This fixes a bug where using the fullscreen toggle keybind (Alt+Enter,
Alt+F, or F11) wouldn't update the color of the "resize to nearest" menu
option. The color doesn't functionally change anything - the option
still won't work, and will still have the message telling you that you
need to be in windowed mode when you move your menu selection to it -
but it's an easy inconsistency to fix; just move the menu recreation in
to Screen::toggleFullScreen() itself.
The game dereferences graphics.screenbuffer without checking it first...
it's unlikely to happen, but the least we can to do be safe is to add a
check and assert here.
If there were two scripts with the same name, removing one of them would
only remove the other script from the script name list, and not also
remove the contents of said script - leading to a desync in state, which
is probably bad.
Fixing this isn't as simple as removing the break statement - I either
also have to decrement the loop variable when removing the script, or
iterate backwards. I chose to iterate backwards here because it
relocates less memory than iterating forwards.
No need to use it when good ol' loops work just fine.
Iterating backwards is correct here, in case there happen to be more
than one of the item in the vectors, and also to minimize the amount of
memory that needs to be relocated.
This is a simple change - we draw minimap.png, instead of the generated
custom map, if it is a per-level mounted custom asset.
Custom levels have already been able to utilize minimap.png, but it was
limited - they could do gamemode(teleporter) in a script, and that would
show their customized minimap.png, but it's not like the player could
look at it during gameplay.
I would have done this earlier if I had figured out how to check if a
specific asset was mounted or not.
Previously, if the game couldn't set the write dir to the base
directory, or couldn't make the base directory, or couldn't calculate
the base directory, it would probably dereference NULL or read from
uninitialized memory or murder your family or something. But now, I've
eliminated the potential Undefined Behavior from the code dealing with
the base path.
Previously, this function had a bug due to failing to account for array
decay. My solution was to just repeat the MAX_PATH again. But in
hindsight I realize that's bad because it hardcodes it, and introduces
the opportunity for an error where we update the size of the original
path but not the size in the function.
So instead, just pass the size through to the function.
I don't want to add too many asserts, because sometimes it's okay if a
file is missing (mmmmmm.vvv). But currently, the game basically expects
all images and sound effects to be present. That might change in the
future, but for now, these asserts are okay.
FILESYSTEM_loadFileToMemory() dereferenced pointers without checking if
they were valid... I don't know of any cases where they could have been
NULL, but better safe than sorry.
So, the codebase was kind of undecided about who is responsible for
initializing the parameters passed to FILESYSTEM_loadFileToMemory() - is
it the caller? Is it FILESYSTEM_loadFileToMemory()? Sometimes callers
would initialize one variable but not the other, and it was always a
toss-up whether or not FILESYSTEM_loadFileToMemory() would end up
initializing everything in the end.
All of this is to say that the game dereferences an uninitialized
pointer if it can't load a sound effect. Which is bad. Now, I could
either fix that single case, or fix every case. Judging by the title of
this commit, you can infer that I decided to fix every case - fixing
every case means not just all cases that currently exist (which, as far
as I know, is only the sound effect one), but all cases that could exist
in the future.
So, FILESYSTEM_loadFileToMemory() is now guaranteed to initialize its
parameters even if the file fails to be loaded. This is better than
passing the responsibility to the caller anyway, because if the caller
initialized it, then that would be wasted work if the file succeeds
anyway because FILESYSTEM_loadFileToMemory() will overwrite it, and if
the file fails to load, well that's when the variables get initialized
anyway.
My next commit will involve using goto to jump to the end of a function
to initialize the variables to NULL, but that results in a compiler
error if we have initializations in the middle of the function. We might
as well put all declarations at the top of each block anyway, to help
the move to C, so I'm doing this now.
Since the length variable in the STDIN block now overshadows the length
variable in the outer block, I've renamed the length variable in the
block to stdin_length.
These casts are sprinkled all throughout the graphics code when creating
and initializing an SDL_Rect on the same line. Unfortunately, most of
these are unnecessary, and at worst are wasteful because they result in
narrowing a 4-byte integer into a 2-byte one when they don't need to
(SDL_Rects are made up of 4-byte integers).
