Whether a ship is a support ship or its target, and whether it is destroyed or vanished, its repair/rearm mode should be cleaned up in a consistent way. This fixes a bug in Series Resurrecta where calling `ship-vanish` on a support ship put things into a broken state.
In the submodel reorganization in commit 575c050da4, subobject movement was moved out of the AI pipeline, which inadvertently removed the play-dead check that prevents subsystem movement. This broke a few missions such as Scroll's 'Thor's Hammer'. This PR restores the check.
Fixes#4933.
PR #4516 did not cover all of the autoscaled AI class attributes, since a few of them were labeled differently. This adds checks to grab the correct autoscale index for the remaining cases:
```
$Afterburner Use Factor:
$Shockwave Evade Chances Per Second:
$Get Away Chance:
$Secondary Range Multiplier:
```
Fortunately, Sushi's investigation was pretty thorough. Between these remaining cases and @EatThePath's autoscaling review, this PR should cover the complete set of autoscaling situations.
Fixes#4922.
Adds gravity! Applies on a per-mission basis, and weapons, ships, debris, and asteroids now have a configurable "gravity const" which multiplies the effect of gravity on them, by default everything except debris is 0 (meaning no effect even when the mission has gravity). AI and the lead indicator will also lead for ballistic trajectories as well, given a gravity-affected weapon. No FRED support or gravity direction control yet, to come in later PRs.
* Enable clients to move turrets
This allows clients to user server data to move turrets according to the timing and target data on the server. Uses a small packet to do so.
This adds an AI profiles flag that filters autoscaling so that only the index and number of autoscaled classes are considered, not the index and number of all AI classes.
Fixes#4516.
* Add AI flag to support ai bay arrivals > 10 seconds
Currently, if an AI ship is emerging from a bay path, the AI will stop following that path after only 10 seconds (and some distance away from the starting point). This results in AI breaking off a path early if a bay arrival path is long and/or uses a slow bay arrive multipliers. For bays that require the ship to follow the exact path this results in AI that then run into the parent ship and each other.
This PR adds a flag which disables that 10 second timeout. Also, in my tests having a flag to disable works best, as opposed to setting a max time (which is hard to calculate with varying ship speeds and varying bay path lengths).
PR is tested and works as expected.
* fix spelling
According to EatThePath's frame profiling, this one ship name lookup takes 2% of the frame time. It can be easily optimized by making use of the ship registry.
The original implementation of `ai_good_time_to_rearm` did not revert to being a "bad" time to rearm when the time was up. This fixes that and adds an AI profiles flag to toggle it.
Also converts `ai_rearm_timestamp` to use the new TIMESTAMP type.
* Allow `test_line_of_sight` to return intersecting object signature
The `test_line_of_sight` can currently provide the closest intersecting distance, but it would be very useful, especially for AI scripting, if this function could also return the signature of the intersecting object. This PR adds that functionality.
Tested and works as expected. Happy to discuss strategies for changing implementation, too. The extra, third return value for object signature was added to `getLineOfSightFirstIntersect` function because that function already returns 2 values. Thus, it will remain compatible with older scripts which only save the first 2 returned values.
* fix documentation
* fix typo
* return object instead of sig
* fix pointer
* use pointer pointer for full fix
* name clarification
* Add Animation Instance Flags
* Add SEXP barebones
* Unify and simplify animation parsing
* Clean up animation calling and use new animationlist api
* Add missing break
`Ai_flag_names[]`, `Object_flag_names[]`, and `Ship_flag_names[]` all had an associated `#define` containing their length that had to manually be incremented whenever new flag names were added to the list. This commit instead makes the number of flags in the list be computed by simple `sizeof` division, so now adding a new entry just requires adding the entry, rather than remembering to bump a `#define`. Tested in FRED to make sure populating `OPF_SHIP_FLAG` entries still works.
This assert catches any weapon's parent object type, which can include OBJ_GHOST, which is a dead player. OBJ_GHOST is being encountered in this function using BPC when the player dies via beam weapon and (likely) when the players weapon are hitting an enemy ship at the same time. Because the player ship can damage an enemy after it is dead, allow for that in this function.
PR #4500 added some checks, but accidentally compared `aip->target_objnum` to an object type (although the assertion message correctly retrieved the object type, leading to the bizarre assertion that `OBJ_SHIP`, i.e. 1, was an invalid target object type). This commit makes them all access the `Objects[]` array and retrieve the `.type`, just as the assertion message does (and the pre-#4500 conditional did).
Related to #4504, in that it's the source of the failed assertions, although I can't be sure it was the cause of the crashes without testing.
* Include more valid object types in this check
The previous version of the fix only included ships because it was believed to be the object that a ship used to be guarding. Now it's clear that this is an object that the ship is currently targeting *after* guarding something.
So remove the UNREACHABLE and manually check the valid object types with an assert. Then only continue with the function if it is the right target type.
* typo
* add subsystem indexer and ensure that ship_get_indexed_subsys and ship_get_first_subsys are independent
* also add reverse lookup from subsys to index
* update indexer to accommodate subsystem pointers being changed
* use clearer function names, and fix the type of subsys_list_indexer
* add assertions
* address latest feedbcak
1) Remove the extra branch in `check_ok_to_fire` that was added in SVN revision 10392, or commit 0d54630e83. If there was an actual Coverity error, this does not fix it at the proper location. More importantly, this breaks certain turret firing behavior.
2) Fix a couple of bounds checks.
3) Remove an incorrect stale comment.
This fixes the broken turret swarm weapons that were uncovered by the fixes in #2959.
For script hooks which use conditions, be more consistent about evaluating both objects involved, as well as evaluating the different types of objects.
Fixes#3874.
My guess is that, in `ai_ship_destroy`, if the dead ship's goal object has already died in the same frame, the `goal_objnum` will be -1 when `ai_formation_object_recalculate_slotnums` is about to be called.
Goober5000