Bug 40651 – Use "typedef wchar_t JSChar" when compiled with RVCT

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Kwang Yul Seo]

Fix Brew MP RVCT build. defined(AEE_STATIC) || defined(DYNAMIC_APP) checks Brew MP.

[Kwang Yul Seo]

Created attachment 58851 [details]
Patch

[Early Warning System Bot]

Attachment 58851 [details] did not build on qt:
Build output: http://webkit-commit-queue.appspot.com/results/3267195

[WebKit Review Bot]

Attachment 58851 [details] did not build on gtk:
Build output: http://webkit-commit-queue.appspot.com/results/3331220

[Kwang Yul Seo]

Created attachment 58852 [details]
Patch

Oops. ')' was missing.

[Laszlo Gombos]

Is __CC_ARM or __ARMCC__ defined for BREW ? 

f it is maybe it is better to remove the "&& defined(__SYMBIAN32__)" test and make this rule generic for all RVCT platforms.

[Kwang Yul Seo]

(In reply to comment #5)
> Is __CC_ARM or __ARMCC__ defined for BREW ? 
> 
> f it is maybe it is better to remove the "&& defined(__SYMBIAN32__)" test and make this rule generic for all RVCT platforms.

__CC_ARM or __ARMCC__ is defined when RVCT is used.

Any suggestion?

[Laszlo Gombos]

How about this ? 

-    && !((defined(__CC_ARM) || defined(__ARMCC__)) && defined(__SYMBIAN32__)) 
+    && !((defined(__CC_ARM) || defined(__ARMCC__)))

[Kwang Yul Seo]

(In reply to comment #7)
> How about this ? 
> 
> -    && !((defined(__CC_ARM) || defined(__ARMCC__)) && defined(__SYMBIAN32__)) 
> +    && !((defined(__CC_ARM) || defined(__ARMCC__)))

Okay. I will update the patch as you suggested.

[Kwang Yul Seo]

Created attachment 59054 [details]
Patch

[Kwang Yul Seo]

Ping, Laszlo!

[Gabor Loki]

Comment on attachment 59054 [details]
Patch

The UChar typedef is different in wtf/unicode/qt4/UnicodeQt4.h in case of RVCT. The same type is required for UChar and JSChar as well.

[Kwang Yul Seo]

(In reply to comment #11)
> (From update of attachment 59054 [details])
> The UChar typedef is different in wtf/unicode/qt4/UnicodeQt4.h in case of RVCT. The same type is required for UChar and JSChar as well.

Why don't we check !defined(__linux__) instead of defined(__SYMBIAN32__)? !defined(__linux__) covers both Symbian and Brew MP. I will modify wtf/unicode/qt4/UnicodeQt4.h as well.

[Kwang Yul Seo]

Created attachment 65732 [details]
Patch

[Gabor Loki]

> Created an attachment (id=65732) [details]

This looks good to me.
Although I do not expect any problem, I will test it on Linux tomorrow.

[Adam Barth]

Comment on attachment 65732 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=65732&action=prettypatch

> JavaScriptCore/wtf/unicode/qt4/UnicodeQt4.h:59
> -#if defined(Q_OS_WIN) || COMPILER(WINSCW) || (COMPILER(RVCT) && OS(SYMBIAN))
> +#if defined(Q_OS_WIN) || COMPILER(WINSCW) || (COMPILER(RVCT) && !OS(LINUX))
It's unclear to me why we're switching from a whitelist to a blacklist for something that's identified as a hack...

[Kwang Yul Seo]

(In reply to comment #15)
> (From update of attachment 65732 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=65732&action=prettypatch
> 
> > JavaScriptCore/wtf/unicode/qt4/UnicodeQt4.h:59
> > -#if defined(Q_OS_WIN) || COMPILER(WINSCW) || (COMPILER(RVCT) && OS(SYMBIAN))
> > +#if defined(Q_OS_WIN) || COMPILER(WINSCW) || (COMPILER(RVCT) && !OS(LINUX))
> It's unclear to me why we're switching from a whitelist to a blacklist for something that's identified as a hack...


We can detect Linux with defined(__linux__) and Symbian with defined(__SYMBIAN32__), but there is no such macro in Brew MP. Therefore, it is very hard to check Brew MP in JSStringRef.h. 

Because Linux is the only exception for now, I think it is okay to switch.

[Gabor Loki]

> > -#if defined(Q_OS_WIN) || COMPILER(WINSCW) || (COMPILER(RVCT) && OS(SYMBIAN))
> > +#if defined(Q_OS_WIN) || COMPILER(WINSCW) || (COMPILER(RVCT) && !OS(LINUX))
> It's unclear to me why we're switching from a whitelist to a blacklist for something that's identified as a hack...

I see no problem using this formula. As far as I know RVCT is used for the following target platforms: Symbian, Linux, WinCE, Brew MP.
Symbian, WinCE and Brew MP use wchar_t. The only exception is the Linux (which uses uint16_t). Well, it looks like a blacklist, not a whitelist to me.

[Kwang Yul Seo]

Comment on attachment 65732 [details]
Patch

r?, q? again for review.

[Kwang Yul Seo]

(In reply to comment #14)
> > Created an attachment (id=65732) [details] [details]
> 
> This looks good to me.
> Although I do not expect any problem, I will test it on Linux tomorrow.

Ping. Have you tested the patch on Linux?

[Gabor Loki]

> Ping. Have you tested the patch on Linux?

Sorry, I forgot to mention that the patch is OK for RVCT and GCC as well. I am sure Laszlo will give an r+ on it.

[Laszlo Gombos]

Comment on attachment 65732 [details]
Patch

r=me.

I agree that we're switching from white-list to blacklist, but I'm not sure if this is really a hack anymore, despite the comment in the source.

[WebKit Commit Bot]

Comment on attachment 65732 [details]
Patch

Clearing flags on attachment: 65732

Committed r68087: <http://trac.webkit.org/changeset/68087>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.