Now, removing them reveals why they were placed there in the first place
- a warning is raised (-Wnarrowing) that implicit narrowing conversions
are prohibited in initializer lists in C++11. (Notably, if the
conversion wasn't narrowing, or implicit, or done in an initializer
list, it would be fine. This is a really specific prohibition that
doesn't apply if any of its sub-cases are true.)
We don't use C++11, but this warning can be easily vanquished by a
simple explicit cast to int (similar to the error of implicitly
converting void* to any other pointer in C++, which works just fine in
C), and we only need to do it when the warning is raised (not every
single time we make an SDL_Rect), so there we go.
This fixes a bug where after loading in to the level editor, pressing
Esc and then switching your option to something other than the first
option, then pressing Esc again to close the menu, then pressing Esc
once more would not keep your menu option.
This is because the code that checks if Menu::ed_settings is already in
the stack doesn't account for if ed_settings is the current menu - the
current menu doesn't get put in to the stack.
In hindsight, maybe I could have designed the new menu system better so
the current menu IS on the stack, and/or should have used a
statically-allocated linked list for each menu name for the stack frames
(instead of an std::vector) and asserted if a menu that already existed
in the stack was created instead... that'll have to be done later,
though.
Pressing Esc to cancel the confirm quit menu didn't play the squeak, in
contrast to pressing ACTION to cancel it, so now it does; pressing Esc
to close the pause menu or pressing ACTION will also now play the
Viridian squeak too.
vx/vy mean x-velocity and y-velocity... except here, where it seems like
they're used as extra parameters that do different things depending on
the entity. But it seems like at one point they were actually meant to
be the speed of the entity (this is the case for the unused decorative
particle entities), and then just never got renamed when they weren't.
The custom levels community named these two parameters meta1 and meta2
in the reference list of entities for the createentity() script command,
so that's what I'm naming them here. This will avoid confusion (I know
that some people reading this function have genuinely mistaken the vx/vy
for actually meaning x-velocity and y-velocity, simply because they were
named that way).
I have spelled out each overloaded version instead, and only the
overloads that are actually used - which just happens to be everything
except the 8-argument one. I don't want to deal with callers right now
(there are too many of them), so I'm not going to change the names that
the callers use, nor do I want to change the amount of arguments any
existing callers use right now - but we will have to deal with them in
one way or another when we move to C.
The script command createentity() is always an int. But not only that,
every time createentity() is used, its arguments are always treated like
ints. Always. I knew that vx/vy were floats because of the int casts in
the function, but I didn't even realize that xp/yp were floats, too,
until I checked just now! That's how much they're treated like ints.
All int casts in createentity() have also been removed, due to being
unnecessary (either because of us suppressing MSVC implicit conversion
warnings, or because there are now no longer any conversions happening).
This boolean is assigned, and it is checked... but it's never assigned
to true, thus making it useless. I also checked 2.2 source and the same
thing happens there; to prevent any confusion, I'm removing this.
So... I did see that map.ypos was a float when I added over-30-FPS mode,
because map.oldypos wasn't there before... I'm guessing that I kind of
just ignored it at the time. But, c'mon, map.ypos and map.oldypos are
always treated as ints, so there's literally no reason for them to be
actually floats in reality. I didn't even know they were anything other
than ints until I checked Map.h.
This is quite simple - whenever the user uses their keyboard or
controller, we hide the mouse cursor. Whenever they move the mouse, we
show it again. This makes it so the cursor gets out of the way when they
play the game, but reappears when they need it.
There is also a timeout, to prevent strobing if the user decides to use
the keyboard/controller and mouse at the same time. There is no timeout
from hiding the mouse cursor, but there is a timeout from showing the
mouse cursor - this because it's okay if the mouse lingers for a few
frames when you start using the keyboard, but really annoying if the
mouse doesn't instantly appear when you move it.
The config option has been removed. I'm going to implement something
that automatically shows and hides the mouse cursor whenever
appropriate, which is better than a config option.
