Reduce glcalls noise

Remove excessive changes to context that instantly get reverted

Explicitly disable multisample

cbuf: recycle node after it executes, clear commands on Host_Error()

This reverts 626658074c8dd5c2288becd836103e2efdb9b741 which looks like a
potentially racy or non-threadsafe approach, and instead prevents
problems by clearing the command buffers if Host_Error() is called and
doesn't shut down DP.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

build: enable strict aliasing optimisations

Extensive code quality improvements have been made since this was last
tried, and compilers are smarter with better warnings now.

This doesn't change the Xonotic pipeline gameplay hash so it's highly
unlikely to cause any subtle regressions.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Change to the C99 standard method of testing for nan

This fixes all remaining strict-aliasing rule warnings in GCC.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

client: improve server connection status messages

Always logs various errors and rejections that previously were only
visible with developer cvar(s) set.  Among other things this fixes
silent failure when trying to join a full server.
Includes IP address in more messages.

Uses CON_ERROR and CON_WARN colours for various errors and warns.

Moves status message generation to the code that sets the relevant
states.  This allows more information to be included, makes the code
easier to follow and reduces polling.

Fixes some cases where a message was always clobbered before it could be
displayed.

Merges the built-in menu and loading screen connection status generation
and buffering.

Simplifies challenge request retry counting.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

README.md: mention `make clean`, more specific build deps

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Rename sv_gameplayfix_customstats to sv_qcstats, improve description

It's an API change not a gameplay change.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Sys_Sleep overhaul

One function to rule them all, and in the darkplaces sleep threads.

Makes sub-millisecond accuracy available on Windows.
Increases the number of sockets we can monitor on Windows.
Improves Curl_Select() a little.
Updates flags of some timing-related cvars.

Cleans up disorganised and duplicate code.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cl_nettimesyncboundmode: use default if invalid

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cl_nolerp: use normal time synchronisation

Fixes constant packet discards on LAN and impossible pings online,
when FPS > ticrate.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cl_netfps: reduce pings by synchronising when possible, related fixes

Fixes possibility of input frequency to exceed cl_netfps in certain
configurations.

Fixes possibility of rate limiting to reduce input frequency more than
it should.

Fixes timescale affecting input frequency (now consistent with client
FPS and server HZ being unaffected by timescale).

History of this code (most recent first):
1b6ea396b616e359cfe98d1e85e8e7a9ac571e91
528b92ffab3a9ab09770310e2796adbe5f0f1e20
79e332ad14e7ec9dbd2702a2d28127f7600f0678
663279ba8feeb4c114df0950bbc820f3bb32dcdc
21beb5fe4fa2e1f1b7591a4a1f6013b82186912f

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cl_maxfps*: various adjustments

Ensures minimum FPS of 10 even while idle to prevent weird issues
(networking assumes at least 10).
If you want to save more resources, hide the game window (disables
rendering entirely) such as by using vid_minimize_on_focus_loss 1.

Disables cl_maxfps_alwayssleep by default, on modern systems with
accurate timers and higher FPS this has a significant impact.

Allows cl_maxfps_alwayssleep to be used with cl_maxfps 0, this was
requested some time ago.

Switches from cl_maxidlefps to cl_maxfps when querying servers to give
pings representative of what can be expected when playing on them.

Cleans up the cl_maxfps* code.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

net_slist: add PGUP PGDN HOME END key support to builtin menu

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

net_slist: overhaul dpmaster and server pinging

Fixes https://gitlab.com/xonotic/darkplaces/-/issues/173
Fixes https://gitlab.com/xonotic/darkplaces/-/issues/211

Adds cvars net_slist_maxping, net_slist_interval, net_slist_debug.

Simplifies state tracking, ping is now either valid or not set (0) and
isn't abused to store state.

Includes a hash-verified timestamp in pings, fixes impossible pings
(happened when querytime was updated before response arrived) and the
ability to add servers to a client's list with unsolicited packets
(abusing the LAN support).

Checks that DP/QW master server list packets originate from a configured
master (cvar net_sourceaddresscheck).

Fixes unreliable LAN support.

Significantly reduces the total time to complete a full query cycle for
DP servers by improving the logic as well as turning up some cvars.

Prints warnings if DP or QW master server queries time out.

Makes net_slist_maxtries actually give better ping reports like its
description says.

