[FFmpeg-devel] libavcodec/als: remove check for predictor order of a block

Thilo Borgmann thilo.borgmann at mail.de
Mon Nov 13 19:36:36 EET 2017


> On Mon, Nov 13, 2017 at 1:09 AM, Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
>> 2017-11-12 20:30 GMT+01:00 Umair Khan <omerjerk at gmail.com>:
>>> Hi,
>>> On Mon, Nov 13, 2017 at 12:45 AM, Carl Eugen Hoyos <ceffmpeg at gmail.com> wrote:
>>>> 2017-11-12 20:05 GMT+01:00 Umair Khan <omerjerk at gmail.com>:
>>>>> The attached patch fixes the address sanitizer issue.
>>>> Breaks compilation here, how did you test?
>>>> libavcodec/alsdec.c: In function ‘decode_var_block_data’:
>>>> libavcodec/alsdec.c:938:7: error: expected ‘}’ before ‘else’
>>> Sorry for the faulty patch. Here is the fixed one.
>> The commit message of your patch is:
>> libavcodec/als: fix address sanitization error in decoder
>> Is there an error in current FFmpeg git head that asan
>> shows? If not, the commit message makes no sense.
>> I believe you should send two patches that are meant
>> to be committed together, one of them fixing ticket #6297.
> This is the complete patchset.

I need some days to find time to test this, earliest during the weekend I fear...

What happens for
block_length < residual_index < opt_order?

Another way of asking would be, where is the second loop from specs page 30 for that case?
(ISO/IEC 14496)

I think what puzzles CE is, that the problematic if() from the other patches is still untouched by your patch. So how could this be a valid solution even if your patch would actually improve the prediction part...
And I wonder the same ;)

Did you run FATE with your patch applied? I assume a big difference in output at the first glance (means FATE aks the conformance files should fail...)

Thanks for driving this forward anyway :)


More information about the ffmpeg-devel mailing list