QtWebengine build error with opus and silk codecs

SGaist

I don't know, but I don't think so but thumb instructions have several advantages that are pretty interesting in the embedded world. You should rather consider enabling it.

Kart

@SGaist ~~AFAIK, emulating neon in software will bring performance loss. My particular CPU hardware does not support neon and thereby, isn't it safe to disable it then ?~~
My CPU (cortex a9) does not support neon nor does the compiler.
-print-multi-lib gives

armv5te_arm9;@mcpu=arm926ej-s
a9;@mcpu=cortex-a9
a7;@mcpu=cortex-a7
armv5te_arm9_soft;@mcpu=arm926ej-s@mfloat-abi=soft
armv5te_arm9_vfp;@mcpu=arm926ej-s@mfloat-abi=softfp@mfpu=vfp
a9_soft;@mcpu=cortex-a9@mfloat-abi=soft
a9_softfp_vfp;@mcpu=cortex-a9@mfloat-abi=softfp@mfpu=vfp
a9_softfp_vfpv3-d16;@mcpu=cortex-a9@mfloat-abi=softfp@mfpu=vfpv3-d16
a7_soft;@mcpu=cortex-a7@mfloat-abi=soft
a7_softfp_vfpv4;@mcpu=cortex-a7@mfloat-abi=softfp@mfpu=vfpv4
a7_softfp_neon-vfpv4;@mcpu=cortex-a7@mfloat-abi=softfp@mfpu=neon-vfpv4
a7_hard_neon-vfpv4;@mcpu=cortex-a7@mfloat-abi=hard@mfpu=neon-vfpv4

I am stuck with this error now

libxxxx.a(yyyyy.o) uses VFP register arguments, libQt5WebEngineCore.so.5.9.1 does not

SGaist

Then you should create your own mkspecs that sets the correct flags for your processor.

Kart

@SGaist I am using my own qmake.conf.

#
# qmake configuration for building with arm400-linux-g++
#

MAKEFILE_GENERATOR      = UNIX
CONFIG                 += incremental
QMAKE_INCREMENTAL_STYLE = sublib


include(../common/linux.conf)
include(../common/gcc-base-unix.conf)
include(../common/g++-unix.conf)

# modifications to g++.conf
QMAKE_INCDIR +=/opt/arm400-linux/target/usr/include 
QMAKE_LIBDIR +=/opt/arm400-linux/target/usr/lib
QMAKE_CC                = arm400-linux-gcc 
QMAKE_CXX               = arm400-linux-g++
QMAKE_LINK              = arm400-linux-g++ -L/opt/arm400-linux/target/usr/lib -Wl,-rpath -Wl,/opt/arm400-linux/target/usr/lib
QMAKE_LINK_SHLIB        = arm400-linux-g++ -L/opt/arm400-linux/target/usr/lib -Wl,-rpath -Wl,/opt/arm400-linux/target/usr/lib

QMAKE_CFLAGS_RELEASE += -march=armv7-a -mcpu=cortex-a9
QMAKE_CXXFLAGS_RELEASE += -march=armv7-a -mcpu=cortex-a9

# modifications to linux.conf
QMAKE_AR                = arm400-linux-ar cqs
QMAKE_OBJCOPY           = arm400-linux-objcopy
QMAKE_STRIP             = arm400-linux-strip

load(qt_config)

I tried

QMAKE_CFLAGS_RELEASE += -march=armv7-a -mcpu=cortex-a9 -mfloat-abi=soft
QMAKE_CXXFLAGS_RELEASE += -march=armv7-a -mcpu=cortex-a9 -mfloat-abi=soft

But it came out something like "arm400-linux-g++: error: -mfloat-abi=soft and -mfloat-abi=hard may not be used together".

SGaist

From the looks of it, your system looks like the linux-tegra2-g++ mkspec, maybe this one will be more suited to copy from.

Kart

@SGaist
I have tried all these combinations

-mcpu=cortex-a9 -mfloat-abi=softfp -mfpu=vfp

-mcpu=cortex-a9 -mfloat-abi=softfp -mfpu=vfpv3-d16

-mtune=cortex-a9 -march=armv7-a -mhard-float -mfloat-abi=softfp 
-mfpu=vfpv3-d16

and i got the same error

libxxxx.a(yyyyy.o) uses VFP register arguments, libQt5WebEngineCore.so.5.9.1 does not

I tried the tegra2 mkspecs, but again another error.

