[webkit-reviews] review denied: [Bug 181560] The WebContent process should not process incoming IPC while waiting for a sync IPC reply : [Attachment 331242] Patch

[bugzilla-daemon at webkit.org]

Alexey Proskuryakov <ap at webkit.org> has denied Chris Dumez <cdumez at apple.com>'s
request for review:
Bug 181560: The WebContent process should not process incoming IPC while
waiting for a sync IPC reply
https://bugs.webkit.org/show_bug.cgi?id=181560

Attachment 331242: Patch

https://bugs.webkit.org/attachment.cgi?id=331242&action=review




--- Comment #18 from Alexey Proskuryakov <ap at webkit.org> ---
Comment on attachment 331242
  --> https://bugs.webkit.org/attachment.cgi?id=331242
Patch

Layout tests are useless to verify this change, as we always use the
UseFullySynchronousModeForTesting mode there. Also, we don't have high fidelity
tests for inline input.

> Note that it is possible to introduce a deadlock in this situation if
WebContent process sends a sync message and UI process needs to send WebContent
process another sync message before answering that sync message. Hopefully, we
don't have such a message, and Chris' testing shows that we don't seem to have
such a message.

UI processes sends sync messages at arbitrary times (text input client calls is
a notable case of sync messages without a timeout). So I don't see why you hope
that there aren't such messages.

This approach continues to look wrong to me. Furthermore, I do not agree that
this behavior is worth being concerned about. It is not unexpected that things
can happen while waiting for a sync response. In fact, it is exactly the same
behavior that we have in legacy WebKit - delegate callouts may and do reenter.