2.2 and earlier had this god-awful thing where it put the closing tag of
an edentity onto the next line, and then kept the indentation the same.
This requires parsing the XML in an extremely specific way (i.e.
ignoring the whitespace) so the newline and indentation isn't taken as
part of the actual contents of the tag.
2.3 removed this awful whitespace entirely to make it easier on parsers.
When I tested #270, I tested against a 2.3 re-save of Dimension Open and
diffed the two, because I thought testing against the original version
of the level would result in a bunch of noise I didn't want due to the
whitespace change. Well, I did exactly what I intended, and ended up
ignoring the whitespace change so much that levels saved in this stupid
format ended up getting broken.
Luckily, we can just tell TinyXML-2 to parse a document exactly like how
TinyXML-1 would've parsed it, by supplying the COLLAPSE_WHITESPACE enum
to it (by default it's on PRESERVE_WHITESPACE).
This removes the TinyXML source files, removes it from CMakeLists.txt,
removes all the includes, and removes the functions
FILESYSTEM_saveTiXmlDocument() and FILESYSTEM_loadTiXmlDocument() (use
FILESYSTEM_saveTiXml2Document() and FILESYSTEM_loadTiXml2Document()
instead).
Additionally I've cleaned up the tinyxml2.h include in FileSystemUtils.h
so that it doesn't actually include tinyxml2.h unnecessarily, meaning a
change to TinyXML2 shouldn't rebuild all files that include
FileSystemUtils.h.
This commit refactors custom level scripts to no longer be stored in one
giant vector containing not only every single script name, but every
single script's contents as well. More specifically,
scriptclass::customscript has been converted to an std::vector<Script>
scriptclass::customscripts (note the extra S), and a Script is just a
struct with an std::string name and std::vector<std::string> contents.
This is an improvement in both performance and maintainability. The game
no longer has to look through script contents in case they're actually
script names, and then manually extract the script contents from there.
Instead, all it has to do is look for script names only. And the
contents are provided for free. This results in a performance gain.
Also, the old system resulted in lots of boilerplate everywhere anytime
scripts had to be handled or parsed. Now, the boilerplate is only done
when saving or loading a custom level. This makes code quality much,
much better.
To be sure I didn't actually change anything, I tested by first saving
Dimension Open in current 2.3 (because current 2.3 gets rid of the
awful edentity whitespace), and then resaved it on this patch. There is
absolutely no difference between the current-2.3-resave and
this-patch-resave.
Main game would retain custom level assets, now fixed. Also, custom fonts load properly. Finally, levels can be stored as a zip and placed in the levels folder, with the .vvvvvv file at the root of the zip and custom asset folders (graphics, sounds etc) also at the root.
Previously, the editor would always say it saved or loaded a level,
even if it was not successful. For example, because a file to load does
not exist, a file to save has illegal characters in its name or the
name is too long to be stored. Now failure is reported. Also, when
quitting the editor and saving before quitting is unsuccessful, the
editor will abort quitting.
I have the feeling that none of the devs understood what extern did, and
they kind of just sprinkled it everywhere until things started working.
But like all other classes, it should just be one line in the class's
respective header file, and shouldn't be so messy.
When the game loads a room in a custom level, previously it would load
the tilemap of that room into ed.swapmap, and then mapclass::loadlevel()
would manually go through each element in ed.swapmap to set each tile in
`contents`. Why do that, when you can just return the vector from
editorclass::loadlevel() and set it directly? ed.swapmap is really
unnecessary.
For some reason, the only way to get a cyan crewmate is by cycling
through an already-existing crewmate by keeping left-clicking on it.
This is because when you cycle through crewmate colors, the allowed
colors are 0-5, but when you place down a crewmate, it picks a random
color from 1-5, which seems to be a bit consistent.
So placing and cycling a crewmate now use the same color ranges.
It's less code being copied and pasted, especially since for my
over-30-FPS patch I would have to make a separate function for each if
both of them were still there, but if they're unified into one then I
will only have to make one more function.
