I think the intention with the cache weirdness was to ensure that vision radius changed as soon as the player crouched or went prone next to a piece of cover, as these actions don't take any time. It would be inappropriate for them to only update one second later after the player passed a turn or something.

I don't have time to compile your PR to test it but you may want to make sure that's still working as intended.

i just tested, and no, even in vanilla, vision cache is invalidated one move after going prone:
Vanilla:

This PR:

Thanks for checking for me. That's a separate issue then, and not something that matters for this PR.

I swear it used to happen on the same turn.

You may have already seen it. The topic offered another solution to the problem. More precisely, they added a condition for checking.

actually i did miss the second Ren suggestion, i'll check it now

Per #73552 (comment) my safer suggestion was flawed as the static was never reassigned. Let's try this one more time!

index bbe88a5b40..aa07232763 100644
--- a/src/lightmap.cpp
+++ b/src/lightmap.cpp
@@ -212,6 +212,7 @@ bool map::build_vision_transparency_cache( const int zlev )
     bool low_profile = player_character.has_effect( effect_quadruped_full ) &&
                        player_character.is_running();
     bool is_prone = player_character.is_prone();
+    static move_mode_id previous_move_mode = player_character.current_movement_mode();
 
     for( const tripoint &loc : points_in_radius( p, 1 ) ) {
         if( loc == p ) {
@@ -219,8 +220,12 @@ bool map::build_vision_transparency_cache( const int zlev )
             vision_transparency_cache[p.x][p.y] = LIGHT_TRANSPARENCY_OPEN_AIR;
         } else if( ( is_crouching || is_prone || low_profile ) && coverage( loc ) >= 30 ) {
             // If we're crouching or prone behind an obstacle, we can't see past it.
-            vision_transparency_cache[loc.x][loc.y] = LIGHT_TRANSPARENCY_SOLID;
-            dirty = true;
+            if( vision_transparency_cache[loc.x][loc.y] != LIGHT_TRANSPARENCY_SOLID ||
+                previous_move_mode != player_character.current_movement_mode() ) {
+                previous_move_mode = player_character.current_movement_mode();
+                vision_transparency_cache[loc.x][loc.y] = LIGHT_TRANSPARENCY_SOLID;
+                dirty = true;
+            }
         }
     }
 

You update the value previous_move_mode 21+ times per game turn.

i'll compile and test it later

@RenechCDDA nope, third attempt do not resolve the bug either

@RenechCDDA nope, third attempt do not resolve the bug either

Here's what I get on master:

and after applying the patch:

Profiling before:

And after:

And you might think our additions inside map::build_vision_transparency_cache are themselves a problem. Nope! It's all the memcpy.

Do you mean something is wrong with my test method or compiling?

I would absolutely ask you to double-check what code you compiled, maybe even do a fresh rebuild to be 200% sure there's no compiler junk going on.

After complete rebuild confirmed the suggestion worsk as expected

This is the one we discussed on discord right? Kevin didn't seem to think it was a problem, and it's passing the test, so let's see what happen, and if someone set up us the bomb.