[webkit-reviews] review canceled: [Bug 231681] REGRESSION (r284079): fast/canvas/gradient-with-clip.html and fast/canvas/gradient-text-with-shadow.html are flaky failures : [Attachment 441553] Initial approach

[bugzilla-daemon at webkit.org]

Wenson Hsieh <wenson_hsieh at apple.com> has canceled Wenson Hsieh
<wenson_hsieh at apple.com>'s request for review:
Bug 231681: REGRESSION (r284079): fast/canvas/gradient-with-clip.html and
fast/canvas/gradient-text-with-shadow.html are flaky failures
https://bugs.webkit.org/show_bug.cgi?id=231681

Attachment 441553: Initial approach

https://bugs.webkit.org/attachment.cgi?id=441553&action=review




--- Comment #20 from Wenson Hsieh <wenson_hsieh at apple.com> ---
Comment on attachment 441553
  --> https://bugs.webkit.org/attachment.cgi?id=441553
Initial approach

Kimmo pointed out that here is still the possibility of reordered messages here
in the following scenario:

1. Receive stream IPC message (A) for an unknown receiver
2. Receive normal IPC message (B)
3. Receive stream IPC message (C) for an unknown receiver

It's possible that (2) causes us to try and dispatch an incoming message, which
(in this patch) will take (A) and append it to the end. Later, when the receive
queue is actually added, we'll end up moving messages (C) and (A) into the new
receive queue, in that order.

I think we could adjust the patch to be robust against this possibility by
adding receiver-less stream messages to a separate queue that is not
`m_incomingMessages`, and taking messages from that separate queue underneath
`enqueueMatchingMessagesToMessageReceiveQueue`.