And since map.scrolldir is now used outside of GAMEMODE, we'll need to
reset it in hardreset() and when exiting playtesting.
Previously, if you had backgrounds disabled in accessibility options,
and went to the editor and opened up the editor menu, it would be drawn
straight on top of what was already there in the editor instead of being
drawn on top of black. So now it's drawn on top of black.
During testing, I made a cursed level that set the flash timer to
precisely 1,000,000 frames. It turns out that if I activated the timer
in playtesting, exited playtesting, and exited the editor without ever
re-entering playtesting, the timer still kept going. So to prevent being
able to do that, we should hardreset() when exiting the editor.
In-game because that's where screen effects are used the most. But on
the title screen, screen effects are used when you press ACTION to start
the game, and when you enable screen effects, too.
Otherwise, we don't need screen effects for any other game-gamestate.
This de-duplicates the screen effects rendering code by putting it
inside a function, Graphics::renderwithscreeneffects(), and using that
instead of copy-pasted code.
The code to decrement the timers for flashing and shaking is now handled
outside the game-gamestate case-switch, instead of having to be
duplicated inside each render function.
As a bonus, I made it so the timer decrements even if screen effects are
disabled. This is to prevent any theoretical situation where the timer
can "pile up" due to disabled screen effects not letting it tick down.
This fixes being able to rack up a large amount of stack frames by
pressing Esc repeatedly in the editor, which would be a problem if you
were to then return to the main menu afterwards.
Instead, if Menu::ed_settings is already in the stack, the game will
simply return to that menu instead of creating it. Else, it will just
create the menu.
Also, as extra attention to detail, I made sure that the menu create or
return only happens if Esc opens the settings menu, and not when Esc is
closes it.
This stabilizes the code that handles the menu that you land on if you
press Esc and quit to the menu.
Instead of using Game::returnmenu(), we now use the new function
Game::returntomenu() to clearly express intent that we want to return to
a specific menu. So I've added another kludge variable
Game::wasinintermission for the was-in-intermission case.
Also, I made it so that if you didn't have a main game telesave or
quicksave, you just get brought back to the main menu. Because you
shouldn't be able to go to the play menu without a quicksave or
telesave.
To find each individual tag quickly, to optimize levels list loading.
I opted to not read the tags <Created>, <Modified>, and <Modifiers> as
they're actually pretty useless.
Also I've added a tag finder for <MetaData> but it's not meant to be
used directly, it's only used to check that the tag exists.
This also replaces some createmenu()s with returnmenu()s as needed even
when said createmenu()s already didn't go to the main menu.
Now when you exit the level editor, you'll be selecting the "level
editor" option in "play levels", and if you exit from a level you'll
still be selecting that level in the levels list.
Furthermore, regardless of what you're exiting, your cursor position
will be remembered.
Much more stylistic, you don't need to repeat "game.currentmenuname" for
each case, and you don't need to deal with the dangling first "if" that
doesn't have an "else".
Stringly-typed things are bad, because if you make a typo when typing
out a string, it's not caught at compile-time. And in the case of this
menu system, you'd have to do an excessive amount of testing to uncover
any bugs caused by a typo. Why do that when you can just use an enum and
catch compile-time errors instead?
Also, you can't use switch-case statements on stringly-typed variables.
So every menu name is now in the enum Menu::MenuName, but you can simply
refer to a menu name by just prefixing it with Menu::.
Unfortunately, I've had to change the "continue" menu name to be
"continuemenu", because "continue" is a keyword in C and C++. Also, it
looks like "timetrialcomplete4" is an unused menu name, even though it
was referenced in Render.cpp.
Firstly, menu options are no longer ad-hoc objects, and are added by
using Game::option() (this is the biggest change). This removes the
vector Game::menuoptionsactive, and Game::menuoptions is now a vector of
MenuOption instead of std::string.
