Skip to contentSkip to conversation

qtjsbackend: fix hardfloat detection in v8

PATCH In review Eric Bénard 3 participants 6 messages May 24, 2013 – May 28, 2013

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 qtjsbackend: fix hardfloat detection in v8

---
 recipes-qt/qt5/qtjsbackend.inc                     |    1 +
 .../qtjsbackend/0001-Fix-hardfloat-detection.patch |   33 ++++++++++++++++++++
 2 files changed, 34 insertions(+)
 create mode 100644 recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch
recipes-qt/qt5/qtjsbackend.inc +1 -0
index 4479654..a0dc632 100644
@@ -13,4 +13,5 @@ INC_PR = "r0"
 
 SRC_URI += " \
     file://0002-v8.pro-respect-external-host-bindir-when-set.patch \
+    file://0001-Fix-hardfloat-detection.patch \
 "
recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch new file +33 -0
new file mode 100644
index 0000000..3d86fbf
@@ -0,0 +1,33 @@
+From 05a87fccb76447ce3cb4b926128af929f26469bb Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Eric=20B=C3=A9nard?= <[email protected]>
+Date: Wed, 22 May 2013 18:50:01 +0200
+Subject: [PATCH] Fix hardfloat detection
+
+based on the same patch for nodejs which says :
+
+gcc has a builtin define to denote hard abi when in use, e.g. when
+using -mfloat-abi=hard it will define __ARM_PCS_VFP to 1 and therefore
+we should check that to determine which calling convention is in use
+and not __VFP_FP__ which merely indicates presence of VFP unit
+
+The fix has been provided by Khem Raj <[email protected]>
+---
+ src/3rdparty/v8/src/platform-linux.cc |    2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/src/3rdparty/v8/src/platform-linux.cc b/src/3rdparty/v8/src/platform-linux.cc
+index f6db423..8ae2249 100644
+--- a/src/3rdparty/v8/src/platform-linux.cc
++++ b/src/3rdparty/v8/src/platform-linux.cc
+@@ -167,7 +167,7 @@ bool OS::ArmCpuHasFeature(CpuFeature feature) {
+ // calling this will return 1.0 and otherwise 0.0.
+ static void ArmUsingHardFloatHelper() {
+   asm("mov r0, #0":::"r0");
+-#if defined(__VFP_FP__) && !defined(__SOFTFP__)
++#if defined(__ARM_PCS_VFP) && !defined(__SOFTFP__)
+   // Load 0x3ff00000 into r1 using instructions available in both ARM
+   // and Thumb mode.
+   asm("mov r1, #3":::"r1");
+-- 
+1.7.10.4
+
Signature · 1 line
1.7.10.4
#1 lore raw

PATCH 2/2 qtbase: add 0004-Disable-mkv8snapshot.patch to 5.0.1

else qml application segfault.

---
 recipes-qt/qt5/qtbase_5.0.1.bb |    2 ++
 1 file changed, 2 insertions(+)
recipes-qt/qt5/qtbase_5.0.1.bb +2 -0
index 2dce2c1..dfcb742 100644
@@ -3,5 +3,7 @@ require ${PN}.inc
 
 PR = "${INC_PR}.0"
 
+SRC_URI += "file://0004-Disable-mkv8snapshot.patch"
+
 SRC_URI[md5sum] = "720e5061d75913dfa97bf8bef4a443df"
 SRC_URI[sha256sum] = "6f8a0489eb057f6cba5e77cabaec13dcd010da562b3b88f91fe960022d7c37c8"
Signature · 1 line
1.7.10.4
#2 lore raw
In reply to Eric Bénard
Quoted 6 lines · Eric Bénard
---
 recipes-qt/qt5/qtjsbackend.inc                     |    1 +
 .../qtjsbackend/0001-Fix-hardfloat-detection.patch |   33 ++++++++++++++++++++
 2 files changed, 34 insertions(+)
 create mode 100644 recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch

Looks good, but have you tried that the same patch does apply to all 4 versions we currently have?

Quoted 52 lines
recipes-qt/qt5/qtjsbackend.inc +1 -0
index 4479654..a0dc632 100644
@@ -13,4 +13,5 @@ INC_PR = "r0"
 
