Re: [meta-browser] [PATCH] chromium: fix null pointer dereference in V8 with gcc-6

From: Andrey Konovalov <andrey.konovalov@linaro.org>
To: Andre McCurdy <armccurdy@gmail.com>
Cc: openembeded-devel <openembedded-devel@lists.openembedded.org>
Subject: Re: [meta-browser] [PATCH] chromium: fix null pointer dereference in V8 with gcc-6
Date: Tue, 11 Apr 2017 16:02:26 +0300	[thread overview]
Message-ID: <58ECD3E2.1090601@linaro.org> (raw)
In-Reply-To: <CAJ86T=US2yATN4uK55O2vjoY-aJvU6Q-mJrkDuGOPEwsueHYug@mail.gmail.com>

On 04/11/2017 03:40 AM, Andre McCurdy wrote:
> On Fri, Apr 7, 2017 at 11:40 AM, Andrey Konovalov
> <andrey.konovalov@linaro.org> wrote:
>> This patch prevents "Aw Snap" error when loading a page with JavaScript.
>>
>> Tested by running:
>> - chromium-wayland on Beagle X15
>> - chromium-wayland on HiKey (with a separate patch to enable aarch64 build)
>> - cromium on dragonboard-410c (with a separate patch to enable aarch64 build)
>>
>> Based on the fix for Gentoo's bug 588596:
>> https://bugs.gentoo.org/show_bug.cgi?id=588596#c10
>>
>> Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
>> ---
>>   recipes-browser/chromium/chromium.inc | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/recipes-browser/chromium/chromium.inc b/recipes-browser/chromium/chromium.inc
>> index 60edccb..7e057ec 100644
>> --- a/recipes-browser/chromium/chromium.inc
>> +++ b/recipes-browser/chromium/chromium.inc
>> @@ -15,7 +15,8 @@ CHROMIUM_BUILD_TYPE ??= "Release"
>>   inherit gettext pythonnative
>>
>>   ARMFPABI_armv7a = "${@bb.utils.contains('TUNE_FEATURES', 'callconvention-hard', 'arm_float_abi=hard', 'arm_float_abi=softfp', d)}"
>
> Not related to the patch, but this over-ride should be duplicated for _armv7ve
>
>> -GYP_DEFINES += "${ARMFPABI} release_extra_cflags='-Wno-error=unused-local-typedefs' sysroot=''"
>> +GYP_DEFINES += " ${ARMFPABI} release_extra_cflags='-Wno-error=unused-local-typedefs' sysroot='' \
>> +       ${@bb.utils.contains("AVAILTUNES", "mips", "", "release_extra_cflags='-fno-delete-null-pointer-checks'", d)}"
>
> What's the significance of skipping the workaround if AVAILTUNES
> contains "mips" ?

This change (to add '-fno-delete-null-pointer-checks' to gcc flags) was initially merged into chromium directly,
but soon reverted as it broke mips somehow:
https://bugs.chromium.org/p/v8/issues/detail?id=3782#c10

>>   GYP_DEFINES_append_x86 = " generate_character_data=0"
>>
>>   do_configure() {
>> --
>> 2.1.4
>>
>> --
>> _______________________________________________
>> Openembedded-devel mailing list
>> Openembedded-devel@lists.openembedded.org
>> http://lists.openembedded.org/mailman/listinfo/openembedded-devel