Secondly, the manual tracker variable of the amount of menu options,
Game::nummenuoptions, has been removed, in favor of using vectors
properly and using Game::menuoptions::size().
As a result, a lot of copy-pasted code has been removed from
Game::createmenu(), mostly due to having to have different versions of
menus depending on whether or not we have certain defines, or having an
mmmmmm.vvv file inside the VVVVVV directory. In the old days, you
couldn't just add or remove a menu option conveniently, you had to
shuffle around the position of every other menu option too, which
resulted in lots of copy-pasted code. But now this copy-pasted code has
been de-duplicated, at least in Game::createmenu().
Previously, it existed solely to count the number of trinkets and
crewmates when loading a level, because we were keeping track of the
amount of them manually, incrementing and decrementing every time a
trinket or crewmate was added or removed, but loading a new level
represented a case that could potentially not be an increment or
decrement.
However, since the amount tracking is now handled automatically, this
function now does nothing, and can be safely removed.
Same principle as removing the separate variable to track number of
collected trinkets. This means it's less error-prone as we're no longer
tracking number of trinkets separately.
In the function that counts the number of trinkets, I would've liked to
have used std::count_if(). However, the most optimal way would require
using a lambda, and lambdas are too new for the C++ standard we're
using. So I just bit the bullet and counted them manually.
It is an exact duplicate of musicclass::haltdasmusik(), so use that
function instead and update callers. Looks like
musicclass::haltdasmusik() came first, anyway (musicclass::stopmusic()
was only used in editor.cpp).
The problem is that it would index out-of-bounds if you did this, but
this UB hasn't caused an exception until my change to refactor
script-related vectors by removing their separate length-trackers.
Just a miscellaneous code cleanup.
There's no glitches that take advantage of the previous situation,
namely that 'temp' was a global variable in Logic.cpp and editor.cpp.
Even if there were, it seems like it would easily lead to some undefined
behavior. So it's good to clean this up.
It looks like this may have been used earlier in development, judging
from the name, obviously, but right now it seems like it's used as an
error message if a main game level is asked for an invalid room (well,
only two of them - the Lab and Warp Zone). It should probably be
formalized into an error system, if we want to keep teststring, and also
people would never see it anyway because I don't think there's a
reliable and consistent way to trigger loading a non-existent room.
I have seen someone manage to load a non-existent Warp Zone room only
one time, but even then this teststring didn't pop up. So this
teststring doesn't even trigger in the right circumstances.
Also, when it does pop up, as far as I can tell it will stay onscreen,
which is kinda annoying. So I'm just removing this ancient relic from
the code.
Looks like this function was created because editorclass::load() takes
in a string by reference, not by value, and thus mutates it afterwards,
so if you passed a string in when you didn't want it to be mutated, bad
things would happen.
However, a better workaround for the above issue would simply to
duplicate the string and pass that string instead, thus the original
string wouldn't be affected.
This changes something like UtilityClass::String to help.String,
basically. It takes less typing this way, and is a neat effect of having
global args actually be global variables.
The 'r', 'g', and 'b' arguments do absolutely nothing. Except unlike
Graphics::drawtile(), there's only one version of Graphics::drawtile2(),
so just remove those args and update callers.
The 'r', 'g', and 'b' arguments do absolutely nothing, even though
they're used in the version of Graphics::drawtile() that's more used. So
delete the other version without those extra arguments, and then remove
the extra arguments from the remaining version. And then update callers.
This removes global arg passing from all functions on editorclass.
Callers have been updated correspondingly. Additionally, all 'dwgfx' has
been replaced with 'graphics' in editor.cpp.
This commit removes all global args from the parameters of each function
on the scriptclass object, and updates all places they are called
accordingly. It also changes all instances of 'dwgfx' to 'graphics' in
Script.cpp.
Interestingly enough, it looks like editor.h depended on Script.h's
class define of the musicclass. I've temporarily placed the class define
in editor.h, but by the end of this patchset it'll be gone.
