[Qt] Single-threaded QtWebKit configuration

• Component: WebKit Qt
• Status: RESOLVED FIXED

[Laszlo Gombos]

The following patch introduces a new configuration option (ENABLE_NO_THREADING) for the Qt port to build single-threaded WebKit. The proposed change impacts only the build system. The single-threaded option builds on top of existing WebKit configuration options.

The motivation of this change is to maintain a single-threaded QtWebKit option for embedded systems where creating threads dynamically is not desired.

[Laszlo Gombos]

Created attachment 30662 [details]
Proposed patch.

[Mark Rowe (bdash)]

ENABLE_NO_THREADING is an awkward name.  Why not DISABLE_THREADING, or set ENABLE_THREADING=0?

[Laszlo Gombos]

(In reply to comment #2)
> ENABLE_NO_THREADING is an awkward name.  Why not DISABLE_THREADING, or set
> ENABLE_THREADING=0?

The guidelines I set to myself
 - stay consistent with the ENABLE_XXX naming schema
 - the build should not change if the newly introduced flag is not set to minimize the impact of this less frequently used configuration. 

How about ENABLE_SINGLE_THREADED ? 

[Laszlo Gombos]

Created attachment 30838 [details]
Updated the patch to use ENABLE_SINGLE_THREADED as the guard.

[Ariya Hidayat]

Landed in 44411

[Laszlo Gombos]

Created attachment 31026 [details]
Fix compilation warnings in ThreadingNone.cpp

It seems to me that changes to ThreadingNone.cpp has not been committed as part of 44411; created a new patch just for those changes.

[Laszlo Gombos]

Reopen as part of https://bugs.webkit.org/attachment.cgi?id=30838 has not been committed as part
of 44411.

[Ariya Hidayat]

I prefer that we use UNUSED_PARAM macro. The argument names are always useful to help understanding the function.

[Laszlo Gombos]

Created attachment 31062 [details]
Use UNUSED_PARAM macro for unused arguments as Ariya suggested.

[Holger Freyther]

(In reply to comment #8)
> I prefer that we use UNUSED_PARAM macro. The argument names are always useful
> to help understanding the function.

Hehe, it is a matter of taste. In general we omit parameter names when it is obvious. I would argue that for a *None file the parameter names are not relevant at all and should be omitted. Something funny I stumbled across is in in UnusedParam.h:

/* don't use this for C++, it should only be used in plain C files or
   ObjC methods, where leaving off the parameter name is not allowed. */

#define UNUSED_PARAM(x) (void)x

we obviously don't stick to that rule...

[Laszlo Gombos]

I share Zecke's opinion for *None files so I canceled the review for https://bugs.webkit.org/attachment.cgi?id=31062. 

[Ariya Hidayat]

Closing it, since the issue is settled

[Laszlo Gombos]

(In reply to comment #12)
> Closing it, since the issue is settled

Ariya, I think we should still commit https://bugs.webkit.org/attachment.cgi?id=31026 before closing this bug. Change the status back to REOPEN.

[Ariya Hidayat]

Oops, sorry. I didn't read it carefully.

Landed in r44607
http://trac.webkit.org/changeset/44607