These are two C++ features that we don't need, don't use, and will never
use in the future. Apparently the best way of doing this in CMake is to
fiddle with the CXX_FLAGS using regex.
Now this is one less flag I need to supply myself when I invoke CMake...
This variable is not defined anywhere and never has been since the
source code release (which is when this CMake file was first created).
To make things clearer, I'm cleaning this variable up.
A function like add_definitions() adds definitions to ALL targets, not
just VVVVVV. This kind of namespace pollution is messy, and could result
in bugs if you pollute with the right kind of pollutant.
So instead of using add_definitions(), use target_compile_definitions().
And instead of using include_directories(), use
target_include_directories().
All the C third-party dependencies are C90, and all the C files we have
are also C90 (well, almost, but that's easily sorted). So we have
basically no reason to not go with C90 here.
The only wrinkle is, turning C extensions off for physfs-static results
in linker errors because PhysFS implicitly uses alloca() without
including it properly (on Linux). I am not the only one who has ran into
this - see https://icculus.org/pipermail/physfs/2020-April/001293.html -
and it's a bug with PhysFS. The workaround I've gone with is to enable C
extensions. (There might also be some funkiness with PhysFS's use of the
`inline` keyword, so enabling extensions will paper over that as well.)
So there were actually only two instances of C99-style end-of-line
comments in C files - and technically one of them was just a C file
including MakeAndPlay.h.
It seems like CMake 3.1.3 introduced the C/C++ standard properties,
while the minimum version of this CMake file is 2.8.12. So we do what
FAudio does, which is print a warning if the CMake version is too old
and otherwise use it if we have the feature.
They're the same thing, but using option() better conveys intent.
However this can't be done for anything that isn't a bool, which the
CUSTOM_LEVEL_SUPPORT option is not (it's a tri-state string).
These were introduced in 098fb77611 - did
Leo not know that they were already there at the top of the file? This
does the same thing, except it only sets it for VVVVVV instead of
everything (so this wouldn't set it for the third-party dependencies).
If a track was restarted after it faded out, then it wouldn't play. This
is because currentsong wasn't set to -1 after fading out, and that is
because the fade out calls pause() instead of haltdasmusik() when it
finishes.
Unlike f196fcd896, this fixes the time
trial music while keeping it to the same behavior as 2.2, and fixes
every single possible case that this music bug could have happened.
This reverts only a part of f196fcd896 -
as the original commit author did not do their changes atomically, they
also squashed in a de-duplication within the same commit. So I'm only
reverting the part of the commit that wasn't the de-duplication, which
is simply the changes to the music.fadeout() calls.
This is being (partially) reverted for several reasons:
1. It's not the correct behavior. What this does instead is persist the
track through after you restart the time trial, instead of fading it
out, then restarting it again. This is in contrast to behavior in
2.2, and I see no reason to not keep the same behavior.
2. It's a single-case patch. The time trials are not the only time in
the game a music track could fade out and then be restarted with the
same track - custom levels could do the same thing too. Instead of
fixing only one case, we should strive to fix EVERY case.
The original commit author (trelbutate) also didn't write anything in
the commit description of f196fcd896. What
you should write in the commit description is things like rationale,
analysis, and other good information that would be useful to anyone
looking at your commit to understand why you did what you did. Having no
commit description leaves readers in the dark as to why you did what you
did.
Thus, I don't know why trelbutate went with this solution, or if they
knew that it was only a single-case patching, or if they knew that it
wasn't 2.2 behavior.
By not writing the commit description, they miss a chance for
reflection; speaking from personal experience, I myself have gone back
and improved my commits countless times because I wrote commit
descriptions for every single one of them, and sometimes whenever I
write them, I think to myself "hang on a minute, that doesn't sound
quite right" and end up finding improvements.
If trelbutate wrote a commit description, they might have realized that
it wasn't 2.2 behavior, and gone back and fixed up their commit to be
correct. As it stands, though, they didn't have to think about it in the
first place because they never bothered to write a commit description.
edteleportent is a global variable that gets assigned whenever the
player collides with a warp token, and gets read from later down the
line in gamelogic(). While I don't know of any way to cause anything bad
with this (and I did try), storing a temporary indexing variable like
this is only bound to be a liability in the future - so we might as well
prevent badness now by adding a bounds check here.