This removes global args from all functions on the Graphics class.
Callers of those functions in other files have been updated accordingly.
Of course, since Graphics.cpp is already in the Graphics namespace,
I do not need to change all 'dwgfx' to 'graphics' in Graphics.cpp.
This is a refactor that turns the script-related arrays `ed.sb`, and
`ed.hooklist` into C++ vectors (`script.commands` was already a vector, it was
just misused). The code handling these vectors now looks more like idiomatic
C++ than sloppily-pasted pseudo-ActionScript. This removes the variables
`script.scriptlength`, `ed.sblength`, and `ed.numhooks`, too.
This reduces the amount of code needed to e.g. simply remove something from
any of these vectors. Previously the code had to manually shift the rest of
the elements down one-by-one, and doing it manually is definitely error-prone
and tedious.
But now we can just use fancy functions like `std::vector::erase()` and
`std::remove()` to do it all in one line!
Don't worry, I checked and `std::remove()` is in the C++ standard since at least
1998.
This patch makes it so the `commands` vector gets cleared when
`scriptclass::load()` is ran. Previously, the `commands` vector never actually
properly got cleared, so there could potentially be glitches that rely on the
game indexing past the bounds set by `scriptlength` but still in-bounds in the
eyes of C++, and people could potentially rely on such an exploit...
However, I checked, and I'm pretty sure that no such glitch previously existed
at all, because the only times the vector gets indexed are when `scriptlength`
is either being incremented after starting from 0 (`add()`) or when it's
underneath a `position < scriptlength` conditional.
Furthermore, I'm unaware of anyone who has actually found or used such an
exploit, and I've been in the custom level community for 6 years.
So I think it's fine.
Text outline is not drawn on roomtext when you're actually playing the
game, so don't draw the outline in the editor, either.
FIQ mistakenly added text outline to roomtext in
ca9f577fc4.
There's an if-else chain that first deals with figuring out if there's
an entity where your left-click happened, and to do this it uses
edentat(), which returns a sentinel value of -1 if there is NOT an
entity where your cursor is.
It's very important to check that the value returned ISN'T -1 before you
start indexing the 'edentity' vector, since if you DO index it with that
-1, it'll result in Undefined Behavior because you're doing an
out-of-bounds array access.
Now, here's what the if-else chain looked like before:
if(tmp==-1 && ed.free(ed.tilex,ed.tiley)==0)
{
...
}
else if(edentity[tmp].t==1)
The bug here is very subtle but it was an easy oversight. Basically, if
'ed.free' ended up not being zero, control flow would jump to the next
"else if" over, which then ends up asking for the -1th index of
'edentity', which is Undefined Behavior.
This undefined behavior has now resulted in a crash on my system after
TerryCavanagh/VVVVVV#172, due it shuffling things around juuuuust enough
such that this UB would end up resulting in a segfault instead of
chugging along and working fine. For me and my system, this meant that
if my first left-click in the editor upon opening the game was me
placing down a tile and not placing down an entity, the game would
crash. But, it would be fine if I first placed down an entity and then
afterwards placed down tiles, because it's UB.
And I'm almost certain this was the cause of the very strange bug where
you couldn't hold down left-click for the foreground-placing tool (but
you COULD for the background-placing tool) that seemed to occur most
often on Windows (TerryCavanagh/VVVVVV#25).
The solution to this is to stick in another conditional in the tree
before any indexing occurs, such that there's no way any other
conditionals with the indexing in the conditional tree could end up
being hit. In summary, the if-else chain looks like this now:
if(tmp==-1 && ed.free(ed.tilex,ed.tiley)==0)
{
...
