It looks like this will be a bit fiddly to fix, so I'd like to revert it 
now and unblock your patch.

https://codereview.chromium.org/1019393003/

Thanks,
Jacob

On Monday, 23 March 2015 15:52:33 UTC, [email protected] wrote:
>
> We are looking at it.
>
> On Monday, 23 March 2015 14:17:26 UTC, Toon Verwaest wrote:
>>
>> As far as I can tell this is broken. See 
>>
>> http://build.chromium.org/p/tryserver.v8/builders/v8_linux_arm64_rel/builds/4101
>>  
>> for a repro. 
>>
>> The reason why it is failing is that GenerateCallFunction pushes 
>> arguments a 
>> single time, before branching on the function to choose the correct   
>> HInstruction 
>> to use for the call. Given that there are 2 call instructions for the 
>> single 
>> push (HInvokeFunction and HCallFunction) we think we're not in the 
>> push-case 
>> anymore in DoInvokeFunction. DoCallFunction already reset   
>> after_push_argument_ 
>> to false. This causes us to load the value relative to the stack pointer 
>> at   
>> an 
>> incorrect offset. 
>>
>> https://codereview.chromium.org/262533006/ 
>>
>

-- 
-- 
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
--- 
You received this message because you are subscribed to the Google Groups 
"v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to