This fixes a bug where quitting to the menu from command-line
playtesting with -playassets specified would always use those assets
when loading back in to any custom level. This also fixes loading in to
a custom level quicksave always using the command-line playtesting
arguments instead of using the actual quicksave.
In a vertically-warping room, the 'height' of the room becomes 232
pixels, regardless of if you have a room name or not. So the remaining 8
rows of pixels at the bottom of the screen corresponds with the first 8
rows of pixels at the top of the screen, and entities in the bottom 8
rows of pixels get teleported to the top of the screen.
The screen wrapping drawing code doesn't draw entities in the top 8 rows
of pixels at the bottom, leading to a discontinuous effect where it
looks like vertically-warping entities don't neatly change from the
bottom to the top or vice versa - this is especially noticeable with
enemies. To fix this, just increase the threshold for drawing top
entities at the bottom of the screen by 8 pixels.
When an entity vertically warps, it teleports upwards or downwards by
232 pixels. However, the graphics code draws them with an offset of 230
pixels. This is off by 2 pixels, but it's enough to make a
downwards-moving enemy look like it suddenly collides with the bottom of
the screen (in a room without a room name) before it warps, especially
if you go frame-by-frame.
It seems like for whatever reason that the frames portion of save files
is never read from, and always zeroed. Well, technically they get parsed
but the result is immediately discarded afterwards.
I see no reason to do this, so I'm removing these zeroes.
This fixes being able to make music fully fade in (or out) by unfocusing
the game, or making the fade bars fully fade in (or out) by unfocusing
the game, or racking up the timer while the game is unfocused.
In 2.2 and previous, the game would call resetgameclock() every frame
for the last 30 frames of the time trial countdown in order to make sure
it gets reset. This was in a render function, and didn't get brought out
in 2.3, so 2.3 resets the game clock *while rendering*, which is kinda
bad and is an oversight on my part for not noticing.
Instead of doing that, just add a conditional to the timer so that it
won't tick during the time trial countdown. This fixes#699 even further
by making it so the time trial par can't even be lost during the
countdown, because the timer won't tick up - so you can never get a sad
squeak to play by pausing the game or unfocus-pausing it during the
countdown.
For some reason, resetgameclock() is only ever used in gamerender(), and
everywhere else just zeroes the clock manually. This is weird to me, so
I've made it so everywhere that zeroes the clock uses the
resetgameclock() function to do so.
Otherwise, if the timer ticked up past the par (via using the unfocus
pause or pause menu), it would result in the sad squeak being played
every frame because the game would constantly be setting
timetrialparlost, then moving to the code block below, assuming that
since timetrialparlost that we haven't lost the par already, and playing
the squeak.
timetrialparlost gets reset in hardreset() and startgamemode() anyways,
so there's no need to be constantly resetting this variable.
Fixes#699.
It turns out this entire chunk of code is simply unneeded (and is
actively harmful) since when we're done with the time trial,
quittomenu() gets called, and that removes the previous stack frame
anyway.
I'm guessing that I added this code, then added quittomenu(), then
didn't consider how this code and quittomenu() would mix. But anyways,
this bug is fixed.
Fixes#714.
This seems to be a comment left by Ethan that he never got around to. So
I did it for him.
What I've done is made it so FileSystemUtils.cpp knows what a binary
blob is, and moved the binary blob loading code directly to
FileSystemUtils.cpp. To do this, I removed the private access modifier
from binaryBlob - I don't think we'll need it, and anyways when we move
to C we can't use it.
Along the way, I also cleaned up the style of the function a bit - the
null termination offset is no longer hardcoded, and the function no
longer mixes code and declarations together in the same block.
I also noticed that when printing all the filenames at the end, a single
invalid header would stop the whole loop instead of just being skipped
over... this seems to be a bug to me, so I've made it so invalid headers
just get skipped over instead of stopping the whole loop.