}
else if(tmp == -1)
{
//Important! Do nothing, or else Undefined Behavior will happen
}
else if(edentity[tmp].t==1)
This turns the array 'edentity' into a proper vector, and removes the need to
use a separate length-tracking variable and manually keep track of the actual
amount of edentities in the level by using the long-winded
'EditorData::GetInstance().numedentities'. This manual tracking was more
error-prone and much less maintainable.
editorclass::naddedentity() has been removed due to now functionally being the
same as editorclass::addedentity() (there's no more
'EditorData::GetInstance().numedentities' to not increment) and for also being
unused in the first place.
editorclass::copyedentity() has been removed because it was only used to shift
the rest of the edentities up manually, but now that we let C++ do all the
hard work it's no longer necessary.
This fixes another way you could end up typing on a non-existent line in
the script editor.
In a script with only 1 line, which is empty, the game would let you
press backspace on it, removing the line. This results in you typing on
a non-existent line.
You will keep typing on it until you either close the script or press
Up. If you press Up, you will be unable to get back to the non-existent
line, for it doesn't exist - but the text you typed on the non-existent
line will still be there, until you close the script and re-open it.
This makes the "[Press ENTER to return to editor]" fade out after a few frames, allowing screenshots of custom levels to be cleaner and to make sure nothing is obscured while the user is editing their level.
This commit also adds alpha support in BlitSurfaceColoured, where it takes into account the alpha of the pixel *and* the alpha of the color.
`graphics::getRGBA(r,g,b,a)` was added to help with this.
In `editor.cpp`, there's a few sections of code that try and index stuff using `string.length()-1`.
This causes issues where if the string is empty, the result is -1, causing undefined behavior.
Flibit fixed a few of these cases, like on line `375` of editor.cpp:
`if((int) tstring.length() - 1 >= 0) // FIXME: This is sketchy. -flibit`
It turns out that one of these weren't caught, over at line `471`.
`tstring=tstring[tstring.length()-1];`
This causes builds compiled on Windows to segfault if you load more than one level in the editor.
I added a quick `if` around it, setting `tstring` to an empty string, which seems to fix the problem.
This uses utfcpp combined with a custom font, in the form of a PNG and text file. By default, the game acts exactly as it did before; custom fonts can be provided by third parties.
This fixes a bug where if warpdir() was used during in-editor
playtesting, the changed warp direction would persist even when leaving
playtesting.
This would be very annoying to correct back every time you playtested
and warpdir() was used, so I've added some kludge to store the actual
warp direction of each room when entering playtesting, and then set the
warp directions back when leaving playtesting.
First, two bug fixes. Room text input mode wasn't properly unset
upon pressing Esc, making the prompt get stuck, requiring you to
add roomtext again and finish it to make it go away. Secondly,
escaping script text input would remove the wrong entity.
I also tweaked the handling slightly so that instead of deleting
the entity if it already existed if escaping from text input,
it merely reverts the change in script name/roomtext to what it
was previously.
I considered refactoring the editor text input handler entirely,
but figured such a change would be a bit too extensive for the
purpose of this repository.
Ingame entities are drawn backwards, probably to draw the player on top,
being entity 0 (usually, at least). Make the level editor draw entities
in the same order.
FillRect() is similar enough to memset when blending isn't used, so the
game will take a fast path drawing the roomname background when the
background is opaque.
This is the variable dwgfx.translucentroomname and <translucentroomname>
in unlock.vvv.
This lets you see through the black background of the roomname at the
bottom of the screen, i.e. it makes the roomname background translucent.
So you can see if someone decides to hide pesky spikes there.
The roomname background used to just be a simple SDL_Rect that was drawn
using SDL_FillRect with a color of 0. Unfortunately, it seems that you
cannot use transparent colors with SDL_FillRect, it just defaults to
being fully opaque. However, you CAN draw surfaces with translucency,
which seems like the easiest thing to do. But the first step is to
convert the roomname background to an SDL_Surface.
This replaces the FillRect()s with SDL_BlitSurface() in the three places
roomnames are drawn: in towerrender, in gamerender, and in editorrender.
For some reason, there are two lines that have been copy-pasted the
exact same way and in the exact same place, namely being at the end of
each branch of the if-else conditional, which makes them be executed no
matter what. If they're going to be executed no matter what, we might as
well make it clearer and take those two lines out of each branch.
