[vlc-devel] [PATCH] decoder: re-order save/restore cancel

[Rémi Denis-Courmont]

Le 12 septembre 2017 13:59:57 GMT+03:00, Thomas Guillem <thomas at gllm.fr> a écrit :
>Mainly to clarify and use the same order when locking/restoring and
>saving/unlocking.
>
>This doesn't change anything since the thread cancellation type is
>deferred.
>---
> src/input/decoder.c | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
>diff --git a/src/input/decoder.c b/src/input/decoder.c
>index 109d5a2205..57273805d5 100644
>--- a/src/input/decoder.c
>+++ b/src/input/decoder.c
>@@ -1531,7 +1531,6 @@ static void *DecoderThread( void *p_data )
> {   /* Flush before/regardless of pause. We do not want to resume just
>       * for the sake of flushing (glitches could otherwise happen). */
>             int canc = vlc_savecancel();
>-
>             vlc_fifo_Unlock( p_owner->p_fifo );
> 
>             /* Flush the decoder (and the output) */
>@@ -1550,10 +1549,10 @@ static void *DecoderThread( void *p_data )
> 
>         if( paused != p_owner->paused )
>         {   /* Update playing/paused status of the output */
>-            int canc = vlc_savecancel();
>             mtime_t date = p_owner->pause_date;
>-
>             paused = p_owner->paused;
>+
>+            int canc = vlc_savecancel();
>             vlc_fifo_Unlock( p_owner->p_fifo );
> 
>         /* NOTE: Only the audio and video outputs care about pause. */
>@@ -1563,8 +1562,8 @@ static void *DecoderThread( void *p_data )
>             if( p_owner->p_aout != NULL )
>                 aout_DecChangePause( p_owner->p_aout, paused, date );
> 
>-            vlc_restorecancel( canc );
>             vlc_fifo_Lock( p_owner->p_fifo );
>+            vlc_restorecancel( canc );
>             continue;
>         }
> 
>@@ -1593,9 +1592,9 @@ static void *DecoderThread( void *p_data )
>              * drain. Pass p_block = NULL to decoder just once. */
>         }
> 
>+        int canc = vlc_savecancel();
>         vlc_fifo_Unlock( p_owner->p_fifo );
> 
>-        int canc = vlc_savecancel();
>         DecoderProcess( p_dec, p_block );
> 
>         if( p_block == NULL )
>@@ -1604,7 +1603,6 @@ static void *DecoderThread( void *p_data )
>             if( p_owner->p_aout != NULL )
>                 aout_DecFlush( p_owner->p_aout, true );
>         }
>-        vlc_restorecancel( canc );
> 
>         /* TODO? Wait for draining instead of polling. */
>         vlc_mutex_lock( &p_owner->lock );
>@@ -1616,6 +1614,8 @@ static void *DecoderThread( void *p_data )
>         vlc_fifo_Lock( p_owner->p_fifo );
>         vlc_cond_signal( &p_owner->wait_acknowledge );
>         vlc_mutex_unlock( &p_owner->lock );
>+
>+        vlc_restorecancel( canc );
>     }
>     vlc_cleanup_pop();
>     vlc_assert_unreachable();
>-- 
>2.11.0
>
>_______________________________________________
>vlc-devel mailing list
>To unsubscribe or modify your subscription options:
>https://mailman.videolan.org/listinfo/vlc-devel

I disagree. cancellation should be disabled the least. This documents the fact that the code is cancellation-safe, one way (no cancellation point) or another (proper clean-up handlers).
-- 
Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser ma brièveté.