In FileSystemUtils.h, I used a forward declaration. In hindsight,
incomplete forward declarations should basically always be done in
header files if possible, otherwise this introduces the possibility of
transitive includes - if a file includes this header and it does a full
include, the file is silently able to use the full header, whereas if
it's a forward declaration, then the moment the file tries to use the
full header it fails, and then it's forced to include the full header
for itself. But uh, that's a code cleanup for later.
While fixing all the other music bugs, I discovered that starting
playtesting in the editor wouldn't play the level music.
The problem is that the editor playtesting start code calls
music.fadeout() before calling music.play(). This queues up the track
from the music.play() call. After that, what should happen is that
processmusic() processes the fade, the fade is then finished, and then
after that it sees that the music is halted so it can play the queued
track.
Instead what happens is that the function first attempts to play the
music before the fade is processed and finished, so play() will re-queue
the music again, but the queue gets cleared right after that (this is a
subtle bit of behavior - it means if the game fails to play a queued
track due to it fading, it's not going to re-queue it again and end up
in some sort of infinite loop).
This is a frame ordering issue - the function is tripping over itself
when it shouldn't be. To fix it, just put the queue processing code
after the fade processing code.
This fixes the 2.2-and-below music blocking workaround not working in
2.3.
The issue was that when the music got halted by the script, the fade
volume would still be processing, silently being decremented in the
background. So the script playing the track afterwards would make the
game queue it (as it was called during the fade), but then the music is
halted so the game would attempt to play it, but the fade is STILL
happening so it wouldn't actually play it and would attempt to queue the
track again.
However, that queue gets discarded immediately afterwards because the
music.play() call happened inside the code responsible for playing the
queued music, and that code unconditionally clears the queue variables
immediately after calling play(). So that's good to know - if the game
queues a song, but fails to play it because of a fade, it's not going to
immediately re-queue it and potentially get stuck in a loop of
infinitely queueing the same song over and over again each frame.
Anyways, the source of the problem is not resetting the fade booleans
when halting music, so I've reset them.
Fixes#701.
The problem here is that even though we start playing the music when the
volume is set to zero, mixer's state doesn't have volume zero, so
whatever it plays next will be the very first quanta of the track but at
the previous volume (in this case, the maximum volume). To fix this,
just update mixer when we update the volume here - it's okay to not
account for user volume because it ends up being zero anyway.
Fixes#710.
This fixes a bug where fading music in but not going through the
music.play() path wouldn't start the fade volume from zero. If this
happened, then the previous volume would persist, and if the previous
volume was the max volume, then that essentially canceled out the
fade-in and prevented it from happening at all. But now all paths to
fadeMusicVolumeIn() set the volume to zero first, instead of only the
caller of music.play().
When you pick up a trinket in the wild, the music gets silenced, so it
silently plays in the background until you advance the trinket text.
However, foundtrinket (used when Victoria or Vitellary give you a
trinket) is inconsistent with this, and halts the music instead of
silencing it.
This was probably due to the musicfadein script command not being
implemented, so Terry or Simon had to simply make do and halt the music
instead. But musicfadein is implemented and is being used in the trinket
cutscenes, so this is another inconsistency that I will fix.
When you pick up a trinket in the wild, the music will fade back in
afterwards. However, the special trinket cutscenes (where Victoria or
Vitellary will directly give you a trinket) are inconsistent with this,
and restart the music instead of fading it back in.
Looking at the scripts themselves, it immediately becomes obvious the
reason for this inconsistency - 2.2 and previous didn't implement the
musicfadein command, so it couldn't be used, and Terry or Simon simply
had to make do with simply restarting the music. However, 2.3 implements
musicfadein, so we can simply swap it out and remove the
trinketscriptmusic command.
This is 2.2 behavior, which I forgot to keep. Otherwise, if music has
halted and you try to play the same track, it simply won't work, because
the current song is the same as the song you're trying to play. This is
what happened with the trinket scripts - the game halted music, then
tried to play the same track.
Fixes#712.
It's not really used because CreateDirectory doesn't support setting
chmod values, but it does clarify intent of the argument.
