Bug 96254 – [WTR] Generated source files should include config.h

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit2

[Ryuan Choi]

As http://www.webkit.org/coding/coding-style.html mentioned,
all implementations should include config.h.

[Ryuan Choi]

Created attachment 163115 [details]
Patch

[Ryuan Choi]

Created attachment 163119 [details]
Patch

[Benjamin Poulain]

Is this solving a real problem? You are just gonna slow down compile time for following the coding style _guidelines_.

[Ryuan Choi]

(In reply to comment #3)
> Is this solving a real problem? You are just gonna slow down compile time for following the coding style _guidelines_.

Yes.

In fact, a real problem is that WTR/Efl can not distinguish compile time options such as ENABLE(TOUCH_EVENTS) because
 - Derived sources does not include config.h
 WebKit/Efl does not use pre compiled header,
 so we should include config.h to determine options

 - CMake build scripts does not provide FEATURE DEFINES to generator. (Bug 96273 for this)

After solving above issues, I want to implement missing functionality of TOUCH EVENTS in EventSenderProxyEfl.cpp

Thank you.

[Benjamin Poulain]

Comment on attachment 163119 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=163119&action=review

> In fact, a real problem is that WTR/Efl can not distinguish compile time options such as ENABLE(TOUCH_EVENTS) because
>  - Derived sources does not include config.h
>  WebKit/Efl does not use pre compiled header,
>  so we should include config.h to determine options
> 
>  - CMake build scripts does not provide FEATURE DEFINES to generator. (Bug 96273 for this)

That's a good justification.

> Tools/WebKitTestRunner/InjectedBundle/Bindings/CodeGeneratorTestRunner.pm:332
>      unshift(@contents, map { "#include \"$_\"\n" } sort keys(%contentsIncludes));
> +    unshift(@contents, "#include \"config.h\"\n");
>      unshift(@contents, @contentsPrefix);

Is this really respecting the order of #includes regarding config.h?

[Ryuan Choi]

Comment on attachment 163119 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=163119&action=review

>> Tools/WebKitTestRunner/InjectedBundle/Bindings/CodeGeneratorTestRunner.pm:332
>>      unshift(@contents, @contentsPrefix);
> 
> Is this really respecting the order of #includes regarding config.h?

Yes, this obey the order of #include rule 1, `All implementation files must #include "config.h" first. Header files should never include "config.h"`.

CodeGeneratorTestRunner.pm still not follow other rules (All other headers are sorted).

[Benjamin Poulain]

Comment on attachment 163119 [details]
Patch

> Yes, this obey the order of #include rule 1, `All implementation files must #include "config.h" first. Header files should never include "config.h"`.

Oh, right! This is unshift, not shift. LGTM.

[Ryuan Choi]

Comment on attachment 163119 [details]
Patch

Thank you.

[WebKit Review Bot]

Comment on attachment 163119 [details]
Patch

Clearing flags on attachment: 163119

Committed r128108: <http://trac.webkit.org/changeset/128108>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.