Makes net_slist_pause completely pause (QC builtin resorthostcache()
used by Xonotic and Nexuiz bypassed it), clarifies description, uses a
callback instead of polling to handle changes.

Changes sv_masters[] and sv_qwmasters[] iteration to a more idiomatic
pattern pattern (numeric limit instead of a "null cvar" terminator).

Checks that sockets are open before writing to them in
NetConn_QueryQueueFrame() to avoid crashing in case we ever close a
lower numbered socket while a higher numbered one is still open.

Separates QW master list parsing into a func (like DP list parsing).

Fixes sending a DP master query string to the QW masters if the first
socket is an IPv6 one.

Fixes double-sending of QW favourite queries.

Minor cleanups/refactoring.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

net_slist: update types, declarations

Mostly this changes array indices and externally-sourced data that
should never trigger signed overflow to unsigned,
and changes to lhnetaddresstype_t where applicable.

Fixes an old bug in the DP Quake server list where if there are less
servers than the most that could fit onscreen, the list jumps around
and/or duplicates entries when moving the cursor down.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

net_slist: code style/formatting fixes

I wouldn't normally nitpick this stuff but the random tabs mixed into
code and comment lines were too annoying and fixing it later commits
would make their diff too noisy.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

build: fix buildstring

The revision was always "-", there was code duplication and
inconsistency, the meaning of some strings was unclear.

Frees up some space in the `status` "version:" line by moving
gamenetworkfiltername to "protocol:" which seems like a more relevant
section anyway.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

build: disallow use of VLAs

They're no longer allowed in Linux for safety and performance reasons.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

PRVM: indicate which VM triggered some common string messages

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix bug where +binds could get latched at low FPS

This was introduced in 6b21c467a83808e3becf61afe7b67a88f870b502 which
didn't ensure the -bind would be executed after the +bind when the
keydown and keyup events are processed in the same frame.

Fixes https://github.com/DarkPlacesEngine/darkplaces/issues/106

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

vid: move sound start/restart out of video start/restart

According to the comments in S_Restart_f() we must not restart sound
while running a map, but `vid_restart` did do that...

Also simplified/consolidated code paths a little.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

vid: add cvar vid_minimize_on_focus_loss

Default of 0 matches current SDL default behaviour.

If set, the environment variable SDL_VIDEO_MINIMIZE_ON_FOCUS_LOSS will
override the cvar.

Changes apply immediately.

Also makes changes to vid_mouse_clickthrough apply immediately.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

macOS: update Info.plist

Declares DP as a game to enable "Game Mode", see
https://gitlab.com/xonotic/xonotic/-/commit/7c57b3b0d8810e69094a338dd8e94ac8ca8a6f83

Includes the other changes from Xonotic.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Windows: opt out of DPI scaling

SDL >= 2.24.0 is required.

Fixes https://gitlab.com/xonotic/darkplaces/-/issues/32
Fixes https://github.com/DarkPlacesEngine/darkplaces/issues/74

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

input: activate DP_BUTTONCHAT if the window is hidden or loses focus

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

input: stop capturing mouse movement if window is hidden without first losing focus

Fixes players looking all around after they alt+tabbed.

Moves the logic for selecting the mouse mode out of screen drawing and
into SDL event handling where it's unaffected if we early-out of drawing
the frame.

Prevents mouse "grab" (raw/relative mode + cursor hiding) in a few
"loading" situations.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

GL: display a more useful error dialog if context creation fails

There's no need to repeatedly retry if context creation fails, the
GPU and driver capabilities won't have changed.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

SDL: print version earlier, once only

No need to print it again at every vid_restart, it can't change.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

vid: implement monitor selection with instant cvar changes

Adds cvars: vid_display, vid_info_displaycount

Prints the monitors detected at startup and on hotplug.

Requires SDL headers >= 2.0.18 at build time.
At runtime older versions should still work but may lack some functionality.

May not work properly or at all if vid_desktopfullscreen is 0.

May require vid_restart after hotplug events.

---

Changes existing cvars vid_fullscreen, vid_desktopfullscreen, vid_resizable,
vid_borderless to apply instantly.

Also (for windowed mode) works around window managers that move the
window after SDL creates and centres it (typically when adding a
titlebar), so the crosshair is always at the centre of the monitor.

