[vlc-devel] [PATCH] decoder: remove misleading comment

[Thomas Guillem]

On Tue, Sep 3, 2019, at 13:22, Steve Lhomme wrote:
> The p_aout and p_vout need locking for reading as well (it wouldn't make much
> sense to only lock when writing).
> 
> For example p_aout can be modified in the DecoderThread during a ReloadDecoder.

ReloadDecoder() will first unload the audio decoder before writing the aout so we are safe here. I could document it.

> 
> p_vout cannot be modified outside of the vout_update_format/spu_new_buffer but
> it is also access in the DecoderThread, for example during DecoderProcessFlush()
> or the OutputChangeXXX functions.

That is what my comment say no ?

> 
> For functions only called by the decoder, it might be OK not to lock as the
> p_out/p_aout should only change during vout_update_format/aout_update_format
> which the decoder should call before doing other calls from various threads.
> ---
>  src/input/decoder.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/src/input/decoder.c b/src/input/decoder.c
> index 65e3dd6f6c..7011c99226 100644
> --- a/src/input/decoder.c
> +++ b/src/input/decoder.c
> @@ -101,12 +101,11 @@ struct decoder_owner
>      vlc_cond_t  wait_acknowledge;
>      vlc_cond_t  wait_fifo; /* TODO: merge with wait_acknowledge */
>  
> -    /* -- These variables need locking on write(only) -- */
> +    /* -- Theses variables need locking on read *and* write -- */
>      audio_output_t *p_aout;
>  
>      vout_thread_t   *p_vout;
>  
> -    /* -- Theses variables need locking on read *and* write -- */
>      /* Preroll */
>      vlc_tick_t i_preroll_end;
>  
> -- 
> 2.17.1
> 
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel