## Summary of the Pull Request
Identifies and scales glyphs in the box and line drawing ranges U+2500-U+259F to fit their cells.
## PR Checklist
* [x] Closes#455
* [x] I work here.
* [x] Manual tests. This is all graphical.
* [x] Metric ton of comments
* [x] Math spreadsheet included in PR.
* [x] Double check RTL glyphs.
* [x] Why is there the extra pixel?
* [x] Scrolling the mouse wheel check is done.
* [x] Not drawing outline?
* [x] Am core contributor. Roar.
* [x] Try suppressing negative scale factors and see if that gets rid of weird shading.
## Detailed Description of the Pull Request / Additional comments
### Background
- We want the Terminal to be fast at drawing. To be fast at drawing, we perform differential drawing, or only drawing what is different from the previous frame. We use DXGI's `Present1` method to help us with this as it helps us compose only the deltas onto the previous frame at drawing time and assists us in scrolling regions from the previous frame without intervention. However, it only works on strictly integer pixel row heights.
- Most of the hit testing and size-calculation logic in both the `conhost` and the Terminal products are based on the size of an individual cell. Historically, a cell was always dictated in a `COORD` structure, or two `SHORT` values... which are integers. As such, when we specify the space for any individual glyph to be displayed inside our terminal drawing region, we want it to fall perfectly inside of an integer box to ensure all these other algorithms work correctly and continue to do so.
- Finally, we want the Terminal to have font fallback and locate glyphs that aren't in the primary selected font from any other font it can find on the system that contains the glyph, per DirectWrite's font fallback mechanisms. These glyphs won't necessarily have the same font or glyph metrics as the base font, but we need them to fit inside the same cell dimensions as if they did because the hit testing and other algorithms aren't aware of which particular font is sourcing each glyph, just the dimensions of the bounding box per cell.
### How does Terminal deal with this?
- When we select a font, we perform some calculations using the design metrics of the font and glyphs to determine how we could fit them inside a cell with integer dimensions. Our process here is that we take the requested font size (which is generally a proxy for height), find the matching glyph width for that height then round it to an integer. We back convert from that now integer width to a height value which is almost certainly now a floating point number. But because we need an integer box value, we add line padding above and below the glyphs to ensure that the height is an integer as well as the width. Finally, we don't add the padding strictly equally. We attempt to align the English baseline of the glyph box directly onto an integer pixel multiple so most characters sit crisply on a line when displayed.
- Note that fonts and their glyphs have a prescribed baseline, line gap, and advance values. We use those as guidelines to get us started, but then to meet our requirements, we pad out from those. This results in fonts that should be properly authored showing gaps. It also results in fonts that are improperly authored looking even worse than they normally would.
### Now how does block and line drawing come in?
- Block and Line drawing glyphs are generally authored so they will look fine when the font and glyph metrics are followed exactly as prescribed by the font. (For some fonts, this still isn't true and we want them to look fine anyway.)
- When we add additional padding or rounding to make glyphs fit inside of a cell, we can be adding more space than was prescribed around these glyphs. This can cause a gap to be visible.
- Additionally, when we move things like baselines to land on a perfect integer pixel, we may be drawing a glyph lower in the bounding box than was prescribed originally.
### And how do we solve it?
- We identify all glyphs in the line and block drawing ranges.
- We find the bounding boxes of both the cell and the glyph.
- We compare the height of the glyph to the height of the cell to see if we need to scale. We prescribe a scale transform if the glyph wouldn't be tall enough to fit the box. (We leave it alone otherwise as some glyphs intentionally overscan the box and scaling them can cause banding effects.)
- We inspect the overhang/underhang above and below the boxes and translate transform them (slide them) so they cover the entire cell area.
- We repeat the previous two steps but in the horizontal direction.
## Validation Steps Performed
- See these commments:
- https://github.com/microsoft/terminal/issues/455#issuecomment-620248375
- https://github.com/microsoft/terminal/issues/455#issuecomment-621533916
- https://github.com/microsoft/terminal/issues/455#issuecomment-622585453
Also see the below one with more screenshots:
- https://github.com/microsoft/terminal/pull/5743#issuecomment-624940567
[Git2Git] Merged PR 4644345: conhost: disable the DX renderer in inbox builds
We're going to be taking on some changes to the Dx renderer that are at
the very least annoying and at the very most inconsequential to the
inbox console. This commit removes support for the DX renderer from the
inbox console.
SizeBench reports that ConRenderDx contributes 55.1kb to the conhost
image (as its third largest constituent library), so this should net us
a couple pleasant WPG improvements down the line.
Related work items: #26291552 Retrieved from https://microsoft.visualstudio.com os OS official/rs_onecore_dep_uxp 6e36786d447b7975298ba31ccd77c5c649fbfbe6
Related work items: #26291552
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
This fixes the RTL regression caused in #4747. We create the rectangle taking the direction (through the BiDi Level) into account, and then the rendering works again. The GlyphRun shaping could still probably use some work to be a polished thingy, and there are still issues with RTL getting chopped up a lot when there's font fallback going on, but this fixes the regression, and it's now functional again.
<!-- Other than the issue solved, is this relevant to any other issues/existing PRs? -->
## References
#4779#4747
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#4779
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [ ] I've discussed this with core contributors already. If not checked, I'm ready to accept this work might be rejected in favor of a different grand plan. Issue number where discussion took place: #xxx
<!-- Provide a more detailed description of the PR, other things fixed or any additional comments/features here -->
## Detailed Description of the Pull Request / Additional comments
The baseline is actually direction dependent. So when it was being initialized, the unconditional baseline as left broke it, setting the box off to right of the text. We just check if the `GlyphRun->bidiLevel` is set, and if so, we adjust it so that the baseline lines up with the right, not with the left.
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
## Summary of the Pull Request
This PR implements a pair of shims for `cmd` and `powershell`, so that their `cls` and `Clear-Host` functions will clear the entire terminal buffer (like they do in conhost), instead of just the viewport. With the conpty viewport and buffer being the same size, there's effectively no way to know if an application is calling these API's in this way with the intention of clearing the buffer or the viewport. We absolutely have to guess.
Each of these shims checks to see if the way that the API is being called exactly matches the way `cmd` or `powershell` would call these APIs. If it does, we manually write a `^[[3J` to the connected terminal, to get he Terminal to clear it's own scrollback.
~~_⚠️ If another application were trying to clear the **viewport** with an exactly similar API call, this would also cause the terminal scrollback to get cleared ⚠️_~~
* [x] Should these shims be restricted to when the process that's calling them is actually `cmd.exe` or `powershell.exe`? Can I even do this? I think we've done such a good job of isolating the client process information from the rest of the host code that I can't figure out how to do this.
- YES, this can be done, and I did it.
* [ ] **TODO**: _While I'm here_, should I have `DoSrvPrivateEraseAll` (the implementation for `^[[2J`, in `getset.cpp`) also manually trigger a EraseAll in the terminal in conpty mode?
## PR Checklist
* [x] Closes#3126
* [x] Actually closes#1305 too, which is really the same thing, but probably deserves a callout
* [x] I work here
* [x] Tests added/passed
* [n/a] Requires documentation to be updated
## Validation Steps Performed
* ran tests
* checked `cls` in the Terminal
* checked `Clear-Host` in the Terminal
* Checked running `powershell clear-host` from `cmd.exe`
Turns out we're still being a bit too aggressive when removing spaces.
If there are spaces at the end of the first run painted to a bottom
line, _and the bottom line was a different color than the previous_,
then we can't trim those spaces off the string. We still need to emit
those to make sure the terminal has colored spaces in it as well.
## References
* there's like 80 PRs in the last month for this function
## PR Checklist
* [x] Closes#5502
* [x] I work here
* [x] Tests added/passed
## Validation Steps
* [x] ran the tests
* [x] checked that vtpipeterm still worked
* [x] Checked that the bug was fixed in the Terminal
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
In tonight's episode of "Can we be even faster?", we will... you know what, just take a look at the code.
<!-- Other than the issue solved, is this relevant to any other issues/existing PRs? -->
## References
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [ ] Closes #xxx
* [ ] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [ ] I've discussed this with core contributors already. If not checked, I'm ready to accept this work might be rejected in favor of a different grand plan. Issue number where discussion took place: #xxx
<!-- Provide a more detailed description of the PR, other things fixed or any additional comments/features here -->
## Detailed Description of the Pull Request / Additional comments
It is actually a quite common technique seen inside the codebase to first reserve the spaces before pushing something into vectors. I don't know why it is not used here.
Before:
After:
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
A tiny performance fix in `renderer.cpp`.
