Add UNIX_DOMAIN_SOCKETS specific bits for supporting NetworkProcess.
Created attachment 188823 [details] Patch
Comment on attachment 188823 [details] Patch Attachment 188823 [details] did not pass mac-wk2-ews (mac-wk2): Output: http://queues.webkit.org/results/16615127 New failing tests: media/video-controls-captions-trackmenu.html
Comment on attachment 188823 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=188823&action=review > Source/WebKit2/NetworkProcess/NetworkProcess.cpp:-185 > -#else Don't you need to keep this #else notImplemented(); here? Or are the only possible implementations protected by PLATFORM(MAC) and USE(UNIX_DOMAIN_SOCKETS) already? > Source/WebKit2/UIProcess/Network/NetworkProcessProxy.cpp:-114 > -#else Ditto. > Source/WebKit2/UIProcess/Network/NetworkProcessProxy.cpp:-164 > -#else Ditto. > Source/WebKit2/WebProcess/WebProcess.cpp:-354 > -#else Ditto.
(In reply to comment #3) > (From update of attachment 188823 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=188823&action=review > > > Source/WebKit2/NetworkProcess/NetworkProcess.cpp:-185 > > -#else > > Don't you need to keep this #else notImplemented(); here? Or are the only possible implementations protected by PLATFORM(MAC) and USE(UNIX_DOMAIN_SOCKETS) already? I was thinking that but now that I think it over it is not true. There are PLATFORM(WINDOWS) and other combinations like (PLATFORM(QT) && OS(WIN)). I am going to keep the #else notImplemented() case.
Created attachment 189070 [details] Patch
Comment on attachment 189070 [details] Patch In CoreIPC do we have a goal of making connectioney stuff work without all of these #if blocks? Because they suck!
(In reply to comment #6) > (From update of attachment 189070 [details]) > In CoreIPC do we have a goal of making connectioney stuff work without all of these #if blocks? > > Because they suck! Probably it would be worthwhile, but I think it should be a separate refactoring step. It's ok to me if you say it should be handled first.
(In reply to comment #7) > (In reply to comment #6) > > (From update of attachment 189070 [details] [details]) > > In CoreIPC do we have a goal of making connectioney stuff work without all of these #if blocks? > > > > Because they suck! > > Probably it would be worthwhile, but I think it should be a separate refactoring step. It's ok to me if you say it should be handled first. Seee: bug 110978
Created attachment 205781 [details] Patch
(In reply to comment #9) I changed the patch not to depend on Bug 110978. I will refactor IPC connection creation code all together once Anders Carlsson reviews Bug 110978.
Comment on attachment 205781 [details] Patch Is the patch in https://bugs.webkit.org/show_bug.cgi?id=110978 the way forward to remove this CoreIPC nonsense so we don't have to add this #ifdef hell to cross platform files?
Because if it is, then we need to renew focus in getting Anders to review/guide on 110978, because it is not okay to be doing this to cross-platform code.
Adding blocking relationship on 110978
(In reply to comment #12) > Because if it is, then we need to renew focus in getting Anders to review/guide on 110978, because it is not okay to be doing this to cross-platform code. Okay. I will rebase the patch in Bug 110978 on top of the HEAD and ask Anders to review it.
I think we can close it, because this change would be unnecessary after bug110978.
Reopen, because WK2 owners don't have time for reviewing the much more bigger refactoring patch in https://bugs.webkit.org/show_bug.cgi?id=110978 Please revise your previous opinion and let this patch land into trunk.
Created attachment 218695 [details] Patch
Comment on attachment 218695 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=218695&action=review > Source/WebKit2/NetworkProcess/NetworkProcess.cpp:200 > + CoreIPC::Connection::ConnectionDescriptor connectionDescriptor(CoreIPC::Connection::createPlatformConnection()); You should probably do: CoreIPC::Connection::SocketPair socketPair = CoreIPC::Connection::createPlatformConnection(); (see below) > Source/WebKit2/Platform/CoreIPC/Connection.h:132 > + class ConnectionDescriptor { > + public: > + Identifier clientIdentifier() const { return m_clientIdentifier; } > + Identifier serverIdentifier() const { return m_serverIdentifier; } > + private: > + friend class Connection; > + ConnectionDescriptor(Identifier clientIdentifier, Identifier serverIdentifier) > + : m_clientIdentifier(clientIdentifier) > + , m_serverIdentifier(serverIdentifier) > + { } > + Identifier m_clientIdentifier; > + Identifier m_serverIdentifier; > + }; I think I'd prefer something like: struct SocketPair { int client; int server; } This seems a little heavy-weight.
Created attachment 218740 [details] Patch
Committed r160308: <http://trac.webkit.org/changeset/160308>