diff mbox series

[v2,1/1] qt5webkit: fix build issue with 32-bits armv8-a

Message ID 20180302115853.4181-1-gael.portay@savoirfairelinux.com
State Superseded
Headers show
Series [v2,1/1] qt5webkit: fix build issue with 32-bits armv8-a | expand

Commit Message

Gaël PORTAY March 2, 2018, 11:58 a.m. UTC 
Adds WTF platform support for the 32-bits armv8-a architectures.

Fixes:

	In file included from ./config.h:30:0,
	                 from ...
	./wtf/Platform.h:323:6: error: #error "Not supported ARM
architecture"
	 #    error "Not supported ARM architecture"
	      ^~~~~

The patch is an adaptation of an upstream fix in version 5.212 of
qtwebkit[1].

Unfortunately, the commit cannot be backported and has to be fixed since
the toolchain does not define __ARM_ARCH_8__ but __ARM_ARCH_8A__.

	$ host/bin/arm-buildroot-linux-gnueabihf-g++ -dM -E - < /dev/null | grep -i ARM
	#define __ARM_SIZEOF_WCHAR_T 4
	#define __ARM_FEATURE_SAT 1
	#define __ARM_ARCH_ISA_ARM 1
	#define __ARMEL__ 1
	#define __ARM_FEATURE_UNALIGNED 1
	#define __ARM_FEATURE_IDIV 1
	#define __ARM_FP 14
	#define __ARM_ARCH_8A__ 1
	#define __ARM_NEON_FP 6
	#define __ARM_SIZEOF_MINIMAL_ENUM 4
	#define __ARM_PCS_VFP 1
	#define __ARM_FEATURE_LDREX 15
	#define __ARM_FEATURE_QBIT 1
	#define __ARM_FEATURE_FMA 1
	#define __ARM_NEON__ 1
	#define __ARM_ARCH_PROFILE 65
	#define __ARM_32BIT_STATE 1
	#define __ARM_FEATURE_CLZ 1
	#define __ARM_ARCH_ISA_THUMB 2
	#define __ARM_ARCH 8
	#define __arm__ 1
	#define __ARM_FEATURE_SIMD32 1
	#define __ARM_FEATURE_CRC32 1
	#define __ARM_NEON 1
	#define __ARM_ARCH_EXT_IDIV__ 1
	#define __ARM_EABI__ 1
	#define __ARM_FEATURE_DSP 1

[1]: https://github.com/qt/qtwebkit/blob/35655d5f4bad248ead1700b59c381cc568b4e98b/Source/WTF/wtf/Platform.h#L241-L242

Cc: Arnout Vandecappelle <arnout@mind.be>
Reported-by: Brock Williams <brock@cottonwoodcomputer.com>
Reviewed-by: Adrian Perez de Castro <aperez@igalia.com>
Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com>
---

Changes since v1:
 - double check for __ARM_ARCH_8__ and __ARM_ARCH_8A__
 - add Reviewed-by Adrian Perez de Castro

 .../0005-Detect-32-bits-armv8-a-architecture.patch | 48 ++++++++++++++++++++++
 .../0003-Detect-32-bits-armv8-a-architecture.patch | 48 ++++++++++++++++++++++
 2 files changed, 96 insertions(+)
 create mode 100644 package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
 create mode 100644 package/qt5/qt5webkit/5.9.1/0003-Detect-32-bits-armv8-a-architecture.patch

Comments

Thomas Petazzoni March 2, 2018, 1:23 p.m. UTC | #1 
Hello,

On Fri,  2 Mar 2018 06:58:53 -0500, Gaël PORTAY wrote:
> Adds WTF platform support for the 32-bits armv8-a architectures.
> 
> Fixes:
> 
> 	In file included from ./config.h:30:0,
> 	                 from ...
> 	./wtf/Platform.h:323:6: error: #error "Not supported ARM
> architecture"
> 	 #    error "Not supported ARM architecture"
> 	      ^~~~~
> 
> The patch is an adaptation of an upstream fix in version 5.212 of
> qtwebkit[1].
> 
> Unfortunately, the commit cannot be backported and has to be fixed since
> the toolchain does not define __ARM_ARCH_8__ but __ARM_ARCH_8A__.

Do you know why we have this difference between __ARM_ARCH_8__ and
__ARM_ARCH_8A__ ?


