VSE: Make time operations self-contained

This patch makes it possible to manipulate strips without need to use
update functions to recalculate effect and meta strips.

Prior to this change function `SEQ_time_update_sequence` had to be used
to update mainly effects and meta strips. This was implemented in a way
that relied on sorted list of strips, which can't always be done and in
rare cases this approach failed.

In case of meta strips, `seqbase` had to be passed and compared with
"active" one to determine whether meta strip should be updated or not.
This is especially weak system that is prone to bugs when functions are
used by python API functions.

Finally, other strip types had startdisp` and `enddisp` fields updated
by this function and a lot of code relied on these fields even if strip
start, length and offsets are available. This is completely
unnecessary.

Implemented changes:
All effects and meta strips are updated when strip handles are moved or
strip is translated, without need to call any update function.

Function `SEQ_time_update_sequence` has been split to
`SEQ_time_update_meta_strip_range` and
`seq_time_update_effects_strip_range`. These functions should be only
used within sequencer module code. Meta update is used for versioning,
which is only reason for it not being declared internally.

Sequence fields `startdisp` and `enddisp` are now only used for
effects to store strip start and end points. These fields should be
used only internally within sequencer module code.
Use function `SEQ_time_*_handle_frame_get` to get strip start and end
points.

To update effects and meta strips with reasonable performance, cache
for "parent" meta strip and attached effects is added to
`SequenceLookup` cache, so it shares invalidation mechanisms.
All caches are populated during single iteration.

There should be no functional changes.

Differential Revision: https://developer.blender.org/D14990

1 files changed, 12 insertions, 10 deletions

diff --git a/source/blender/sequencer/intern/render.c b/source/blender/sequencer/intern/render.c
index b0898be3765..e7a1bbeb9d0 100644
--- a/source/blender/sequencer/intern/render.c
+++ b/source/blender/sequencer/intern/render.c
@@ -1994,11 +1994,12 @@ ImBuf *SEQ_render_give_ibuf_direct(const SeqRenderData *context,
 
 float SEQ_render_thumbnail_first_frame_get(Sequence *seq, float frame_step, rctf *view_area)
 {
-  int first_drawable_frame = max_iii(seq->startdisp, seq->start, view_area->xmin);
+  int first_drawable_frame = max_iii(
+      SEQ_time_left_handle_frame_get(seq), seq->start, view_area->xmin);
 
   /* First frame should correspond to handle position. */
-  if (first_drawable_frame == seq->startdisp) {
-    return seq->startdisp;
+  if (first_drawable_frame == SEQ_time_left_handle_frame_get(seq)) {
+    return SEQ_time_left_handle_frame_get(seq);
   }
 
   float aligned_frame_offset = (int)((first_drawable_frame - seq->start) / frame_step) *
@@ -2011,7 +2012,7 @@ float SEQ_render_thumbnail_next_frame_get(Sequence *seq, float last_frame, float
   float next_frame = last_frame + frame_step;
 
   /* If handle position was displayed, align next frame with `seq->start`. */
-  if (last_frame == seq->startdisp) {
+  if (last_frame == SEQ_time_left_handle_frame_get(seq)) {
     next_frame = seq->start + ((int)((last_frame - seq->start) / frame_step) + 1) * frame_step;
   }
 
@@ -2088,8 +2089,9 @@ void SEQ_render_thumbnails(const SeqRenderData *context,
 
   /* Adding the hold offset value (seq->anim_startofs) to the start frame. Position of image not
    * affected, but frame loaded affected. */
-  float upper_thumb_bound = SEQ_time_has_right_still_frames(seq) ? (seq->start + seq->len) :
-                                                                   seq->enddisp;
+  float upper_thumb_bound = SEQ_time_has_right_still_frames(seq) ?
+                                (seq->start + seq->len) :
+                                SEQ_time_right_handle_frame_get(seq);
   upper_thumb_bound = (upper_thumb_bound > view_area->xmax) ? view_area->xmax + frame_step :
                                                               upper_thumb_bound;
 
@@ -2122,8 +2124,8 @@ void SEQ_render_thumbnails(const SeqRenderData *context,
 
 int SEQ_render_thumbnails_guaranteed_set_frame_step_get(const Sequence *seq)
 {
-  const int content_start = max_ii(seq->startdisp, seq->start);
-  const int content_end = min_ii(seq->enddisp, seq->start + seq->len);
+  const int content_start = max_ii(SEQ_time_left_handle_frame_get(seq), seq->start);
+  const int content_end = min_ii(SEQ_time_right_handle_frame_get(seq), seq->start + seq->len);
   const int content_len = content_end - content_start;
 
   /* Arbitrary, but due to performance reasons should be as low as possible. */
@@ -2143,10 +2145,10 @@ void SEQ_render_thumbnails_base_set(const SeqRenderData *context,
   SeqRenderState state;
   seq_render_state_init(&state);
 
-  int timeline_frame = seq->startdisp;
+  int timeline_frame = SEQ_time_left_handle_frame_get(seq);
   const int frame_step = SEQ_render_thumbnails_guaranteed_set_frame_step_get(seq);
 
-  while (timeline_frame < seq->enddisp && !*stop) {
+  while (timeline_frame < SEQ_time_right_handle_frame_get(seq) && !*stop) {
     ImBuf *ibuf = seq_cache_get(
         context, seq_orig, roundf(timeline_frame), SEQ_CACHE_STORE_THUMBNAIL);
     if (ibuf) {