<!-- Other than the issue solved, is this relevant to any other issues/existing PRs? -->
## References
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [ ] Closes #xxx
* [X] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [ ] I've discussed this with core contributors already. If not checked, I'm ready to accept this work might be rejected in favor of a different grand plan. Issue number where discussion took place: #xxx
<!-- Provide a more detailed description of the PR, other things fixed or any additional comments/features here -->
## Detailed Description of the Pull Request / Additional comments
The cluster construction code is intensively called during rendering. Even though a single `back()` is fast, but accumulated `back()`s still take a noticiable amount of CPU cycles.
Before:
After:
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
Manually validated.
## Summary of the Pull Request
- Adjusts scaling practices in `DxEngine` (and related scaling practices in `TerminalControl`) for pixel-perfect row baselines and spacing at High DPI such that differential row-by-row rendering can be applied at High DPI.
## References
- #5185
## PR Checklist
* [x] Closes#5320, closes#3515, closes#1064
* [x] I work here.
* [x] Manually tested.
* [x] No doc.
* [x] Am core contributor. Also discussed with some of them already via Teams.
## Detailed Description of the Pull Request / Additional comments
**WAS:**
- We were using implicit DPI scaling on the `ID2D1RenderTarget` and running all of our processing in DIPs (Device-Independent Pixels). That's all well and good for getting things bootstrapped quickly, but it leaves the actual scaling of the draw commands up to the discretion of the rendering target.
- When we don't get to explicitly choose exactly how many pixels tall/wide and our X/Y placement perfectly, the nature of floating point multiplication and division required to do the presentation can cause us to drift off slightly out of our control depending on what the final display resolution actually is.
- Differential drawing cannot work unless we can know the exact integer pixels that need to be copied/moved/preserved/replaced between frames to give to the `IDXGISwapChain1::Present1` method. If things spill into fractional pixels or the sizes of rows/columns vary as they are rounded up and down implicitly, then we cannot do the differential rendering.
**NOW:**
- When deciding on a font, the `DxEngine` will take the scale factor into account and adjust the proposed height of the requested font. Then the remainder of the existing code that adjusts the baseline and integer-ifies each character cell will run naturally from there. That code already works correctly to align the height at normal DPI and scale out the font heights and advances to take an exact integer of pixels.
- `TermControl` has to use the scale now, in some places, and stop scaling in other places. This has to do with how the target's nature used to be implicit and is now explicit. For instance, determining where the cursor click hits must be scaled now. And determining the pixel size of the display canvas must no longer be scaled.
- `DxEngine` will no longer attempt to scale the invalid regions per my attempts in #5185 because the cell size is scaled. So it should work the same as at 96 DPI.
- The block is removed from the `DxEngine` that was causing a full invalidate on every frame at High DPI.
- A TODO was removed from `TermControl` that was invalidating everything when the DPI changed because the underlying renderer will already do that.
## Validation Steps Performed
* [x] Check at 150% DPI. Print text, scroll text down and up, do selection.
* [x] Check at 100% DPI. Print text, scroll text down and up, do selection.
* [x] Span two different DPI monitors and drag between them.
* [x] Giant pile of tests in https://github.com/microsoft/terminal/pull/5345#issuecomment-614127648
Co-authored-by: Dustin Howett <duhowett@microsoft.com>
Co-authored-by: Mike Griese <migrie@microsoft.com>
The logic here, regarding deleting the spaces and just instantly adding
them bad, is incredibly suspect. Given that we're close to 0.11, I don't
think I can change it.
I've added a TODO with an issue number to figure out the right logic
here.
Fixes#5386.
## PR Checklist
* [x] Closes#5386
* [x] CLA signed
* [x] Tests added/passed
* [ ] Requires documentation to be updated
* [x] I'm horrified.
## Validation Steps Performed
Tests, manual validation of the scenario in 5386 and a repro program.
# Summary of the Pull Request
This PR will allow the cursor to be double width when on top of a double width character. This required changing `IsCursorDoubleWidth` to check whether the glyph the cursor's on top of is double width. This code is exactly the same as the original PR that addressed this issue in #2932. That one got reverted at some point due to the crashes related to it, but due to a combination of Terminal having come further since that PR and other changes to address use-after-frees, some of the crashes may/may not be relevant now. The ones that seemed to be relevant/repro-able, I attempt to address in this PR.