> diff --git a/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch b/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
> new file mode 100644
> index 0000000000..1cb655c078
> --- /dev/null
> +++ b/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
> @@ -0,0 +1,48 @@
> +From 068bf2d2d91382ea0d8ec24a142a30ea429704db Mon Sep 17 00:00:00 2001
> +From: =?utf-8?q?Ga=C3=ABl=20PORTAY?= <gael.portay@savoirfairelinux.com>
> +Date: Tue, 15 Aug 2017 18:28:49 -0400
> +Subject: [PATCH 1/1] Detect 32-bits armv8-a architecture

Nit: generate your patch with "git format-patch -N" to get rid of the
"1/1" in [PATCH 1/1]. Indeed, this patch is the fifth patch in the
series :)

> +Upstream-Status: Inappropriate

This is not really correct: your patch is a backport from upstream,
with adaptations.

> +Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com>

(Both comments also apply to the 5.9.3 patch, of course).

Thanks!

Thomas
Gaël PORTAY March 2, 2018, 4:16 p.m. UTC | #2 
Thomas,

On Fri, Mar 02, 2018 at 02:23:01PM +0100, Thomas Petazzoni wrote:
> Hello,
> 
> On Fri,  2 Mar 2018 06:58:53 -0500, Gaël PORTAY wrote:
> > Adds WTF platform support for the 32-bits armv8-a architectures.
> > 
> > Fixes:
> > 
> > 	In file included from ./config.h:30:0,
> > 	                 from ...
> > 	./wtf/Platform.h:323:6: error: #error "Not supported ARM
> > architecture"
> > 	 #    error "Not supported ARM architecture"
> > 	      ^~~~~
> > 
> > The patch is an adaptation of an upstream fix in version 5.212 of
> > qtwebkit[1].
> > 
> > Unfortunately, the commit cannot be backported and has to be fixed since
> > the toolchain does not define __ARM_ARCH_8__ but __ARM_ARCH_8A__.
> 
> Do you know why we have this difference between __ARM_ARCH_8__ and
> __ARM_ARCH_8A__ ?
> 

Honestly I don't.

I think that the __ARM_ARCH_8__ is never defined and probably does not
exist (at least with gcc). But I may be wrong.

I tried to all the armv8 -mcpu option with gcc and I never had
__ARM_ARCH_8__ defined.

> 
> > diff --git a/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch b/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
> > new file mode 100644
> > index 0000000000..1cb655c078
> > --- /dev/null
> > +++ b/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
> > @@ -0,0 +1,48 @@
> > +From 068bf2d2d91382ea0d8ec24a142a30ea429704db Mon Sep 17 00:00:00 2001
> > +From: =?utf-8?q?Ga=C3=ABl=20PORTAY?= <gael.portay@savoirfairelinux.com>
> > +Date: Tue, 15 Aug 2017 18:28:49 -0400
> > +Subject: [PATCH 1/1] Detect 32-bits armv8-a architecture
> 
> Nit: generate your patch with "git format-patch -N" to get rid of the
> "1/1" in [PATCH 1/1]. Indeed, this patch is the fifth patch in the
> series :)
> 

Fixed for v3.

> > +Upstream-Status: Inappropriate
> 
> This is not really correct: your patch is a backport from upstream,
> with adaptations.
>

Well, okay.

> > +Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com>
> 
> (Both comments also apply to the 5.9.3 patch, of course).
> 
> Thanks!
> 
> Thomas
> -- 
> Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
> Embedded Linux and Kernel engineering
> http://bootlin.com
> _______________________________________________
> buildroot mailing list
> buildroot@busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

Regards,
Gael
Adrian Perez de Castro March 2, 2018, 9:19 p.m. UTC | #3 
Hello,

On Fri, 2 Mar 2018 11:16:24 -0500, Gaël PORTAY <gael.portay@savoirfairelinux.com> wrote:
 
> On Fri, Mar 02, 2018 at 02:23:01PM +0100, Thomas Petazzoni wrote:
> > Hello,
> > 
> > On Fri,  2 Mar 2018 06:58:53 -0500, Gaël PORTAY wrote:
> > >
> > > [...]
> > 
> > Do you know why we have this difference between __ARM_ARCH_8__ and
> > __ARM_ARCH_8A__ ?
> 
> Honestly I don't.
> 
> I think that the __ARM_ARCH_8__ is never defined and probably does not
> exist (at least with gcc). But I may be wrong.
> 
> I tried to all the armv8 -mcpu option with gcc and I never had
> __ARM_ARCH_8__ defined.

I haven't got the time to check right now, but I *suspect* that Clang (or
Apple's Clang) uses __ARM_ARCH_8__, because the initial support for 64-bit
ARM was done by Apple and they use Clang. 