Co-authored-by: Ethan Lee <flibitijibibo@gmail.com>
In #52 I fixed VVVVVV not being able to handle filepaths with non-ASCII
characters on Windows. 2f0a0bce4c and
aa5c2d9dc2 reintroduce this problem,
however, by reverting the definition of mkdir to how it was before the
fix and using the non-Unicode version of CreateDirectory. And I can
confirm that VVVVVV indeed doesn't make its folder anymore with a
Windows username of "тест". This commit fixes that issue.
This adds music and volume sliders to the audio options. To use the
sliders, you navigate to the given option, then press ACTION, and your
selection will be transferred to the slider. Pressing left or right will
move the slider accordingly. Then you can press ACTION to confirm the
volume is what you want and deselect it, or you can press Esc to cancel
the volume change, and it will revert to the previous volume; both
actions will write your settings to disk.
Most of this commit is just adding infrastructure to support having
sliders in menus (without copy-pasting code), which is a totally
completely new user interface that has never been used before in this
game. If we're going to be adding something new, I want to make sure
that it at least is done the RIGHT way.
Closes#706.
This adds <musicvolume> and <soundvolume> tags to unlock.vvv and
settings.vvv, so users' volume preferences will be persistent across
game sessions. This does not add the user interface to change them from
in-game; the next commit will do that.
This function is simple - it takes a given buffer and its size, fills it
with a certain character, and null-terminates it. It's meant to be used
with freshly-created buffers, so we don't copy-paste code.
Pressing return in gameplay options would send you back to the pause
menu instead of the general options menu, and pressing return in graphic
options would send you back to the pause menu instead of the general
options menu, too. Additionally, pressing Esc in graphic options would
also send you back to the pause menu instead of the general options
menu.
Like I said before, the menu system is still a bit hardcoded in some
places, and these happened because Terry forgot to update them when he
changed the menus around.
Fixes#711.
The in-game menu code is better than it was in 2.2 but still pretty
hardcoded, so to fix this just change each individual case around. This
bug happened because the "options" button was in the place where "quit
to menu" was previously, but Terry forgot to update it when changing all
the options around.
PhysFS requires a write dir to create a directory, so the first PHYSFS_mkdir
never could have worked. Because of that we need to go back to the old mkdir,
and since we're bringing that back we can reuse it for saves/levels, because we
know it works and we don't have to worry about middlewares ruining anything.
When a text box in the script system (not the gamestate system) is
displayed onscreen and "- Press ACTION to advance text -" is up, the
game sets pausescript to true, so the script system won't blare past the
text box and keep executing. Then it also sets advancetext to true.
Crucially, these two variables are different, so if you have pausescript
true but advancetext false, then what happens?
Well, you get softlocked. There's no way to continue the script.
How is this possible? Well, you can teleport to the (0,0) teleporter
(the teleporter in the very top-left of the map) and regain control
during the teleporter animation. To do that, in 2.2 and below, you have
to press R at the same time you press Enter on the teleporter, or in 2.3
you can simply press R during the cutscene. Then once you teleport to
the room, it's really precise and a bit difficult (especially if
Viridian is invisible), but you can quickly walk over to the terminal in
that room and press Enter on it.
Then what will happen is the terminal script will run, but the
teleporter gamestate sequence will finish and turn advancetext off in
the middle of it. And then you're softlocked.
To fix this, just add a check so if we're in gamestate 0 and there's a
script running, but we have pausescript on and advancetext off, just
turn pausescript off so the game automatically advances the script.
This softlock was reported by Tzann on the VVVVVV speedrunning Discord.
If you manage to get softlocked by being stuck in completestop, the next
thing you'll notice is that quitting to the menu or loading back in will
not reset this.
So you can actually softlock yourself in 2.3 by doing the trinket
cutscene, then quitting to the menu (because 2.3 lets you open the pause
menu during completestop). This is a bug, and should be fixed.
You can skip the "You have found a shiny trinket!" cutscene. The
conditions are that this can only be done in the main game, in the main
dimension (no Polar Dimension), the checkpoint that you last touched
must not be in the same room as the trinket, and you have to have
skipped the Comms Relay cutscene. To do the skip, you press R on the
exact frame (or previous frame, if input delay is enabled) that Viridian
touches the trinket. Then, the gamestate will be immediately set to 0
(because of the gotoroom) and the cutscene will be skipped.