The `IsCursorDoubleWidth` check would fail during the `TextBuffer::Reflow` call inside of `Terminal::UserResize` occasionally, particularly when `newCursor.EndDeferDrawing()` is called. This is because when we tell the newCursor to `EndDefer`, the renderer will attempt to redraw the cursor. As part of this redraw, it'll ask if `IsCursorDoubleWidth`, and if the renderer managed to ask this before `UserResize` swapped out the old buffer with the new one from `Reflow`, the renderer will be asking the old buffer if its out-of-bounds cursor is double width. This was pretty easily repro'd using `cmatrix -u0` and resizing the window like a madman.
As a solution, I've moved the Start/End DeferDrawing calls out of `Reflow` and into `UserResize`. This way, I can "clamp" the portion of the code where the newBuffer is getting created and reflowed and swapped into the Terminal buffer, and only allow the renderer to draw once the swap is done. This also means that ConHost's `ResizeWithReflow` needed to change slightly.
In addition, I've added a WriteLock to `SetCursorOn`. It was mentioned as a fix for a crash in #2965 (although I can't repro), and I also figured it would be good to try to emulate where ConHost locks with regards to Cursor operations, and this seemed to be one that we were missing.
## PR Checklist
* [x] Closes#2713
* [x] CLA signed
* [x] Tests added/passed
## Validation Steps Performed
Manual validation that the cursor is indeed chonky, added a test case to check that we are correctly saying that the cursor is double width (not too sure if I put it in the right place). Also open to other test case ideas and thoughts on what else I should be careful for since I am quite nervous about what other crashes might occur.
## Summary of the Pull Request
When WSL vim prints the initial empty buffer (the one that's just a bunch of '\~'s), it prints this by doing the following:
* Print '\~' followed by enough spaces to clear the line
* Use CUP (`^[[H`) to move the cursor to the start of the next line
* repeat until the buffer is full
When we'd get the line of "\~ "... in conhost, we'd mark that line as wrapped.
Logically, it doesn't really make any sense that when we follow that up by moving the cursor, the line is wrapped. However, this is just how conhost is right now.
This wasn't ever a problem in just conhost before, because we really didn't care if lines in the alt buffer were "wrapped" or not. Plus, when vim would get resized, it would just reprint it's own buffer anyways. Nor was this a problem in conpty before this year (2020). We've only just recently added logic to conpty to try and preserve wrapped lines.
Initially, I tried fixing this by breaking the line manually when the cursor was moved. This seemed to work great, except for the win32 vim.exe. Vim.exe doesn't emit a newline or a CUP to get to the next line. It just _goes for it_ and keeps printing. So there's _no way_ for us to know the line broke, because they're essentially just printing one long line, assuming we'll automatically move the cursor.
So instead, I'm making sure to emit the proper number of spaces at the end of a line when the line is wrapped. We won't do any funny business in that scenario and try to optimize for them, we'll _just print the spaces_.
## References
* #5181 - This change regressed this
* #4415 - Actually implemented wrapped lines in conpty
## PR Checklist
* [x] Closes#5291
* [x] I work here
* [x] Tests added/passed
* [n/a] Requires documentation to be updated
## Validation Steps Performed
* Wrote a unittest first and foremost
* Checked vtpipeterm to make sure vim still works
* checked Terminal to make sure vim still works
## Summary of the Pull Request
Renderer: Add support for backoff and auto-disable on failed retry
This commit introduces a backoff (150ms * number of tries) to the
renderer's retry logic (introduced in #2830). It also changes the
FAIL_FAST to a less globally-harmful render thread disable, so that we
stop blowing up any application hosting a terminal when the graphics
driver goes away.
In addition, it adds a callback that a Renderer consumer can use to
determine when the renderer _has_ failed, and a public method to kick it
back into life.
Fixes#5340.
This PR also wires up TermControl so that it shows some UI when the renderer tastes clay.
## PR Checklist
* [x] Closes#5340
* [x] cla
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [x] I've discussed this with core contributors already.
## Validation Steps Performed
I tested this by dropping the number of retries to 1 and forcing a TDR while doing `wsl cmatrix -u0`. It picked up exactly where it left off.
As a bonus, you can actually still type into the terminal when it's graphically suspended (and `exit` still works.). The block is _entirely graphical_.
## Summary of the Pull Request
Adjusts DirectX renderer to use `til::bitmap` to track invalidation
regions. Uses special modification to invalidate a row-at-a-time to
ensure ligatures and NxM glyphs continue to work.
