Report leaks after static destruction via _CRTDBG_LEAK_CHECK_DF instead of calling _CrtDumpMemoryLeaks() at the end of main(), which incorrectly reported every global container as a memory leak.
Also fix the three remaining real leaks which were discovered after all the noise was cleared up:
- Mc_point_list was never freed, because its atexit cleanup was only registered by code that no longer runs. But the list was only a temporary array of pointers used while building the collision tree, so remove it and have model_collide_parse_bsp_defpoints() copy the vertices directly into the tree's point list.
- model_batch_buffer's Mem_alloc copy was never freed. But the copy is actually not necessary, so remove it and pass Submodel_matrices directly.
- Goal_target_names strings were only freed when the next mission was parsed, so the last mission's names leaked at exit. Hold them in SCP_vm_unique_ptr so they are freed with the vector.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
* rename misleading variable check_again to never_check_again
The first element of collision_result means "never check again" per the
typedef comment, but was destructured as check_again (opposite meaning).
Rename to never_check_again to match the actual semantics and prevent
future confusion.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* fix data races in ship-weapon collision threading
Two thread-safety bugs in ship_weapon_check_collision, which runs on
worker threads:
1. update_danger_weapon() wrote to Ai_info[] (danger_weapon_objnum,
danger_weapon_signature) without synchronization. Multiple weapons
targeting the same ship would race on the same Ai_info entry.
2. Homing missile detonation logic wrote wp->lifeleft and
wp->weapon_flags directly on the Weapons[] global array.
Fix: defer both writes to the main-thread post-processing phase via
new flags in ship_weapon_collision_data. Also convert the collision
data from a 6-element tuple to a named struct for readability.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* fix data race in ship-ship collision submodel checking
ship_ship_check_collision() toggled pmi->submodel[].collision_checked
flags on the heavy ship's polymodel_instance to selectively enable/
disable submodel collision checks. When two collision pairs shared the
same heavy ship, worker threads would race on the same pmi entries.
Fix: add a collision_checked_override pointer to mc_info that lets
callers provide a thread-local array of collision_checked values.
model_collide() uses the override when set, falling back to the
pmi->submodel[] field otherwise. ship_ship_check_collision() now
builds a local SCP_vector<char> and passes it via the override,
completely avoiding writes to shared polymodel_instance state.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* Ensure threads have spun down before continuing non-threaded execution
* address review feedback for ship-weapon collision threading
- postproc flag now accumulates incrementally instead of being OR'd
at the final return, ensuring all return paths respect it
- early return for predictive collisions now preserves the
should_update_danger_weapon flag in the collision data
- moved nprintf detonation debug print to the main-thread process
pass for thread safety
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* move collision_checked from pmi into mc_info for thread safety
collision_checked flags were transient per-collision-pass state that
didn't belong in the shared polymodel_instance. Now mc_info owns an
SCP_vector<char> that model_collide auto-initializes from pmi when
empty. All collision callers (ship-ship, asteroid, debris, prop)
updated to manipulate mc.collision_checked instead of pmi->submodel[].
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* Convert to CV and spinup counter
* Update comment
---------
Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]>
Co-authored-by: BMagnu <[email protected]>
* Add flag to respect detailboxes in model collide checks
* Don't check detailboxes that are invisible for sunglare checks
* Use local coords
* Actualyl also run the check for cockpits & show ship
* Be less stupid
Add assertions to the model parsing macros to guard against accessing out-of-bounds memory. Patch the model parsing routines to avoid going out of bounds, even if the out of bounds situations are temporary and benign.
* Expose NoCollide and NoCollideThisOnly as virtual variables for submodels
* Allow submodels to be optionally specified in object:checkRayCollision
* Be sure to calculate ship-to-ship hit position before the override collision hook is called, in case the hook needs this value. (The other collision hooks already calculate the hit position before the override hook.)
Mostly renaming of fields and functions, and moving some modelread sections into functions. This reorganization will put the code on a better footing for the model translation feature. No model translation code is included yet, and there should be no change in functionality.
1) Changes the submodel flags to use a flagset instead of several booleans
2) Adds `rotate_accel` to allow submodel rotation acceleration to be specified
Submodel acceleration can now be instant; dumb-rotating submodels always instantly accelerate (and have always done so)
This makes it easier to determine what takes the most time while loading
the mission (loading models) and how we could optimize that load time
(optimize model loading).
This is not a bug fix but it also shouldn't cause any regressions since
there are no actual code changes so this could be merged before the
release of 3.8.
