[OE-core] [PATCH 1/1] webkitgtk: turn off JIT on armv4 and armv7a

[Khem Raj]

> On May 18, 2016, at 8:16 PM, Gary Thomas <gary at mlbassoc.com> wrote:
> 
> On 2016-05-18 22:07, Khem Raj wrote:
>> On Wed, May 18, 2016 at 2:03 AM, Robert Yang <liezhi.yang at windriver.com> wrote:
>>> * It doesn't build on armv4:
>>> {standard input}: Assembler messages:
>>> {standard input}:52: Error: selected processor does not support `blx llint_throw_stack_overflow_error' in ARM mode
>>> {standard input}:126: Error: selected processor does not support `bkpt #0' in ARM mode
>>> {standard input}:128: Error: selected processor does not support `blx r0' in ARM mode
>>> {standard input}:134: Error: selected processor does not support `bkpt #0' in ARM mode
>>> {standard input}:185: Error: selected processor does not support `blx llint_throw_stack_overflow_error' in ARM mode
>>> {standard input}:256: Error: selected processor does not support `blx r4' in ARM mode
>>> {standard input}:310: Error: selected processor does not support `movw r2,#:lower16:.Lllint_op_enter-.LrelativePCBase' in ARM mode
>>> {standard input}:311: Error: selected processor does not support `movt r2,#:upper16:.Lllint_op_enter-.LrelativePCBase' in ARM mode
>>> {standard input}:315: Error: selected processor does not support `movw r2,#:lower16:.Lllint_op_get_scope-.LrelativePCBase' in ARM mode
>>> {standard input}:316: Error: selected processor does not support `movt r2,#:upper16:.Lllint_op_get_scope-.LrelativePCBase' in ARM mode
>>> [snip]
>>> 
>>> * It can build on armv7a, but doesn't work on runtime, cause
>>>   displaying problems or ephiphany hang.
>>> 
>>> [YOCTO #9474]
>>> 
>>> Signed-off-by: Robert Yang <liezhi.yang at windriver.com>
>>> ---
>>>  meta/recipes-sato/webkit/webkitgtk_2.12.1.bb | 7 ++++++-
>>>  1 file changed, 6 insertions(+), 1 deletion(-)
>>> 
>>> diff --git a/meta/recipes-sato/webkit/webkitgtk_2.12.1.bb b/meta/recipes-sato/webkit/webkitgtk_2.12.1.bb
>>> index bdbcbea..23ead72 100644
>>> --- a/meta/recipes-sato/webkit/webkitgtk_2.12.1.bb
>>> +++ b/meta/recipes-sato/webkit/webkitgtk_2.12.1.bb
>>> @@ -62,9 +62,14 @@ EXTRA_OECMAKE = " \
>>>  EXTRA_OECMAKE_append_powerpc = " -DENABLE_JIT=OFF "
>>>  EXTRA_OECMAKE_append_powerpc64 = " -DENABLE_JIT=OFF "
>>> 
>>> -# ARM JIT code does not build on ARMv5/6 anymore, apparently they test only on v7 onwards
>>> +# ARM JIT code does not build on ARMv4/5/6 anymore
>>>  EXTRA_OECMAKE_append_armv5 = " -DENABLE_JIT=OFF "
>>>  EXTRA_OECMAKE_append_armv6 = " -DENABLE_JIT=OFF "
>>> +EXTRA_OECMAKE_append_armv4 = " -DENABLE_JIT=OFF "
>>> +
>>> +# ARM JIT can build on armv7a, but doesnt' work on runtime, cause
>>> +# displaying problems or ephiphany hang.
>>> +EXTRA_OECMAKE_append_armv7a = " -DENABLE_JIT=OFF "
>> 
>> this should work fine with thumb2 e.g. so this is a little broad brush
>> here to diable it across
>> all armv7
> 
> Why do you think that changing the instruction set (to thumb2) would
> make the JIT work any better?

Assembler implementation for JIT has always worked better with thumb2.

> 
> If you tell me how, I'll test it on my i.MX6 (Cortex-A9) which is
> the platform that inspired the original bug report.


Try adding

ARM_INSTRUCTION_SET_armv7a = “thumb"
ARM_INSTRUCTION_SET_armv7ve = “thumb"

in the webkitgtk recipe and see if it helps

> 
>> 
>>> 
>>>  # binutils 2.25.1 has a bug on aarch64:
>>>  # https://sourceware.org/bugzilla/show_bug.cgi?id=18430
>>> --
>>> 2.7.4
>>> 
>>> --
>>> _______________________________________________
>>> Openembedded-core mailing list
>>> Openembedded-core at lists.openembedded.org
>>> http://lists.openembedded.org/mailman/listinfo/openembedded-core
> 
> 
> --
> ------------------------------------------------------------
> Gary Thomas                 |  Consulting for the
> MLB Associates              |    Embedded world
> ------------------------------------------------------------
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core at lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 204 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20160518/d9bb0e92/attachment-0002.sig>