## References
Likely helps #1064
## PR Checklist
* [x] Closes#778
* [x] I work here.
* [x] Manual testing performed. See Performance traces in #778.
* [x] Automated tests for `til` changes.
* [x] Am core contributor. And discussed with @DHowett-MSFT.
## Detailed Description of the Pull Request / Additional comments
- Applies `til::bitmap` as the new invalidation scheme inside the
DirectX renderer and updates all entrypoints for collecting
invalidation data to coalesce into this structure.
- Semi-permanently routes all invalidations through a helper method
`_InvalidateRectangle` that will expand any invalidation to cover the
entire line. This ensures that ligatures and NxM glyphs will continue
to render appropriately while still allowing us to dramatically reduce
the number of lines drawn overall. In the future, we may come up with
a tighter solution than line-by-line invalidation and can modify this
helper method appropriately at that later date to further scope the
invalid region.
- Ensures that the `experimental.retroTerminalEffects` feature continues
to invalidate the entire display on start of frame as the shader is
applied at the end of the frame composition and will stack on itself
in an amusing fashion when we only redraw part of the display.
- Moves many member variables inside the DirectX renderer into the new
`til::size`, `til::point`, and `til::rectangle` methods to facilitate
easier management and mathematical operations. Consequently adds
`try/catch` blocks around many of the already-existing `noexcept`
methods to deal with mathematical or casting failures now detected by
using the support classes.
- Corrects `TerminalCore` redraw triggers to appropriately communicate
scrolling circumstances to the renderer so it can optimize the draw
regions appropriately.
- Fixes an issue in the base `Renderer` that was causing overlapping
scroll regions due to behavior of `Viewport::TrimToViewport` modifying
the local. This fix is "good enough" for now and should go away when
`Viewport` is fully migrated to `til::rectangle`.
- Adds multiplication and division operators to `til::rectangle` and
supporting tests. These operates will help scale back and forth
between a cell-based rectangle and a pixel-based rectangle. They take
special care to ensure that a pixel rectangle being divided downward
back to cells will expand (with the ceiling division methods) to cover
a full cell when even one pixel inside the cell is touched (as is how
a redraw would have to occur).
- Blocks off trace logging of invalid regions if no one is listening to
optimize performance.
- Restores full usage of `IDXGISwapChain1::Present1` to accurately and
fully communicate dirty and scroll regions to the underlying DirectX
framework. This additional information allows the framework to
optimize drawing between frames by eliminating data transfer of
regions that aren't modified and shuffling frames in place. See
[Remarks](https://docs.microsoft.com/en-us/windows/win32/api/dxgi1_2/nf-dxgi1_2-idxgiswapchain1-present1#remarks)
for more details.
- Updates `til::bitmap` set methods to use more optimized versions of
the setters on the `dynamic_bitset<>` that can bulk fill bits as the
existing algorithm was noticeably slow after applying the
"expand-to-row" helper to the DirectX renderer invalidation.
- All `til` import hierarchy is now handled in the parent `til.h` file
and not in the child files to prevent circular imports from happening.
We don't expect the import of any individual library file, only the
base one. So this should be OK for now.
## Validation Steps Performed
- Ran `cmatrix`, `cmatrix -u0`, and `cacafire` after changes were made.
- Made a bunch of ligatures with `Cascadia Code` in the Terminal
before/after the changes and confirmed they still ligate.
- Ran `dir` in Powershell and fixed the scrolling issues
- Clicked all over the place and dragged to make sure selection works.
- Checked retro terminal effect manually with Powershell.
Now that the Terminal is doing a better job of actually marking which
lines were and were not wrapped, we're not always copying lines as
"wrapped" when they should be. We're more correctly marking lines as not
wrapped, when previously we'd leave them marked wrapped.
The real problem is here in the `ScrollFrame` method - we'd manually
newline the cursor to make the terminal's viewport shift down to a new
line. If we had to scroll the viewport for a _wrapped_ line, this would
cause the Terminal to mark that line as broken, because conpty would
emit an extra `\n` that didn't actually exist.
This more correctly implements `ScrollFrame`. Now, well move where we
"thought" the cursor was, so when we get to the next `PaintBufferLine`,
if the cursor needs to newline for the next line, it'll newline, but if
we're in the middle of a wrapped line, we'll just keep printing the
wrapped line.
A couple follow up bugs were found to be caused by the same bad logic.
