[WK2] Improve log channel decision routine on WebKit

• Component: WebKit2
• Status: RESOLVED INVALID

[KwangYong Choi]

initializeLogChannel() on GTK and QT on WebKit2 is not effective. It's not necessary to call initializeLogChannel() for every log channels.

[KwangYong Choi]

Created attachment 163603 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 163603 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=163603&action=review

This

> Source/WebKit2/Platform/mac/Logging.mac.mm:54
> +void platformInitializeLogChannels()
> +{
> +    initializeLogChannel(&LogContextMenu);
> +    initializeLogChannel(&LogIconDatabase);
> +    initializeLogChannel(&LogKeyHandling);
> +    initializeLogChannel(&LogSessionState);
> +    initializeLogChannel(&LogTextInput);
> +    initializeLogChannel(&LogView);

As a result of this patch, each platform is responsible for knowing about all log channels, not just about how to initialize them. This is not maintainable - adding a logging channel should not require changing platform specific code.

[KwangYong Choi]

Comment on attachment 163603 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=163603&action=review

>> Source/WebKit2/Platform/mac/Logging.mac.mm:54
>> +    initializeLogChannel(&LogView);
> 
> As a result of this patch, each platform is responsible for knowing about all log channels, not just about how to initialize them. This is not maintainable - adding a logging channel should not require changing platform specific code.

The reason why I am not changing the routine of Mac port is because its routine in WebCore is same. And I'm not familiar with Objective-C.

GTK and QT will find the enabled log channels based on environment variable. It need not to be executed if there is no environment variable.

[Alexey Proskuryakov]

This issue needs to be addressed in a way that doesn't introduce the design problem I pointed out. Perhaps Gtk and Qt could cache some results to avoid recalculating them unnecessarily. 

Also, I'm not sure why this is a problem to begin with, as this code is debug only.

[KwangYong Choi]

(In reply to comment #4)
> This issue needs to be addressed in a way that doesn't introduce the design problem I pointed out. Perhaps Gtk and Qt could cache some results to avoid recalculating them unnecessarily. 
> 
> Also, I'm not sure why this is a problem to begin with, as this code is debug only.

I just wanted to add this feature to EFL port. And I think, it can be modified to WebCore style implementation before I add it to EFL port.

I will found out how to add this feature to EFL without changing previous implementation.

Thank you for review.