Cheers,


--
 Adrián 🎩
diff mbox series

Patch

diff --git a/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch b/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
new file mode 100644
index 0000000000..1cb655c078
--- /dev/null
+++ b/package/qt5/qt5webkit/5.6.3/0005-Detect-32-bits-armv8-a-architecture.patch
@@ -0,0 +1,48 @@ 
+From 068bf2d2d91382ea0d8ec24a142a30ea429704db Mon Sep 17 00:00:00 2001
+From: =?utf-8?q?Ga=C3=ABl=20PORTAY?= <gael.portay@savoirfairelinux.com>
+Date: Tue, 15 Aug 2017 18:28:49 -0400
+Subject: [PATCH 1/1] Detect 32-bits armv8-a architecture
+MIME-Version: 1.0
+Content-Type: text/plain; charset=utf-8
+Content-Transfer-Encoding: 8bit
+
+Adds WTF platform support for the 32-bits armv8-a architectures.
+
+Theses toolchains define __ARM_ARCH_8A__ (for ARM architecture version)
+and __arm__ (for 32-bits word-size; __aarch64__ for 64-bits).
+
+This commit catches this new architecture (armv8a) within a #ifdef/#endif
+inside the if statement dedicated for 32-bits ARM detection.
+
+Fixes:
+
+	In file included from ./config.h:30:0,
+	                 from ...
+	./wtf/Platform.h:323:6: error: #error "Not supported ARM architecture"
+	 #    error "Not supported ARM architecture"
+	      ^~~~~
+
+Upstream-Status: Inappropriate
+Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com>
+---
+ Source/WTF/wtf/Platform.h | 4 ++++
+ 1 file changed, 4 insertions(+)
+
+diff --git a/Source/WTF/wtf/Platform.h b/Source/WTF/wtf/Platform.h
+index 562840cf7..9cf656845 100644
+--- a/Source/WTF/wtf/Platform.h
++++ b/Source/WTF/wtf/Platform.h
+@@ -243,6 +243,10 @@
+     || defined(__ARM_ARCH_7S__)
+ #define WTF_ARM_ARCH_VERSION 7
+ 
++#elif defined(__ARM_ARCH_8__) \
++    || defined(__ARM_ARCH_8A__)
++#define WTF_ARM_ARCH_VERSION 8
++
+ /* MSVC sets _M_ARM */
+ #elif defined(_M_ARM)
+ #define WTF_ARM_ARCH_VERSION _M_ARM
+-- 
+2.16.1
+
diff --git a/package/qt5/qt5webkit/5.9.1/0003-Detect-32-bits-armv8-a-architecture.patch b/package/qt5/qt5webkit/5.9.1/0003-Detect-32-bits-armv8-a-architecture.patch
new file mode 100644
index 0000000000..1cb655c078
--- /dev/null
+++ b/package/qt5/qt5webkit/5.9.1/0003-Detect-32-bits-armv8-a-architecture.patch
@@ -0,0 +1,48 @@ 
+From 068bf2d2d91382ea0d8ec24a142a30ea429704db Mon Sep 17 00:00:00 2001
+From: =?utf-8?q?Ga=C3=ABl=20PORTAY?= <gael.portay@savoirfairelinux.com>
+Date: Tue, 15 Aug 2017 18:28:49 -0400
+Subject: [PATCH 1/1] Detect 32-bits armv8-a architecture
+MIME-Version: 1.0
+Content-Type: text/plain; charset=utf-8
+Content-Transfer-Encoding: 8bit
+
+Adds WTF platform support for the 32-bits armv8-a architectures.
+
+Theses toolchains define __ARM_ARCH_8A__ (for ARM architecture version)
+and __arm__ (for 32-bits word-size; __aarch64__ for 64-bits).
+
+This commit catches this new architecture (armv8a) within a #ifdef/#endif
+inside the if statement dedicated for 32-bits ARM detection.
+
+Fixes:
+
+	In file included from ./config.h:30:0,
+	                 from ...
+	./wtf/Platform.h:323:6: error: #error "Not supported ARM architecture"
+	 #    error "Not supported ARM architecture"
+	      ^~~~~
+
+Upstream-Status: Inappropriate
+Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com>
+---
+ Source/WTF/wtf/Platform.h | 4 ++++
+ 1 file changed, 4 insertions(+)
+
+diff --git a/Source/WTF/wtf/Platform.h b/Source/WTF/wtf/Platform.h
+index 562840cf7..9cf656845 100644
+--- a/Source/WTF/wtf/Platform.h
++++ b/Source/WTF/wtf/Platform.h
+@@ -243,6 +243,10 @@
+     || defined(__ARM_ARCH_7S__)
+ #define WTF_ARM_ARCH_VERSION 7
+ 
++#elif defined(__ARM_ARCH_8__) \
++    || defined(__ARM_ARCH_8A__)
++#define WTF_ARM_ARCH_VERSION 8
++
+ /* MSVC sets _M_ARM */
+ #elif defined(_M_ARM)
+ #define WTF_ARM_ARCH_VERSION _M_ARM
+-- 
+2.16.1
+