- GCC is pickier than Visual C++. GCC requires that structs with constructors, etc that are
used in a union must be defined outside the union. VC++ lets you do it inline.
ZDoom defaulted to Boom's (buggy) angle adjustment.
Changed it so that
* Mode 0 is like Hexen, performing no adjustment at all. This still should match all known maps using this special.
* Mode 1 remains unchanged.
* Mode 2 replicates Boom's broken angle adjustment and is used in the xlat file.
* Mode 3 implements the correct angle adjustment that Boom originally intended.
(Note: Should some map require something different it should be handled with compatibility.txt instead of reverting this back to the broken way it was before.)
- fixed: Visplane checks should only compare the plane flags that are relevant for rendering and mask out the rest.
- floatified FTransform and made the visplane checks a bit less verbose by moving the comparison as an operator into FTransform.
Note that this operator needs forceinline on Visual Studio so that it won't get called as a function.
In some places, P_UndoPlayerMorph was called with the 'force' argument placed in the 'unmorphflag' parameter, so that 'forced' unmorphs would be not completely forceful.
I hope no mod relied on this weirdness...
Note that this required splitting P_SerializeWorld, because sector_t and FSectorPortal contain some actor pointers, for which the same rule applies: Portal linking can only be done after all sectors have been read, meaning it cannot be done along with the rest of the data in these structures.
Obviously such a change breaks savegame compatibility so the min. savegame version had to be increased again.
- This reverts commit 06216d733e.
- I don't know what I was thinking. Since stateowner is always available
to the wrapper function, and this code is only generated for the wrapper
function, it's a nonissue. The state is already located before calling
any function that uses it.
- This reverts commit 39df62b20e.
- Anything that needs to lookup a state also needs stateowner. See
FxMultiNameState::Emit(). I will need to be more selective when
de-actionifying functions.
- An actor function really only needs to be an action function if:
1. It can be called with no parameters specified, either because it takes
none or because all its parameters are optional. This lets SetState()
call it directly without creating a wrapper function for it.
2. It wants access to the callingstate or stateowner parameters. Most
functions don't care about them, so passing them is superfluous.
- DECORATE now has atan2(y,x) and VectorAngle(x,y) functions. They are
identical except for the order of their parameters. The returned angle
is in degrees (not radians).
Two issues:
Portal linking requires all skyboxes in the sectors to be present, without them some info is not there when needed.
UpdateWaterLevel was called in AActor::Serialize, which operated on the freshly loaded level where lots of things haven't been set up yet and plane heights may be wrong.
All the PortalBlocks* functions were turned into real function calls which for such a frequent check should be avoided. So now any linked portal sets a flag in the attached sectors so that it isn't necessary to check the skybox pointer each time which makes the functions subject to inlining again.
- When the camera texture for a player view was rendered, its psprites
would overwrite the ones stored for hardware acceleration.
* Use a temporary vissprite in R_DrawPSprite() until it's known that it
will be accelerated.
* Do not write to any of the VisPSprites*[] arrays until it's known that
the sprite will be accelerated.
This got accidentally committed. Even if this gets extended to reach through portals it needs to be done differently. FMultiBlockLinesIterator can't guarantee to get every sector that's being touched.
This is for rendering the sprite properly in all areas the actor touches. The original thinglist is not sufficient for this and Boom's touching thinglist has other purposes and collects too much data.
This new list will only get filled in when the actor is actually crossing a portal plane, for the normal sector thinglist will still be used.
This piggybacks on the msecnode_t code which has been extended to be able to handle more than one list by passing the sector's membert pointers as parameters.
The old code kept the dead thinker, resulting in constant deletion and recreation of the subsector links and PolyBSP because the interpolation kept running.
Changed it so that the thinker is destroyed and the polyobject gets blocked by setting a new flag.
This required some changes to the Trace function because it turned out that the original was incapable of collecting the required information:
* actors are now also linked into blockmap blocks on both sides if they occupy the boundary of a sector portal.
* Trace will no longer set up parallel traces in all parts connected with sector portal, but only use one trace and relocate that on the actual boundary.
This check is only active for linedef based portals, due to the large amount of maps that did it wrong with thing based portals.
Although it may well be that there are some maps that abuse this omission for linedef portals as well, these are better handled with a compatibility option if the need arises.
The main reason this was added is to streamline and optimize the portal handling between renderers in ZDoom and GZDoom. For that both need to show the same general behavior and for linedef portals it is also important to handle the same as in Eternity.
In some situations it can happen that the sector here is not the frontsector of the anchor linedef, because some colinear node line with opposite direction causes this to be positioned on the wrong side. The only remedy here is to explicitly set the correct sector after spawning these things.
This isn't necessary. When rendering no actors from other groups may ever come into view directly - only when the respective part of the level is rendered through a portal. But at that point the camera is in a position where it's already correctly placed with relation to that actor.
