1658776 - EventSource's locking regime is inconsistent

Status: RESOLVED FIXED

(Core :: DOM: Networking, defect, P2)

Description

[Yaron Tausky [:ytausky]]

EventSource's implementation has a mutex with a very wrong comment attached to it. Several other members of both EventSource and EventSourceImpl are potentially accessed from two different threads and usage of the mutex is inconsistent.

Comment 1

[Yaron Tausky [:ytausky]]

Attachment: Bug 1658776 - Remove some potential data races r=asuth,#dom-workers-and-storage-reviewers

Some code in EventSource's implementation was accessing data
members of the DOM object's base class DOMEventTargetHelper from
the main thread, event though that class is not thread-safe. This
commit makes those accesses happen only if the DOM object was
created on the main thread.

Note that there are still other data races in EventSource's
implementation; those will be fixed in subsequent patches.

Comment 2

[Yaron Tausky [:ytausky]]

Attachment: Bug 1658776 - Make some data members atomic or const r=asuth,#dom-workers-and-storage-reviewers

With this commit a few of EventSource's and EventSourceImpl's data
members are now atomic, since a mutex isn't really necessary for
their use case. Also, several data members are now marked const.

Depends on D101163

Comment 3

[Phabricator Automation]

Comment on attachment 9196069 [details]
Bug 1658776 - Remove some potential data races r=asuth,#dom-workers-and-storage-reviewers

Revision D101163 was moved to bug 1682928. Setting attachment 9196069 [details] to obsolete.

Comment 4

[Phabricator Automation]

Comment on attachment 9196135 [details]
Bug 1658776 - Make some data members atomic or const r=asuth,#dom-workers-and-storage-reviewers

Revision D101210 was moved to bug 1682928. Setting attachment 9196135 [details] to obsolete.

Comment 5

[Jens Stutte [:jstutte]]

Hi Randell, would you want to give this a look? Thanks!

Comment 6

[Randell Jesup [:jesup] (needinfo me)]

patches were landed on another bug