ERROR at //build/config/android/internal_rules.gni:11:1: Assertion failed.
assert(is_android)
^-----
See //build/config/android/rules.gni:8:1: whence it was imported.
import("//build/config/android/internal_rules.gni")
^-------------------------------------------------
See //third_party/android_tools/BUILD.gn:5:1: whence it was imported.
import("//build/config/android/rules.gni")
^----------------------------------------
See //third_party/openmax_dl/dl/BUILD.gn:199:16: which caused the file to be included.
      deps = [ "//third_party/android_tools:cpu_features" ]
               ^-----------------------------------------
Project ERROR: GN run error!
Makefile:80: recipe for target 'sub-gn_run-pro-make_first' failed
make[3]: *** [sub-gn_run-pro-make_first] Error 3
make[3]: Leaving directory '/opt/qt/qt5.9.1arm_source/qt-everywhere-opensource-src-5.9.2/qtwebengine/src/core'
Makefile:77: recipe for target 'sub-core-make_first' failed
make[2]: *** [sub-core-make_first] Error 2
make[2]: Leaving directory '/opt/qt/qt5.9.1arm_source/qt-everywhere-opensource-src-5.9.2/qtwebengine/src'
Makefile:46: recipe for target 'sub-src-make_first' failed
make[1]: *** [sub-src-make_first] Error 2
make[1]: Leaving directory '/opt/qt/qt5.9.1arm_source/qt-everywhere-opensource-src-5.9.2/qtwebengine'
Makefile:834: recipe for target 'module-qtwebengine-make_first' failed
make: *** [module-qtwebengine-make_first] Error 2

SGaist

Might be a silly question but did you nuke properly the build folder between tries ?

Kart

@SGaist I am deleting the source folder and doing all the steps from the beginning each and everytime.

SGaist

Ok, so when you said you tried the Tegra2 mkspec, did you just use it or did you copy from it into yours like I suggested ?

Kart

@SGaist This is how i used it.

include(../common/linux_device_pre.conf)

QMAKE_INCDIR_POST      += $$[QT_SYSROOT]/usr/include

QMAKE_LIBDIR_POST      += $$[QT_SYSROOT]/usr/lib

QMAKE_RPATHLINKDIR_POST += $$[QT_SYSROOT]/usr/lib

QMAKE_CFLAGS           += -mtune=cortex-a9 -march=armv7-a -mhard-float -mfloat-abi=softfp -mfpu=vfpv3-d16
QMAKE_CXXFLAGS         += -mtune=cortex-a9 -march=armv7-a -mhard-float -mfloat-abi=softfp -mfpu=vfpv3-d16

include(../common/linux_arm_device_post.conf)

load(qt_config)

This is my qmake.conf.

Kart

Does anyone knows how ninja build in qtwebengine is getting these architecture values by default ?

-mfloat-abi=hard -mtune=generic-armv7-a -mfpu=vfpv3-d16 -mthumb

SGaist

From a quick look, somewhere around:

src/3rdparty/chromium/breakpad/src/build/common.gypi
src/3rdparty/chromium/build/config/compiler/BUILD.gn
src/3rdparty/chromium/build/common.gypi
src/3rdparty/chromium/native_client/build/standalone_flags.gypi
src/3rdparty/chromium/native_client/build/untrusted.gypi
./src/core/config/linux.pri:    MFLOAT = $$extractCFlag("-mfloat-abi=.*")

Kart

Thanks @SGaist. This file had some information. I am not sure with the syntax to make changes in this file.

./src/core/config/linux.pri:    MFLOAT = $$extractCFlag("-mfloat-abi=.*")

Whatever values I am giving to QMAKE_CFLAGS_RELEASE and QMAKE_CXXXFLAGS_RELEASE is not what GN build is taking.

Maybe if I figure out how to make changes in the file you have mentioned, I might be able to solve this issue.

Kart

In the file

qtwebengine/src/3rdparty/chromium/third_party/opus/BUILD.gn:     use_opus_rtcd = current_cpu == "arm" && (is_win || is_android || is_linux)

qtwebengine/src/3rdparty/chromium/third_party/opus/BUILD.gn: 
 if (use_opus_rtcd) {
      sources += [
        "$target_gen_dir/celt_pitch_xcorr_arm_gnu.S",
        "src/celt/arm/arm_celt_map.c",
        "src/celt/arm/armcpu.c",
        "src/celt/arm/armcpu.h",
        "src/celt/arm/celt_neon_intr.c",
        "src/celt/arm/fft_arm.h",
        "src/celt/arm/mdct_arm.h",
        "src/celt/arm/pitch_arm.h",
        "src/silk/arm/NSQ_neon.c",
        "src/silk/arm/NSQ_neon.h",
        "src/silk/arm/arm_silk_map.c",
      ]

As far as I understood if the CPU is based on ARM and the OS is linux use_opus_rtcd becomes TRUE. If it becomes true, qtwebengine/src/3rdparty/chromium/third_party/opus/src/celt/arm/armcpu.c decides whether it have the NEON capabilities. I am trying to see that changing armcpu.c to not to use NEON might work or not.

Torsten Krause

The chromium snapshot is outdated. There is already a fix for that!

link text