Adds some missing SDL_WINDOWEVENT enumerations.

Changes the GL viewport size retrieval from screen coordinates to
pixels, which is a prerequisite for "HIGHDPI" aka "Apple Retina"
support.

Fixes https://gitlab.com/xonotic/darkplaces/-/issues/385

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

vid_vsync: fully support adaptive, remove polling, print better errors

Adaptive vsync was partially supported: it worked if set before window
creation but not if toggled on afterwards, and wasn't mentioned in the
cvar description.

The immediate application of `vid_vsync` cvar changes was implented by
polling, which was ok for one cvar but I'll be making more vid_ cvars
apply immediately (no vid_restart needed), and don't want to poll for
things that rarely change every frame.

SDL is now queried for the current state of vsync instead of storing
what we believe to be the current state.

If the setting can't be applied the reason (as reported by SDL) is now
printed.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

vid: remove obsolete cvars

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

vid: misc cleanups

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Save cvars cl_rollangle and cl_rollspeed to config.cfg

Closes https://gitlab.com/xonotic/darkplaces/-/issues/393

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Merge MR 'off-switch cvar for dlcache unload on server/map change'

See https://gitlab.com/xonotic/darkplaces/-/merge_requests/141

off-switch cvar for dlcache unload on server/map change

Fix GCC 13 warns: ‘new_velocity[0]’ may be used uninitialized

Pretty sure this was a false positive, for it to happen `numplanes`
would have to be <= 0 in the loop which sets `new_velocity` which should
be impossible.  Declaring `numplanes` as unsigned placates the compiler,
and these ints should be unsigned anyway because they index an array.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Work around black texture bug on AMD's Windows driver

Closes https://gitlab.com/xonotic/darkplaces/-/issues/228

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Sys_Error(): don't leave a dead window covering the OS UI or crash dialog

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

VID_Shutdown(): Delete OpenGL context

Destroying the window doesn't do this so for a cleaner vid_restart we need to.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

VID_InitModeGL(): fix clear to black of new window

The window was left with whatever was already in the framebuffer.
Also we can perform the clear a bit earlier.

Changes initial state of vid_activewindow to match the logic that sets
it during normal frames.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Implement OpenGL version checking, use it for occlusion queries

This allows DP to use core features added after 3.2 without crashing,
in this case query buffer objects.

In div0-stable GL_ARB_query_buffer_object was checked but that extension
was written against 4.3 and became part of core in 4.4 so we can just
check for 4.4.

Fixes https://gitlab.com/xonotic/darkplaces/-/issues/176

Moves GL function pointer initialisation to its own function to support
a workaround for NVIDIA's opinion about GL profiles.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Cleanup some legacy cruft in vid_sdl.c

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Move QC extension reporting out of video initialisation

Once at startup is enough, no need to do this during vid_restart (etc).

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Merge PR 'Small fixes (indices out of bounds)'

See https://github.com/DarkPlacesEngine/darkplaces/pull/92

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Merge PR 'Fix some UB'

See https://github.com/DarkPlacesEngine/darkplaces/pull/89

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix yet more UB. Provide explanatory comments.

Signed-off-by: notcancername <notcancername@protomail.com>

Delete (broken) cmake submodule that somehow got in from my cmake branch

Signed-off-by: Cloudwalk <cloudwalk@icculus.org>

host: Move host cvar exports to host.h that don't belong in darkplaces.h

Signed-off-by: Cloudwalk <cloudwalk@icculus.org>

.gitignore: Exclude *.obj and *.tlog. Remove build artifacts

Signed-off-by: Cloudwalk <cloudwalk@icculus.org>

Merge PR 'Fix a bug on FreeBSD caused by assuming that short reads are errors'

This also enables files larger than 2GiB to be accessed on all platforms.

See https://github.com/DarkPlacesEngine/darkplaces/pull/88

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix a bug on FreeBSD caused by assuming that short reads are errors

Signed-off-by: notcancername <119271574+notcancername@users.noreply.github.com>

cvar: reset colours after printing values, clarify virtual/alias message

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Make `name` be a command or virtual cvar on a game-specific basis

This allows matching the client console behaviour of div0-stable in
games that want it.