Speedrunners of the main game, well, run the main game already, the
only trinket in the Polar Dimension is not one you want to do a death
warp at, and they have a habit of automatically skipping over the Comms
Relay cutscene because they press R at the beginning of the run when
Viridian teleports to Welcome Aboard, to warp back to the Ship and so
they can leave rescuing Violet for later.
So someone reported softlocking themselves by doing the trinket text
skip in 2.3. The softlock is because they're stuck in a state where
completestop is true but can't advance to a state that turns it off. How
does this happen? It's because they pressed R too late and interrupted
the gamestate sequence. In 2.2 and previous, if you're in the gamestate
sequence then you can't press R at all, but 2.3 removes this restriction
(on account of aiming to prevent softlocks). So only on the very first
frame can you death warp and interrupt the gamestate sequence before it
happens at all.
Anyways to fix this, just turn completestop off automatically if we're
in gamestate 0 and there's no script running.
This softlock was reported by Euni on the VVVVVV speedrunning Discord.
So some people reported the levels list crashing when they loaded it.
But this wasn't reproducible every time. They didn't provide any
debugging information, so I had to use my backup plan: doing a full
audit of the code path taken for loading the levels list.
And then I found this. It turns out this was because I used a
LOAD_ARRAY_RENAME() macro on an std::vector. You can't do that because
you need to use push_back() to resize a vector, so the macro will end up
indexing into nothing, causing a segfault. However, this code path would
only be taken if you have an old levelstats.vvv, from 2.2 and previous -
which explains why it wasn't 100% reproducible. But now that I know you
need an old levelstats.vvv, this bug happens 100% of the time.
Anyways, to fix this, just ditch the macro and expand it manually, while
replacing the indexing with a proper usage of push_back().
While the game does support playing levels with filenames that don't
have the .vvvvvv extension, it doesn't do it well.
Namely, those files can't be loaded or saved into the editor (because a
.vvvvvv always gets tacked on to your input when saving or loading). In
2.3, this gets worse because you can't load a level without a .vvvvvv
extension from command-line playtesting (because a .vvvvvv automatically
gets added) and you can't load per-level custom assets.
The only place where extensionless level files are supported is when
loading level metadata. But this makes it so they no longer work. This
is technically an API break, but it's easily fixed (just add the
.vvvvvv), plus there's nothing to be gained from not having an
extension, plus basically no one ever actually did this in the first
place (as far as I know, I am the only person to have ever done this,
and no one else ever has).
This fixes an issue where you would be able to mount things other than
custom assets in per-level custom asset directories and zips.
To be fair, the effects of this issue were fairly limited - about the
only thing I could do with it was to override a user-made quicksave of a
custom level with one of my own. However, since the quicksave check
happens before assets are mounted, if the user didn't have an existing
quicksave then they wouldn't be able load my quicksave. Furthermore,
mounting things like settings.vvv simply doesn't work because assets
only get mounted when the level gets loaded, but the game only reads
from settings.vvv on startup.
Still, this is an issue, and just because it only has one effect doesn't
mean we should single-case patch that one effect only. So what can we
do?
I was thinking that we should (1) mount custom assets in a dedicated
directory, and then from there (2) mount each specific asset directly -
namely, mount the graphics/ and sounds/ folders, and mount the
vvvvvvmusic.vvv and mmmmmm.vvv files. For (1), assets are now mounted at
a (non-existent) location named .vvv-mnt/assets/. However, (2) doesn't
fully work due to how PhysFS works.
What DOES work is being able to mount the graphics/ and sounds/ folders,
but only if the custom assets directory is a directory. And, you
actually have to use the real directory where those graphics/ and
sounds/ folders are located, and not the mounted directory, because
PHYSFS_mount() only accepts real directories. (In which case why bother
mounting the directory in the first place if we have to use real
directories anyway?) So already this seems like having different
directory and zip mounting paths, which I don't want...
