While working on #824, I found out that cleaning up that particular piece of code would not be doable cleanly without a refactor of the multi_sexp infrastructure. This is that refactor; it encapsulates the code of multi_sexp inside a class, an instance of which is then used to manage all the network stuff that sexps need to do.
This introduces a new mechanism for handling bit flags. Instead of using an int (or two, or three) to write bits into, this uses the std::bitset in a wrapper class that accepts an enum class as input. This means that certain classes of errors we've had a few of in the engine, where a coder used the wrong set of flags or checked them against the wrong variable (SF_STEALTH vs SIF_STEALTH, for example) can now be detected at a compiler level rather than not at all.
This allows to use standard TrueType Fonts for text rendering in FSO. It uses the NanoVG library which provides the ability of drawing 2D shapes. This library is used for most 2D operations now and can also be used directly by using the PathRenderer class.
Replaces calls to double-precision floating point trig functions with
calls to single-precision floating point trig functions as follows:
cos() -> cosf()
sin() -> sinf()
tan() -> tanf()
acos() -> acosf()
asin() -> asinf()
atan() -> atanf()
This allows the compiler to avoid generating unnecessary instructions
that convert a float to a double and then a double back to a float.
Signed-off-by: Peter Mitsis <[email protected]>
Both the ANG_TO_RAD() and fl_radians() macros were serving the same
purpose; they converted an angle from degrees to radians. To reduce
code duplication, ANG_TO_RAD() was replaced with the existing
fl_radians().
Signed-off-by: Peter Mitsis <[email protected]>
We were declaring an array of NUM_TURRET_ORDER_TYPES ints, and then only specifying the first oindex of them; the memcpy() could've been changed to multiply by oindex instead, but it seemed cleaner to just reset new_target_order to -1 in all values and get rid of the redundant per-ship reset.
Discovered by triggering an Int3() in aiturret.cpp (line 1167 as of this writing) and finding uninitialized values in the turret_targeting_order array.
Affected SEXPs include turret-set-target-order, turret-set-direction-preference, turret-set-rate-of-fire, turret-set-optimum-range, ship-turret-target-order, turret-get-primary-ammo, turret-set-primary-ammo, turret-get-secondary-ammo, and turret-set-secondary-ammo.
It's worth noting that while turret-set-direction-preference, turret-set-rate-of-fire, and turret-set-optimum-range were wrong, they were wrong in a way that wouldn't actually affect the outcome most of the time (thanks to the "if(turret == NULL)" checks simply skipping invalid subsystem names).
In 9ec10dd I thought I was fixing a memory leak, but the actual scenario was that the freeing was just delayed until later. This removes the freeing added by 9ec10dd and changes the free() calls into vm_free() calls.
ssm_create() subtracts the start coordinates from the position of the target, and then normalizes that. Since call-ssm-strike was using the position of the target as the start coordinates, this was obviously resulting in normalizing a null vec3d, generating the spurious warning. This just makes call-ssm-strike use a point immediately behind the target's position (using the target's orientation) as the starting coordinate, thereby avoiding the warning.
This bug was found by, of all people, woutersmits, while FREDding his happynewyear campaign. See http://www.hard-light.net/forums/index.php?topic=90069.0 for discussion. The crash was triggered by the use of ( validate-all-arguments ) as the entire sexp tree for one of his mission events.
This was a two-part fix. The first part was to ensure that every *-argument sexp (of which validate-all-arguments is but one) is actually used within a when-argument or every-time-argument tree. More specifically, every sexp operation that expects an argument handler should first check that the arg_handler node actually exists.
The second part was to partially revert one of Taylor's old commits:
Revision: 3644
Author: taylor
Date: Tuesday, October 24, 2006 9:17:24 AM
Message:
various bits of cleanup (warning messages, things that Valgrind complained about, etc.)
fix an out-of-bounds issue on Sexp_nodes[] which was leading to stack corruption
----
Modified : /branches/fs2_open_3_6_9/fs2_open/code/parse/sexp.cpp
This was intended to prevent a -1 array reference into Sexp_nodes[]. What Taylor did was to cause repeated calls of the parent_operator function to loop endlessly at the topmost parent. But there are certain edge cases (such as the one woutersmits discovered) where you need to know whether a parent is the topmost one. What Taylor should have done, and what this change does, is to keep the -1 as a valid return value but check for it in the places where it may cause an issue.
If the subsystem was <none>, sexp_beam_floating_fire() wasn't consuming the node, resulting in it also being parsed as the X coordinate (making it always zero, because atoi("<none>") == 0).
`multi_sexp_set_turret_primary_ammo()` and `multi_sexp_set_turret_secondary_ammo()` were using an uninitialized pointer instead of an uninitialized array, like every other instance of multi_get_string(). This is just begging to write to invalid memory.
Fabian Woltermann