I found it's easy to get the LLVM optimizer to emit code that hits this
warning, see the test I added:

https://github.com/kripken/emscripten/commit/308e98787f88789431e5210d4defa76c275454ac

The optimizer sees a union with an int write and a float read, and just
forwards those bits into a float.

On Tue, Nov 24, 2015 at 3:27 PM, Jukka Jylänki <[email protected]> wrote:

> Btw, I did write a PR for being able to handle such arbitrary NaN literals
> without warnings, but it was chosen not to merge that in, if I recall
> correctly, the thinking was that it would be an impossible scenario to
> occur. The related PR was here:
> https://github.com/kripken/emscripten-fastcomp/pull/116 which was amended
> to demote the handling to the warning you are seeing. It would be
> interesting to hear what kind of code pattern does cause this.
>
> 2015-11-25 0:36 GMT+02:00 Alon Zakai <[email protected]>:
>
>> I added logging of the instruction and function on incoming now. It will
>> also print out debug info if available.
>>
>> On Tue, Nov 24, 2015 at 8:29 AM, Robert Goulet <
>> [email protected]> wrote:
>>
>>> Unfortunately that doesn't really help us track down what produced this
>>> warning in our code.
>>>
>>>
>>> On Tuesday, November 24, 2015 at 10:51:29 AM UTC-5, arnab choudhury
>>> wrote:
>>>>
>>>> This error originates from the JS backend modifications to Clang-++.exe
>>>> in the emscripten fastcomp project:
>>>>
>>>>
>>>> https://github.com/kripken/emscripten-fastcomp/blob/4e83be90903250ec5142edc57971ed4c633c5e25/lib/Target/JSBackend/JSBackend.cpp
>>>>
>>>> Hope this helps.
>>>>
>>>> On Tuesday, November 24, 2015 at 7:28:15 AM UTC-8, Robert Goulet wrote:
>>>>>
>>>>> I got this warning at link time. I searched the code for this value
>>>>> but couldn't find it. Is there any way to find which source file produce
>>>>> this warning? EMCC_DEBUG=1 didn't tell more details about this.
>>>>> Thanks!
>>>>>
>>>>> emcc: warning: cannot represent a NaN literal '0x5fec1ef10' with
>>>>> custom bit pattern in NaN-canonicalizing JS engines (e.g. Firefox and
>>>>> Safari) without erasing bits!
>>>>>
>>>>> Thanks!
>>>>>
>>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "emscripten-discuss" 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.
>>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "emscripten-discuss" 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.
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "emscripten-discuss" 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.
>

-- 
You received this message because you are subscribed to the Google Groups 
"emscripten-discuss" 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