I tried to unify the directory and zip paths and get around the real
directory limitation. So for mounting each individual asset (i.e.
graphics/, sounds/, but especially vvvvvvmusic.vvv and mmmmmm.vvv), I
tried doing PHYSFS_openRead() followed by PHYSFS_mountHandle() with that
PHYSFS_File, but this simply doesn't work, because PHYSFS_mountHandle()
will always create a PHYSFS_Io object, and pass it to a PhysFS internal
helper function named openDirectory() which will only attempt to treat
it as a directory if the PHYSFS_Io* passed is NULL. Since
PHYSFS_mountHandle() always passes a non-NULL PHYSFS_Io*,
openDirectory() will always treat it like a zip file and never as a
directory - in contrast, PHYSFS_mount() will always pass a NULL
PHYSFS_Io* to openDirectory(), so PHYSFS_mount() is the only function
that works for mounting directories.
(And even if this did work, having to keep the file open (because of the
PHYSFS_openRead()) results in the user being unable to touch the file on
Windows until it gets closed, which I also don't want.)
As for zip files, PHYSFS_mount() works just fine on them, but then we
run into the issue of accessing the individual assets inside it. As
covered above, PHYSFS_mount() only accepts real directories, so we can't
use it to access the assets inside, but then if we do the
PHYSFS_openRead() and PHYSFS_mountHandle() approach,
PHYSFS_mountHandle() will treat the assets inside as zip files instead
of just mounting them normally!
So in short, PhysFS only seems to be able to mount directories and zip
files, and not any loose individual files (like vvvvvvmusic.vvv and
mmmmmm.vvv). Furthermore, directories inside directories works, but
directories inside zip files doesn't (only zip files inside zip files
work).
It seems like our asset paths don't really work well with PhysFS's
design. Currently, graphics/, sounds/, vvvvvvmusic.vvv, and mmmmmm.vvv
all live at the root directory of the VVVVVV folder. But what would work
better is if all of those items were organized into a subfolder, for
example, a folder named assets/. So the previous assets mounting system
before this patch would just have mounted assets/ and be done with it,
and there would be no risk of mounting extraneous files that could do
bad things. However, due to our unorganized asset paths, the previous
system has to mount assets at the root of the VVVVVV folder, which
invites the possibility of those extraneous bad files being mounted.
Well, we can't change the asset paths now, that would be a pretty big
API break (maybe it should be a 2.4 thing). So what can we do?
What I've done is, after mounting the assets at .vvv-mnt/assets/, when
the game loads an asset, it checks if there's an override available
inside .vvv-mnt/assets/, and if so, the game will load that asset
instead of the regular one. This is basically reimplementing what PhysFS
SHOULD be able to do for us, but can't. This fixes the issue of being
able to mount a quicksave for a custom level inside its asset directory.
I should also note, the unorganized asset paths issue also means that
for .zip files (which contain the level file), the level file itself is
also technically mounted at .vvv-mnt/assets/. This is harmless (because
when we load a level file, we never load it as an asset) but it's still
a bit ugly. Changing the asset paths now seems more and more like a good
thing to do...
This will clarify which directory, exactly, failed to mount. I know it
gets printed earlier in the mounting process, but it can't hurt to print
it twice, just to be sure. Also this is for consistency.
Default function arguments are the devil, and it's better to be more
explicit about what you're passing into the function. Also because we
might become C-only in the future and to help faciliate that, we should
get rid of C++-isms like default function arguments now.
PHYSFS_getDirSeparator() already gets called and stored in pathSep at
the top of FILESYSTEM_init(). So clearly, two people worked on this
function and forgot that both pieces of code existed at the same time
(or it was one person independently forgetting both).
PhysFS uses platform-independent notation, so we really don't need to
care about getting the correct dir separator here. Especially since we
don't ever do so anywhere else (e.g. load/saveTiXml2Document()), either.
This is to make it clear that this is not a general-purpose mounting
function; it is a helper function for FILESYSTEM_mountAssets()
specifically for treating a directory or file as an assets directory,
and mounting assets from there.