See #5039 and #5161 for more details on the investigations there.
## References
* #4741 RwR, which probably made this worse
* #5122, which I branched off of
* #1245, #357 - a pair of other conpty wrapped lines bugs
* #5228 - A followup issue for this PR
## PR Checklist
* [x] Closes#5113
* [x] Closes#5180 (by fixing DECRST 25)
* [x] Closes#5039
* [x] Closes#5161 (by ensuring we only `removeSpaces` on the actual
bottom line)
* [x] I work here
* [x] Tests added/passed
* [n/a] Requires documentation to be updated
## Validation Steps Performed
* Checked the cases from #1245, #357 to validate that they still work
* Added more and more tests for these scenarios, and then I added MORE
tests
* The entire team played with this in selfhost builds
## Summary of the Pull Request
Reduce the number of times we dispatch a cursor changed event. We were firing it every time the renderer had to do anything related to the cursor. Unfortunately, blinking the cursor triggered this behavior. Now we just check if the position has changed.
## PR Checklist
* [X] Closes#5143
## Validation Steps Performed
Verified using Narrator
Also verified #3791 still works right
Correct scrolling invalidation region for tmux in pty w/ bitmap
Add tracing for circling and scrolling operations. Fix improper
invalidation within AdjustCursorPosition routine in the subsection about
scrolling down at the bottom with a set of margins enabled.
## References
- Introduced with #5024
## Detailed Description of the Pull Request / Additional comments
- This occurs when there is a scroll region restriction applied and a
newline operation is performed to attempt to spin the contents of just
the scroll region. This is a frequent behavior of tmux.
- Right now, the Terminal doesn't support any sort of "scroll content"
operation, so what happens here generally speaking is that the PTY in
the ConHost will repaint everything when this happens.
- The PTY when doing `AdjustCursorPosition` with a scroll region
restriction would do the following things:
1. Slide literally everything in the direction it needed to go to take
advantage of rotating the circular buffer. (This would force a
repaint in PTY as the PTY always forces repaint when the buffer
circles.)
2. Copy the lines that weren't supposed to move back to where they were
supposed to go.
3. Backfill the "revealed" region that encompasses what was supposed to
be the newline.
- The invalidations for the three operations above were:
1. Invalidate the number of rows of the delta at the top of the buffer
(this part was wrong)
2. Invalidate the lines that got copied back into position (probably
unnecessary, but OK)
3. Invalidate the revealed/filled-with-spaces line (this is good).
- When we were using a simple single rectangle for invalidation, the
union of the top row of the buffer from 1 and the bottom row of the
buffer from 2 (and 3 was irrelevant as it was already unioned it)
resulted in repainting the entire buffer and all was good.
- When we switched to a bitmap, it dutifully only repainted the top line
and the bottom two lines as the middle ones weren't a consequence of
intersect.
- The logic was wrong. We shouldn't be invalidating rows-from-the-top
for the amount of the delta. The 1 part should be invalidating
everything BUT the lines that were invalidated in parts 2 and 3.
(Arguably part 2 shouldn't be happening at all, but I'm not optimizing
for that right now.)
- So this solves it by restoring an entire screen repaint for this sort
of slide data operation by giving the correct number of invalidated
lines to the bitmap.
## Validation Steps Performed
- Manual validation with the steps described in #5104
- Automatic test `ConptyRoundtripTests::ScrollWithMargins`.
Closes#5104
Retrieved from https://microsoft.visualstudio.com os OS official/rs_onecore_dep_uxp cba60cafaadfcc7890a45dea3e1a24412c3d0ec6
Related work items: MSFT:25631386
This commit introduces a github action to check our spelling and fixes
the following misspelled words so that we come up green.
It also renames TfEditSes to TfEditSession, because Ses is not a word.
currently, excerpt, fallthrough, identified, occurred, propagate,
provided, rendered, resetting, separate, succeeded, successfully,
terminal, transferred, adheres, breaks, combining, preceded,
architecture, populated, previous, setter, visible, window, within,
appxmanifest, hyphen, control, offset, powerpoint, suppress, parsing,
prioritized, aforementioned, check in, build, filling, indices, layout,
mapping, trying, scroll, terabyte, vetoes, viewport, whose
## Summary of the Pull Request
Moves the ConPTY drawing mechanism (`VtRenderer`) to use the fine-grained `til::bitmap` individual-dirty-bit tracking mechanism instead of coarse-grained rectangle unions to improve drawing performance by dramatically reducing the total area redrawn.
