[pulseaudio-discuss] [PATCH] sink-input: Don't access resampler to get silence memchunk

[Arun Raghavan]

On 16 January 2016 at 13:08, Tanu Kaskinen <tanuk at iki.fi> wrote:
> On Wed, 2015-12-16 at 09:52 +0530, Arun Raghavan wrote:
>> On 16 December 2015 at 00:36, Tanu Kaskinen <tanuk at iki.fi> wrote:
>> > On Tue, 2015-12-15 at 21:49 +0530, arun at accosted.net wrote:
>> > > From: Arun Raghavan <git at arunraghavan.net>
>> > >
>> > > There doesn't appear to be a good reason to restrict the memchunk length
>> > > to the resample max block size -- we're going to have the memory around
>> > > anyway.
>> >
>> > I think the reason is to make sure that we don't feed the resampler
>> > bigger chunks than what it can handle. The resampler has to allocate
>> > other memblocks during its operation, and those memblocks may be bigger
>> > than the input block, so if the input block is too large, the
>> > requirements for the other blocks will grow beyond the mempool max
>> > block size.
>> >
>> > However, pa_sink_input_peek() seems to protect against this anyway when
>> > doing resampling (it processes the input in smaller pieces if it's
>> > larger than the resampler max block size), so maybe this change is safe
>> > anyway.
>> >
>> > > Moreover, callers of pa_sink_input_get_silence() don't seem to
>> > > actually care about the chunk itself, just the memblock for creating
>> > > their own pa_memblockq.
>> >
>> > I don't understand this comment. pa_memblockq cares about the chunk
>> > itself, not just the memblock.
>>
>> I misread that code. It does work with the chunk, not the memblock of course.
>
> Do you plan to send v2 of this patch some time soon?

Do you see a need to change anything other than the commit message? I
can just drop the last sentence.

-- Arun