An earlier change caused TinyXml to prettyprint specifically the
contents of entities (script names and roomtext) a bit more than
before in level files, and added an unusually high amount of whitespace
(particularly, it added an empty line to every entity). This happens
because, for some reason, an empty string is explicitly being added
when creating an entity XML element. The line is so mysterious it feels
like it probably somehow solved a nasty bug long ago and shouldn't be
touched, but it was probably just a mistake, and with all that
whitespace it doesn't look good.
2.2 will handle this just fine, because there are 100 slots allocated
for trinkets and crewmates, and it will save and load all 100 slots just
fine. Except, it only resets the first 20 slots when starting a level
from the beginning, but that's minor and already fixed in 2.3 anyway.
This commit makes it so you can now place up to one hundred trinkets and
crewmates in the editor.
When editorclass::reset() was resetting the contents of the level
previously, it was mixing up the X and Y bounds. The Y bound was
supposed to be 30*maxheight, and the X bound was supposed to be
40*maxwidth. Instead, it took 30*maxwidth as its Y bound and
40*maxheight as its X bound.
Then, when it actually indexes the contents vector to set each tile to
0, it used 30*maxwidth instead of 40*maxwidth.
The difference between width and height is a bit hard to spot, but one
thing you can do to remember the difference is to remember the fact that
X corresponds with width, and Y corresponds with height. Also, rooms are
40 by 30 tiles, and so X (and therefore width) should correspond with
40, and Y (and therefore height) should correspond with 30.
As a result of mixing up the variables, whenever you played a 20x20 map,
quit the level and then started making a new 20x20 map, the tiles of the
last four rows of the previous map would persist, from y=16 (1-indexed)
all the way to y=20 (1-indexed).
I don't recall anyone ever running into this bug before, which is a bit
strange. But if no one truly has ever ran into this bug before, then I'm
genuinely surprised.
While working on the patch to fix the enemy type room property of each
room not getting reset, and testing the fix, I noticed that for some
reason some contents of the previous level I played in order to test the
enemy type property persisting was ALSO persisting alongside the enemy
type property.
Then I read the code and when I realized that the X and Y bounds were
getting mixed up I groaned. Very loudly.
This fixes a bug where if you loaded a level, then started making a new
level in the editor, the enemy types from the previous level would
persist.
While working on VVVVVV: Community Edition and adding a new room
property for enemy speed, I noticed that enemy type was not getting
reset at all. After some testing, I confirmed that this was the case. So
this bug is fixed now.
Out-of-bounds array access is Undefined Behavior, which means Bad
Things.
In this particular case, it was indexing an array by using the
`testeditor` variable. Which is fine, except it was indexing that array
*in a conditional that only happens if `testeditor` is -1*. So it was
indexing an array at position -1, which is Out of Bounds and is Not
Good.
There's a long-standing issue where the Direct Mode status of the loaded
custom map in memory never resets properly. So if you load a level with
a certain layout of Direct Mode rooms, that same layout will be
preserved if you start making a new level in the editor. This commit
fixes that issue.
Similar to 2ebccbc3e9, there's also a
long-standing bug where if you backspace an empty line (and this time,
the line IS actually already empty, not merely
emptied-earlier-in-the-frame), the game will quickly delete more blank
lines if there are any above the blank line you deleted. Again, this is
annoying too, if you so happen to need to use lots of blank lines.
To fix this, it's simple - just set ed.keydelay to 6 when the game
backspaces an empty line. Then it won't be so trigger-happy in deleting
blank lines.
There is a long-standing bug with the script editor where if you delete
the last character of a line, it IMMEDIATELY deletes the line you're on,
and then moves your cursor back to the previous line. This is annoying,
to say the least.
