Bug 108373 – [harfbuzz] WebKit fails to build with MinGW compiler because of invalid cast in HarfBuzzShaper.cpp

• Status: RESOLVED
• Resolution: FIXED
• Component: Platform

Build dies with the following error

  GEN      generate-testwebkitapi-forwarding-headers
  GEN      DerivedSources/JavaScriptCore/LLIntAssembly.h
offlineasm: Parsing ./Source/JavaScriptCore/llint/LowLevelInterpreter.asm and Programs/LLIntOffsetsExtractor.exe and creating assembly file DerivedSources/JavaScriptCore/LLIntAssembly.h.
offlineasm: No magic values found. Skipping assembly file generation.
make  all-am
make[1]: Wejście do katalogu `/home/pawel/src/webkit'
/usr/bin/mkdir -p ./.deps/DerivedSources
  CXX      Source/WebCore/platform/graphics/harfbuzz/ng/libWebCore_la-HarfBuzzShaper.lo
Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp: In member function 'bool WebCore::HarfBuzzShaper::shapeHarfBuzzRuns(bool)':
Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:340:138: error: invalid conversion from 'const UChar* {aka const wchar_t*}' to 'const uint16_t* {aka const short unsigned int*}' [-fpermissive]
In file included from /usr/i686-w64-mingw32/sys-root/mingw/include/harfbuzz/hb.h:32:0,
                 from Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.h:38,
                 from Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:32:
/usr/i686-w64-mingw32/sys-root/mingw/include/harfbuzz/hb-buffer.h:230:1: error:   initializing argument 2 of 'void hb_buffer_add_utf16(hb_buffer_t*, const uint16_t*, int, unsigned int, int)' [-fpermissive]
Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:342:167: error: invalid conversion from 'WTF::OwnArrayPtr<wchar_t>::PtrType {aka wchar_t*}' to 'const uint16_t* {aka const short unsigned int*}' [-fpermissive]
In file included from /usr/i686-w64-mingw32/sys-root/mingw/include/harfbuzz/hb.h:32:0,
                 from Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.h:38,
                 from Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:32:
/usr/i686-w64-mingw32/sys-root/mingw/include/harfbuzz/hb-buffer.h:230:1: error:   initializing argument 2 of 'void hb_buffer_add_utf16(hb_buffer_t*, const uint16_t*, int, unsigned int, int)' [-fpermissive]
make[1]: *** [Source/WebCore/platform/graphics/harfbuzz/ng/libWebCore_la-HarfBuzzShaper.lo] Błąd 1
make[1]: Opuszczenie katalogu `/home/pawel/src/webkit'
make: *** [all] Błąd 2

Created attachment 185553 [details]
proposed fix

Created attachment 185739 [details]
Update for directory rename

[EFL EWS Bot]

Comment on attachment 185739 [details]
Update for directory rename

Attachment 185739 [details] did not pass efl-ews (efl):
Output: http://queues.webkit.org/results/16266173

[WebKit Review Bot]

Comment on attachment 185739 [details]
Update for directory rename

Attachment 185739 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/16271193

[Peter Beverloo (cr-android ews)]

Comment on attachment 185739 [details]
Update for directory rename

Attachment 185739 [details] did not pass cr-android-ews (chromium-android):
Output: http://queues.webkit.org/results/16252352

Created attachment 185745 [details]
don't ommit const

[Martin Robinson]

Comment on attachment 185745 [details]
don't ommit const

This kind of cast always makes me nervous.

[Dominik Röttsches (drott)]

(In reply to comment #0)
> Build dies with the following error
> 
>   GEN      generate-testwebkitapi-forwarding-headers
>   GEN      DerivedSources/JavaScriptCore/LLIntAssembly.h
> offlineasm: Parsing ./Source/JavaScriptCore/llint/LowLevelInterpreter.asm and Programs/LLIntOffsetsExtractor.exe and creating assembly file DerivedSources/JavaScriptCore/LLIntAssembly.h.
> offlineasm: No magic values found. Skipping assembly file generation.
> make  all-am
> make[1]: Wejście do katalogu `/home/pawel/src/webkit'
> /usr/bin/mkdir -p ./.deps/DerivedSources
>   CXX      Source/WebCore/platform/graphics/harfbuzz/ng/libWebCore_la-HarfBuzzShaper.lo
> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp: In member function 'bool WebCore::HarfBuzzShaper::shapeHarfBuzzRuns(bool)':
> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:340:138: error: invalid conversion from 'const UChar* {aka const wchar_t*}' to 'const uint16_t* {aka const short unsigned int*}' [-fpermissive]

m_normalizedBuffer is declared as follow:
OwnArrayPtr<UChar> m_normalizedBuffer;
Are you sure you have a correctly sized UChar typedef/declaration on your platform? What configuration are you building? Does it match one of the bots?

(In reply to comment #8)

> 
> m_normalizedBuffer is declared as follow:
> OwnArrayPtr<UChar> m_normalizedBuffer;
> Are you sure you have a correctly sized UChar typedef/declaration on your platform? What  configuration are you building? Does it match one of the bots?

I try to cross compile WebkitGTK for Windows using MinGW stack, there is no such build bot like
that as far as i know.

[Dominik Röttsches (drott)]

(In reply to comment #9)
> (In reply to comment #8)
> 
> > 
> > m_normalizedBuffer is declared as follow:
> > OwnArrayPtr<UChar> m_normalizedBuffer;
> > Are you sure you have a correctly sized UChar typedef/declaration on your platform? What  configuration are you building? Does it match one of the bots?
> 
> I try to cross compile WebkitGTK for Windows using MinGW stack, there is no such build bot like
> that as far as i know.

I would try to follow up why UChar/wchar_t is not matching uint16_t in this configuration, maybe there is something wrong with your build configuration of ICU.

[Darin Adler]

Comment on attachment 185745 [details]
don't ommit const

Change looks fine. Patch needs a change log.

[Kalev Lember]

UChar is always defined as a 16 bit type. Depending on the build configuration, the exact type can vary, but it's always a 16 bit type.

In the GTK+/MinGW configuration, the preprocessor magic chooses it to be wchar_t, but only after making sure wchar_t is a 16 bit type.

Unfortunately, even though the source and target types are both 16 bit, gcc considers them to be different, so it needs an explicit reinterpret_cast.

From Source/JavaScriptCore/icu/unicode/umachine.h:
#if U_SIZEOF_WCHAR_T==2
    typedef wchar_t UChar;
#elif U_GNUC_UTF16_STRING
#if defined _GCC_
    typedef __CHAR16_TYPE__ char16_t;
#endif
    typedef char16_t UChar;
#else
    typedef uint16_t UChar;
#endif

Created attachment 203629 [details]
Patch updated with the changelog

[Martin Robinson]

Comment on attachment 203629 [details]
Patch updated with the changelog

View in context: https://bugs.webkit.org/attachment.cgi?id=203629&action=review

> Source/WebCore/ChangeLog:10
> +        No new tests (OOPS!).

You need to remove this line for the patch to pass the tools.

Created attachment 203631 [details]
updated

[WebKit Commit Bot]

Comment on attachment 203631 [details]
updated

Clearing flags on attachment: 203631

Committed r151141: <http://trac.webkit.org/changeset/151141>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.