Closes https://gitlab.com/xonotic/darkplaces/-/issues/337

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cmd: inline Arg helper funcs

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cbuf: catch runaway command loops, respects -norunaway cmdline arg

Prevents DP from spinlocking permanently.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cbuf: prepend keybind commands instead of appending to skip waits

This prevents user inputs being delayed by `wait` commands that other
sources may have added to the buffer.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cbuf: fix a 1 frame delay of keybind inputs

This was a regression (from div0-stable) introduced in the refactoring.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cbuf: use standard parsing and buffering of deferred strings

Fixes https://gitlab.com/xonotic/darkplaces/-/issues/377

Matches div0-stable in that the deferred string is not parsed until
execution time and is printed verbatim in the pending command list.

Also fixes unsafe linked list removal and removes its workaround.

Slightly improves timing precsion and PRVM_64 / 32 support.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

cbuf: refactor and fix parsing of text blocks into the linked list

Simplifies the parsing to a single loop with less state tracking.
Reorganises the code for readability.

Fixes https://gitlab.com/xonotic/darkplaces/-/issues/378

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix occasional NaN sdev in server timing reports

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Add some physics-related comments

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

SV_FlyMove(): defer SV_Impact() call until after ground detection

Quake didn't use SV_PushEntity in SV_FlyMove() and id1 fiends depend on
the original behaviour of setting ONGROUND (etc) before calling
SV_Impact().
Fixes https://github.com/DarkPlacesEngine/darkplaces/issues/94

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Rework handling of stuck entities in engine movetypes

Avoids most false positives and unnecessary work.
Mentions likely code path and cvar deprecations.

Fixes nades ending up where they shouldn't, which looked like:
Unstuck entity 450 (classname "ogre_grenade") with offset 0.000000 0.000000 28.000000.

Allows certain entities to move while stuck in the world (see comments).
Fixes https://github.com/DarkPlacesEngine/darkplaces/issues/18

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Add missing DP_QC_FINDCHAIN_TOFIELD declarations to CSQC defs

Fix potential index out of bounds

Fix potential index out of bounds

Fix mod_q1bsp_zero_hullsize_cutoff only applying to HLBSP in some cases

In 0dfd32e1e2bb2c1bd5e58b3202271c391be423de in one code path this
applied to Q1BSP and HLBSP, but the other code path was slightly
different and there it only applied to HLBSP.  Oops :)

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix some UB

caused by shifting such that the value becomes unrepresentable and
overflow respectively

Improve server lag detection and reporting

On a dedicated server, significant lag is reported to players (so they
know it wasn't the network or their machine).

Cvar sv_lagreporting_strict logs even slight lag that was "recovered" to
console.

sv_lagreporting_always enables these for empty servers or singleplayer.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Deduct frame execution time from sleep time

Fixes failure to sustain cl_maxfps/sys_ticrate when the system can
exceed it if unlimited, should improve frame pacing too.

Includes related refactoring. Actually this commit is mostly refactoring.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Properly reset server timers and perf stats after level change

This prevents the server running "catchup" physics frames immediately
after level change (when sv_maxphysicsframesperserverframe > 1) and
makes accurate stats available almost immediately (all 0 before that).

Moves perf stats from persistent to temporary storage.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix host.framecount

It never incremented if sleep was called, and could have eventually
become negative.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

curl: use much less conservative performance defaults

Matches defaults used in Xonotic >= 0.8.2

Dedicated server throughput is still limited by ticrate to avoid a
reduction in frame pacing precision, see 025fbac31da7f654e296e20b3037b5ce88074f8b

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix flags and naming of cvars needed by the dedicated server

The cl_curl* cvars also apply to the dedicated server so not
registering them in 3a58ad52e63635c93e95955f984fd6052156e56b broke HTTP.
These now have aliases for backwards compat (eg Xonotic, Nexuiz menus).

cl_maxphysicsframesperserverframe is server-specific and was also used
by the dedicated server for "catch up" if it got behind.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

strcat(): update QC defs to match implementation

Xonotic was defining strcat1n to allow single-arg cases without
boilerplate...

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

checkextension(): check libcurl is loaded before returning it's available

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Fix flags and naming of Q3BSP curve collision cvars, related tidying

r_subdivisions_collision* were marked as client-only, so not registering
them in the dedicated server in 3a58ad52e63635c93e95955f984fd6052156e56b
broke Q3BSP curve collisions.  Renamed to mod_q3bsp_curves_subdivisions*
for consistency.

