Connecting to GPU process may hang if UI process sends sync message simultaneously
Created attachment 458598 [details] WIP
<rdar://91713027>
Created attachment 458799 [details] Patch
Created attachment 458801 [details] Patch
Comment on attachment 458801 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=458801&action=review r=me with fixes. We may want to do the same for the network process? > Source/WebKit/ChangeLog:23 > + connection initialization. Send the GPUPg connection initialization result, audit token and Typo: GPUPg > Source/WebKit/GPUProcess/GPUProcess.cpp:134 > + WTFLogAlways("GOT Create gpu process connection"); Should drop. > Source/WebKit/UIProcess/GPU/GPUProcessProxy.cpp:374 > + WTFLogAlways("Create gpu process connection"); Should drop > Source/WebKit/WebProcess/GPU/GPUProcessConnection.cpp:132 > + if (parentConnection.ignoreInvalidMessageForTesting()) indentation is wrong > Source/WebKit/WebProcess/GPU/GPUProcessConnection.cpp:133 > + instance->connection().setIgnoreInvalidMessageForTesting(); ditto. > Source/WebKit/WebProcess/GPU/GPUProcessConnection.cpp:306 > + WTFLogAlways("Got didinitialize"); Should drop.
Created attachment 458858 [details] For landing
(In reply to Chris Dumez from comment #5) > Comment on attachment 458801 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=458801&action=review > > r=me with fixes. We may want to do the same for the network process? Thanks! I think network process, authn process and plugin process could all use this pattern. However, I see the network process startup seems to be a bit intricate too, given the retries (?) and all. Maybe if there's no immediate need we should do it a bit later? Maybe after it is apparent that this change didn't cause any unforeseen issues.
Committed r293829 (250302@main): <https://commits.webkit.org/250302@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 458858 [details].