 SRC_URI += " \
     file://0002-v8.pro-respect-external-host-bindir-when-set.patch \
+    file://0001-Fix-hardfloat-detection.patch \
 "
recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch new file +33 -0
new file mode 100644
index 0000000..3d86fbf
@@ -0,0 +1,33 @@
+From 05a87fccb76447ce3cb4b926128af929f26469bb Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Eric=20B=C3=A9nard?= <[email protected]>
+Date: Wed, 22 May 2013 18:50:01 +0200
+Subject: [PATCH] Fix hardfloat detection
+
+based on the same patch for nodejs which says :
+
+gcc has a builtin define to denote hard abi when in use, e.g. when
+using -mfloat-abi=hard it will define __ARM_PCS_VFP to 1 and therefore
+we should check that to determine which calling convention is in use
+and not __VFP_FP__ which merely indicates presence of VFP unit
+
+The fix has been provided by Khem Raj <[email protected]>
+---
+ src/3rdparty/v8/src/platform-linux.cc |    2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/src/3rdparty/v8/src/platform-linux.cc b/src/3rdparty/v8/src/platform-linux.cc
+index f6db423..8ae2249 100644
+--- a/src/3rdparty/v8/src/platform-linux.cc
++++ b/src/3rdparty/v8/src/platform-linux.cc
+@@ -167,7 +167,7 @@ bool OS::ArmCpuHasFeature(CpuFeature feature) {
+ // calling this will return 1.0 and otherwise 0.0.
+ static void ArmUsingHardFloatHelper() {
+   asm("mov r0, #0":::"r0");
+-#if defined(__VFP_FP__) && !defined(__SOFTFP__)
++#if defined(__ARM_PCS_VFP) && !defined(__SOFTFP__)
+   // Load 0x3ff00000 into r1 using instructions available in both ARM
+   // and Thumb mode.
+   asm("mov r1, #3":::"r1");
+-- 
+1.7.10.4
+
Signature · 1 line
1.7.10.4
Signature · 1 line
Martin 'JaMa' Jansa     jabber: [email protected]
#3 lore raw
Quoted 11 lines · Martin Jansa
Quoted 6 lines · Eric Bénard
---
 recipes-qt/qt5/qtjsbackend.inc                     |    1 +
 .../qtjsbackend/0001-Fix-hardfloat-detection.patch |   33 ++++++++++++++++++++
 2 files changed, 34 insertions(+)
 create mode 100644 recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch

Looks good, but have you tried that the same patch does apply to all 4 versions we currently have?

still need to test it on _git.

Eric

#4 lore raw
In reply to Eric Bénard

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Op 24-05-13 11:44, Eric Bénard schreef:

Quoted 11 lines

recipes-qt/qt5/qtjsbackend.inc | 1 +
.../qtjsbackend/0001-Fix-hardfloat-detection.patch | 33
++++++++++++++++++++ 2 files changed, 34 insertions(+) create mode 100644 recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch

(quoted diff) +0 -0

b/recipes-qt/qt5/qtjsbackend.inc index 4479654..a0dc632 100644 --- a/recipes-qt/qt5/qtjsbackend.inc +++ b/recipes-qt/qt5/qtjsbackend.inc @@ -13,4 +13,5 @@ INC_PR = "r0"

Quoted 16 lines

SRC_URI += " \
file://0002-v8.pro-respect-external-host-bindir-when-set.patch \ + file://0001-Fix-hardfloat-detection.patch \ " diff --git a/recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch b/recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch new file mode 100644 index 0000000..3d86fbf --- /dev/null +++ b/recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch @@ -0,0 +1,33 @@ +From 05a87fccb76447ce3cb4b926128af929f26469bb Mon Sep 17 00:00:00 2001 +From: =?UTF-8?q?Eric=20B=C3=A9nard?= <[email protected]> +Date: Wed, 22 May 2013 18:50:01 +0200 +Subject: [PATCH] Fix hardfloat detection + +based on the same patch for nodejs which says : + +gcc has a builtin define to denote hard abi when in use, e.g. when +using -mfloat-abi=hard it will define __ARM_PCS_VFP to 1 and therefore +we should check that to determine which calling convention is in use +and not __VFP_FP__ which merely indicates presence of VFP unit + +The fix has been provided by Khem Raj <[email protected]>

Does chromium need a similar patch?

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
Comment: GPGTools - http://gpgtools.org

iD8DBQFRn1F/MkyGM64RGpERAljbAJ9SkkFECqTmkYc8gxj/5KwjR726IACfd5O7 oNta6Xz+YZDKhT/tnsZEwEI=
=P43l
-----END PGP SIGNATURE-----

#5 lore raw

Hi Koen,

Quoted 12 lines · Koen Kooi

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Op 24-05-13 11:44, Eric Bénard schreef:

Quoted 6 lines

recipes-qt/qt5/qtjsbackend.inc | 1 +
.../qtjsbackend/0001-Fix-hardfloat-detection.patch | 33
++++++++++++++++++++ 2 files changed, 34 insertions(+) create mode 100644 recipes-qt/qt5/qtjsbackend/0001-Fix-hardfloat-detection.patch

Does chromium need a similar patch?

we have one on 19.0.1049.3, no idea if 24.0.1312.52 was tested on armv7 but v8 seems to integrate the fix (no idea since when).

Eric