Removed 3 obsolete mod_q3bsp cvars, moved r_fxaa to gl_rmain.c

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Add some missing csprogs defs

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

mod() builtin: match gmqcc and fteqw

Also catch div0, he escaped again

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Add cvar mod_q1bsp_zero_hullsize_cutoff for games using "big" projectiles on Q1BSP

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Merge MR 'Fix inability to access non-ascii path under Windows with WTF-8'

See https://gitlab.com/xonotic/darkplaces/-/merge_requests/142

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

dedicated server: don't call CL_UpdateScreen() and SCR_BeginLoadingPlaque()

dedicated server: don't register client-only cvars

Fixes random client cvars like r_trippy and cl_netpacketloss_send
appearing in the server.

Adds a warn (instead of crashing) for if we try to set an unregistered
engine cvar somewhere.

mod_bsp_portalize: remove from dedicated server, update description

sv_cullentities_portal was removed in
d3b158411b1d81181e05f750d9a7b0b2268438e2 so the server has no use for it
now.

Merge MR 'Implement DP_QC_NUDGEOUTOFSOLID extension, fix Quake item bugs' into 'master'

See merge request https://gitlab.com/xonotic/darkplaces/-/merge_requests/144

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Use qbool for some trace_s members, other tidying

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Implement DP_QC_NUDGEOUTOFSOLID extension

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Q1BSP: Fix TraceBox failing to move a bbox out of solid

Fixes https://github.com/DarkPlacesEngine/darkplaces/issues/85

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

droptofloor(): refactor and use correct conditions for each BSP format

Behaviour changes:

Although startsolid is used by Q3 and Q2, it's too strict for Quake
which used allsolid, so we now use the correct conditions for each BSP
format.

The downtrace endpoint is now calculated after
sv_gameplayfix_droptofloorstartsolid_nudgetocorrect.

The downtrace endpoint is now offset when using the
sv_gameplayfix_droptofloorstartsolid fallback path, otherwise it would
trace a diagonal line for entities whose origin is not horizontally centred
in their bbox.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

droptofloor(): fix collisions between entities, including on id1 maps

The presence of a monster bbox shouldn't cause a health to be considered
"in solid".

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Add prefix to VM_Warning indicating which VM it came from

A few messages include this but many don't, and in future there will be
less functions with SV or CL in their name.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

droptofloor(): add/improve warnings for badly placed entities

In the past, sv_gameplayfix_droptofloorstartsolid and
sv_gameplayfix_droptofloorstartsolid_nudgetocorrect were enabled by
default when playing Quake, which seems to have led to some maps
depending on them.

These need to be off by default in Quake to avoid gameplay changes,
but we can at least warn people about map entity issues.

See: 34e819a626f6a7ff2ad190a8e7d42e38ca8cef0e

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

NudgeOutOfSolid: fix "random" failures on slightly inclined planes

History:
This epsilon value was first added to the nudge distance in
31e7715cffcebfd400b8473f3b2d3dc80937fc5a
and was replaced with bbox expansion in
25c09ca585e81bf9eb236a339ae5dcd64f63b679

With the expansion approach we need to compare to separation to make
"good location" detection reliable, partly to avoid float precision
problems, partly to avoid compatibility problems with normal TraceBox.
For example: we drop an entity to floor successfully with TraceBox, but
then NudgeOutOfSolid decides it's in solid.

Also updated the startsolid conditions to suit the 2-pass approach.

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

NudgeOutOfSolid: add a wasn't-stuck result

It can be handy for calling code to know if the entity was initially
stuck (mainly for warnings).

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

Move SV_NudgeOutOfSolid() into a common namespace and unit

Signed-off-by: bones_was_here <bones_was_here@xonotic.au>

(wtf8) wrap 4-bytes-sequence detection in ifdef U32

Signed-off-by: NaitLee <naitli@foxmail.com>

rename and add comments for wtf8 macro flags

Signed-off-by: NaitLee <naitli@foxmail.com>

use tabs for wtf8 code indent

Signed-off-by: NaitLee <naitli@foxmail.com>

(non-ascii path) fix an Win32 API annotation, fix "mygames" folder

Signed-off-by: NaitLee <naitli@foxmail.com>