revision: https://phabricator.services.mozilla.com/D331272

Comment 9

[Pulsebot]

https://hg.mozilla.org/releases/mozilla-esr153/rev/e943d271231b

Comment 10

[Pulsebot]

https://hg.mozilla.org/releases/mozilla-esr153/rev/74b64e9b121f