The only reason this even existed was that ZDoom's original VC projects used __fastcall. The CMake generated project do not, they stick to __cdecl.
Since no performance gain can be seen by using __fastcall the best course of action is to just remove all traces of it from the source and forget that it ever existed.
Added temporary solution for the same foreground and background colors of the title in OS X startup window
It's used in graphical startup screen, with Hexen style in particular (for example WolfenDoom - Blade of Agony)
Native OS X backend doesn't implement this yet
Due to the VC++ 2015 headers being rather bloated (the average include size per source is 400-500kb) this provides a noticable compile speedup, although right now this only covers the game code, so there should be more room for improvement.
This means that the varargs functions themselves are now responsible for parsing them into DrawParms.
This was done because DrawTextV made a blanket assumption that every single vararg has the size of a 32 bit integer and caused crashes when anything else was passed. It also failed to eliminate any tag that is incompatible with text display. These will now abort DrawText and trigger an assert.
In order to avoid passing around tag lists, DrawTextV needs to parse everything itself and then pass a fully initialized structure to DrawTexture. This cannot be done if all variants require a varargs tag list.
Apparently the only reason for the old approach was the 'hw' parameter which was never used.
What was the point of this strange setup anyway? MoveFloor and MoveCeiling were inlines calling the universal MovePlane, which had nothing better to do than a switch/case with two cases - floor and ceiling!
Normally this will adjust relative to the actual direction to the target, but with arbitrary portals that cannot be calculated so using the actual attack angle is the only option.
Sadly the mappers cannot be trusted to use a feature correctly. Despite repeatedly telling that portals on one-sided lines are problematic, everybody seems to do it this way - and then report bugs if it doesn't work. Under such circumstances the only safe option is to block such portals entirely.
See http://forum.zdoom.org/viewtopic.php?f=2&t=51511&p=898522#p898522 for a typical example of the problems this might cause.
This code was ported from the 2005 floating point version, at one point it replaced 128 with 0.5, but 128 as a fixed point value needs to be 1/512. as a floating point value.
This will fail when a trace starts directly on a block boundary in which case x is a whole number. It should always use 'floor(x)+1' to ensure that the calculated point is at the right or upper edge of a block.
* we do not really need compatibility PointOnLineSide here. Unlike the movement code it'd only affect some extreme edge cases.
* removed the special case for very short traces. This was a result of the original and very imprecise PointOnLine functions. Since those no longer get used here and floating point precision is a lot higher there is no need for this kind of treatment.
* PointOnLine checks for the sides of an actor's bounding box don't need a full PointOnLineSide call, a simple coordinate comparison is fully sufficient, and this can easily be done in the existing switch/case block.
This was accidentally deleted during one round of portal refactoring but is essential to prevent multiple teleport activations in one move.
Fixing this also allowed removing the fudging that was added to work around the issue in P_TryMove.
- This reverts commit c90a1c0c96.
- DECORATE looks to be very dependant on functions that take strings as
parameters receiving those strings as constants and not as expressions,
so being able to declare string variables with DECORATE is pretty much
useless.
- This is "support" in the very most basic sense. You can declare them,
but you can't actually do anything with them, since the decorate parser
can't handle expressions when it's parsing string arguments. However,
they seem to be getting properly initialized and destroyed, which is
what this was added to test. If it doesn't look like too much trouble, I
might try to turn them into something actually worth something.
- Values are tagged to allow for some measure of changing variable types
without automatically breaking savegames.
- Use these new methods to serialize the non-native variables in an
object. This allows for achiving non-ints.
This could happen if the damage calculations resulted in a value between 0 and 1, which for the actual check was multiplied with the damage parameter of P_RadiusAttack which inflated the fractional value to something that looked like actual damage but was later truncated.
This is the last bit of play code that needed to be altered, what's left is the underlying data representations of vertices, linedefs and sectors.
# Conflicts:
# src/p_setup.cpp
# src/r_things.cpp
* typo in calculating end position from a trace vector
* must use floor to convert from floating point block coordinate to block index to account for running off the negative side of the blockmap. (Int cast always rounds toward zero which is wrong here.)
* bad calculation of sight checking slopes - they has the actor's z coordinate duplicated.
- fixed scaling of automap markers.
- Since DECORATE already allows reading all declared variables in a class,
where's the utility in keeping this restriction in ACS?
- Variables must still be numeric types.
- SetUserVariable is still restricted to user variables only.
While testing this it became clear that with the higher precision of doubles it has to be avoided at all costs to compare an actor's z position with a value retrieved from ZatPoint to check if it is standing on a floor. There can be some minor variations, depending on what was done with this value. Added isAbove, isBelow and isAtZ checking methods to AActor which properly deal with the problem.
