Thie removes the http-service hiscore api, which we haven't used in
awhile and doesn't work in practice due to upstream rate limits, as well
as the xp tracker which also hasn't been used in a long time since there
are now multiple quality community run xptrackers.
The examine text, buy limit, trade price, and fee info does not fit on
the interface anymore since the addition of the fee info. This compacts
the examine text as well as places the limit / trade price on the same
line in order to prevent pushing the fee off of the bottom of the
border.
This adds a new createMenuEntry api method to make MenuEntries instead.
Menu entries now have an associated callback called when they are
clicked on, avoiding most plugins from having to hook separately to
detect the menu click. Additionally get/set type has changed to take a
MenuAction.
The config service accepts in strings instead of json strings, however
this causes the normal control flow to throw a json parsing exception.
Since this happens so frequently it is using a measurable amount of CPU
time, so avoid it in the common case by testing if the string is json
first
Apply the color changes at chat build time directly to the message being
set on the component, instead of prepending the color change onto the
message itself.
This is a bit more robust since it doesn't break things which depend on
the chat message not starting with a color, which they would not
otherwise, such as the chat filter.
The flashing is caused by the vanilla overlays being hidden by the
detached overlay after being drawn, and so the remaining frames before
the next tick would have them hidden. Just always hide the overlay in
BeforeRender so that they are never drawn when the plugin is on.
Also add a quick check if the barrows region is loaded before trying to
draw the barrows brothers minimap overlay.
This allows the additional OFF mode which has the client manage its own
frame rate target without the use of vsync. On lower end systems which
have a hard time delivering a frame in the vsync interval this can give
a noticible performance improvement since it doesn't stall the cpu
waiting for the next vblank when a frame is skipped
This lets plain latin-character filters to match messages with accents
and diacritics which are not easily typed on all keyboard layouts.
Co-authored-by: Jordan Atwood <jordan.atwood423@gmail.com>
This hasn't worked in awhile since it was assining the scroll value to the
wrong variable. Now we are computing the scroll in onScriptPreFixed for
[proc,bankmain_finishbuilding], so it is no longer necessary here.
With tag tab tab or "Remove tab separators", the scrollbar height needs to be
adjusted to account for either many tag tabs (which requires scrolling to see)
or the hiding of the tab separators. This is because we artifically filter the
bank results without keeping the bank in searching mode, causing the vanilla
code which usually computes the scroll height to be incorrect.
The existing code does this by manually invoking [proc,update_scrollbar] later
with the recomputed scroll height. Now that invokeLater runs next tick and not
next frame, this leaves several frames with the scrollbar the incorrect size.
The desired scroll height is passed to [proc,bankmain_finishbuilding] in
argument int12, so instead we recompute the scroll height and overwrite the
script argument prior to execution.
From the opengl wiki:
Recent GL drivers implement a new WGL/GLX extension called EXT_swap_control_tear.
This extension brings "adaptive vsync" as featured in modern gaming consoles to the PC.
Adaptive vsync enables v-blank synchronisation when the frame rate is higher than the
sync rate, but disables synchronisation when the frame rate drops below the sync rate.
Disabling the synchronisation on low frame rates prevents the common problem where the
frame rate syncs to a integer fraction of the screen's refresh rate in a complex scene.
If the EXT_swap_control_tear extension is not supported, JOGL uses the absolute value
as the swap interval, 1, the same as before.
Shortly after logging in there several frames without a tli, and so no scene is
rendered on them. Since we glClear the entire buffer each frame and re-render
the most recent scene draw to emulate the vanilla effect of not painting over
the viewport, the viewport offset x/y values at the time we draw the frame may
be -1 due to that frame not having the scene rendered. Store the viewport
offsets at the time of the most recent scene draw and use them for each
subsequent frame.
With invokeLater() now running next tick instead of next frame, with uncapped fps there can be several frames with the bank pin visible in the chatbox. This blocks the keyboard input on this tick, preventing having to clear the chatbox later.
Also move the RL Instructions from the script assembler to cache too, so it can be reused by the updater. The api Opcodes class has no use in the public api and so is removed.
Doing the xp calculation with floats and then converting it to double is
less precise, and causes minor errors that accumulate with very large xp
ranges
XP cannot be more precise than a 10th of an xp. This was compute more precise xp values and when applied over many actions caused the computed actions to be incorrect
When the wiki entity lookup option is off the parent layer isn't
positioned at all, causing the resulting wiki icon we create to be
incorrectly positioned.
This method checks if the model is on screen by projecting points at the model
extremes to screen and seeing if they are in the viewport bounds.
When doing this, it offsets the model height for the top of the viewport check,
but does not offset the bottom of the model for the bottom viewport check. This
offsets the bottom height to the model y position before doing this check,
which improves the test on models which are beneath the tile they are placed on.
This is most noticible in the Karamja dungeon and Hydra dungeon which has
objects for the lava instead of tiles, and they are placed at y=0 with some or
all of the model being below that.
Without this, the minimum size will be the same size as was prior to
maximizing the frame. If the side bar or panel is opened or closed when
maximized this can cause the minimum size to be incorrect after a
restore.
Co-authored-by: Mark Koester <mark.koester1@gmail.com>
Co-authored-by: Jeremy Plsek <jeremyplsek@gmail.com>
When the varrock armour effect triggers it will smelt two bars at once,
using two charges, unless there is only one ore remaining where it will
use only one charge.
Co-authored-by: Jason Barr <58792686+JBarr2710@users.noreply.github.com>
The ground marker minimap overlay gets a local point for the tile to mark from fromWorld()
and passes it to localToMinimap() - but because the adjustment of -1, when the player position
is subtracted from it to compute the minimap location it causes the computed position to be off
by a full tile.
Additionally, drawRect() width and height is inclusive so the width/height must be subtracted by 1
GPU uses an alternative projection to avoid vertex snapping. Usually
this doesn't matter much, and in the worst case causes the discrepancy
to be only a few pixels, but with the model outline feature it causes
the outlines to be noticibly off when GPU is enabled.
This adds a second model-to-canvas method using the alternative
projection and uses it when GPU is enabled.