The reason for this is that, in the sequence of events that happens in
one frame (known as frame ordering), the code that backspaces one
character from the line when you press Backspace is ran BEFORE the code
to remove an empty line if you backspace it is ran. The former is
located in key.Poll(), and the latter is located in editorinput().
Thus, when you press Backspace, the game first runs key.Poll(), sees
that you've pressed Backspace, and dutifully removes the last character
from a line. The line is now empty. Then, when the game gets around to
the "Are you pressing Backspace on an empty line?" check in
editorinput(), it thinks that you're pressing Backspace on an empty
line, and then does the usual line-removing stuff.
And actually, when it does the check in editorinput(), it ACTUALLY asks
"Are you pressing Backspace on THIS frame and was the line empty LAST
frame?" because it's checking against its own copy of the input buffer,
before copying the input buffer to its own local copy. So the problem
only happens if you press and hold Backspace for more than 1 frame.
It's a small consolation prize for this annoyance, getting to
tap-tap-tap Backspace in the hopes that you only press it for 1 frame,
while in the middle of something more important to do like, oh I don't
know, writing a script.
So there are two potential solutions here:
(1) Just change the frame ordering around.
This is risky to say the least, because I'm not sure what behavior
depends on exactly which frame order. It's not like it's key.Poll()
and then IMMEDIATELY afterwards editorinput() is run, it's more
like key.Poll(), some things that obviously depend on key.Poll()
running before them, and THEN editorinput(). Also, editorinput() is
only one possible thing that could be ran afterwards, on the next
frame we could be running something else entirely instead.
(2) Add a kludge variable to signal when the line is ALREADY empty so
the game doesn't re-check the already-empty line and conclude that
you're already immediately backspacing an empty line.
I went with (2) for this commit, and I've added the kludge variable
key.linealreadyemptykludge.
However, that by itself isn't enough to fix it. It only adds about a
frame or so of delay before the game goes right back to saying "Oh,
you're ALREADY somehow pressing backspace again? I'll just delete this
line real quick" and the behavior is basically the same as before,
except now you have to hit Backspace for TWO frames or less instead of
one in order to not have it happen.
What we need is to have a delay set as well, when the game deletes the
last line of a char. So I set ed.keydelay to 6 as well if editorinput()
sses that key.linealreadyemptykludge is on.
For a long time, the script editor has had a bug where it would let you
put the cursor on a nonexistent script line, which would APPEAR to be
the last line of the script... but in reality, it WASN'T the last line
of the script, and in fact, the ACTUAL last line was the line ABOVE the
script.
So, if you typed anything on this nonexistent line, it would appear to
get erased when exiting the script. Thus, people have (erroneously)
misdiagnosed this as the script editor somehow being trigger-happy and
erasing lines when it shouldn't be, when in reality it should've have
let you gone onto that line in the first place!
The TinyXml functions to load and save files don't properly support
unicode file paths on Windows, so in order to support that properly, I
saw no other option than to do the actual loading and saving via PHYSFS
(or to use the Windows API on Windows and retain doc.LoadFile and
doc.SaveFile on other OSes, but that'd be more complicated and
unnecessary, we already have PHYSFS, right?).
There are two new functions in FileSystemUtils:
bool FILESYSTEM_saveTiXmlDocument(const char *name, TiXmlDocument *doc)
bool FILESYSTEM_loadTiXmlDocument(const char *name, TiXmlDocument *doc)
Any instances of doc.SaveFile(<FULL_PATH>) have been replaced by
FILESYSTEM_saveTiXmlDocument(<VVVVVV_FOLDER_PATH>, &doc), where
<FULL_PATH> included the full path to the saves or levels directory,
and <VVVVVV_FOLDER_PATH> only includes the path relative to the VVVVVV
directory.
When loading a document, a TiXmlDocument used to be created with a full
path in its constructor and doc.LoadFile() would then be called, now a
TiXmlDocument is constructed with no path name and
FILESYSTEM_loadTiXmlDocument(<VVVVVV_FOLDER_PATH>, &doc) is called.