## PR Checklist
* [x] Part of #778 and #1064
* [x] I work here
* [x] Tests added and updated.
* [x] I'm a core contributor
## Detailed Description of the Pull Request / Additional comments
- Converted `GetDirtyArea()` interface from `IRenderEngine` to use a vector of `til::rectangle` instead of the `SMALL_RECT` to banhammer inclusive rectangles.
- `VtEngine` now holds and operates on the `til::bitmap` for invalidation regions. All invalidation operation functions that used to be embedded inside `VtEngine` are deleted in favor of using the ones in `til::bitmap`.
- Updated `VtEngine` tracing to use new `til::bitmap` on trace and the new `to_string()` methods detailed below.
- Comparison operators for `til::bitmap` and complementary tests.
- Fixed an issue where the dirty rectangle shortcut in `til::bitmap` was set to 0,0,0,0 by default which means that `|=` on it with each `set()` operation was stretching the rectangle from 0,0. Now it's a `std::optional` so it has no value after just being cleared and will build from whatever the first invalidated rectangle is. Complementary tests added.
- Optional run caching for `til::bitmap` in the `runs()` method since both VT and DX renderers will likely want to generate the set of runs at the beginning of a frame and refer to them over and over through that frame. Saves the iteration and creation and caches inside `til::bitmap` where the chance of invalidation of the underlying data is known best. It is still possible to iterate manually with `begin()` and `end()` from the outside without caching, if desired. Complementary tests added.
- WEX templates added for `til::bitmap` and used in tests.
- `translate()` method for `til::bitmap` which will slide the dirty points in the direction specified by a `til::point` and optionally back-fill the uncovered area as dirty. Complementary tests added.
- Moves all string generation for `til` types `size`, `point`, `rectangle`, and `some` into a `to_string` method on each object such that it can be used in both ETW tracing scenarios AND in the TAEF templates uniformly. Adds a similar method for `bitmap`.
- Add tagging to `_bitmap_const_iterator` such that it appears as a valid **Input Iterator** to STL collections and can be used in a `std::vector` constructor as a range. Adds and cleans up operators on this iterator to match the theoretical requirements for an **Input Iterator**. Complementary tests added.
- Add loose operators to `til` which will allow some basic math operations (+, -, *, /) between `til::size` and `til::point` and vice versa. Complementary tests added. Complementary tests added.
- Adds operators to `til::rectangle` to allow scaling with basic math operations (+, -, *) versus `til::size` and translation with basic math operations (+, -) against `til::point`. Complementary tests added.
- In-place variants of some operations added to assorted `til` objects. Complementary tests added.
- Update VT tests to compare invalidation against the new map structure instead of raw rectangles where possible.
## Validation Steps Performed
- Wrote additional til Unit Tests for all additional operators and functions added to the project to support this operation
- Updated the existing VT renderer tests
- Ran perf check
When we painted spaces up until the character right before the right
edge of the screen, we would erroneously use Erase in Line instead of
Erase Character due to an off-by-one.
Fixes#4727
## Summary of the Pull Request
This notifies automation clients (i.e.: NVDA, narrator, etc...) of new output being rendered to the screen.
## References
Close#2447 - Signaling for new output and cursor
Close#3791 - fixed by signaling cursor changes
## Detailed Description of the Pull Request / Additional comments
- Added tracing for UiaRenderer. This makes it easier to debug issues with notifying an automation client.
- Fire TextChanged automation events when new content is output to the screen.
## Validation Steps Performed
Verified with NVDA [1]
## Narrator
Narrator works _better_, but is unable to detect new output consistently. There is no harm for narrator when this change goes in.
[1] https://github.com/microsoft/terminal/issues/2447#issuecomment-595879890
Retrieved from https://microsoft.visualstudio.com os OS official/rs_onecore_dep_uxp 5b3acd8b5bac38da02fc86a29c81dfd252e79d1f
Related work items: MSFT:25505535
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
Always use the system's locale to render text to ensure the correct font variants are used.
`_ResolveFontFaceWithFallback()` overrides the last argument with the locale name of the font, but users normally configure fonts with latin alphabet only and use font linking to display non-latin characters, which causes the the locale names of the latin fonts are used to render the non-latin fonts. https://github.com/microsoft/terminal/issues/4508#issuecomment-598552472
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#4508
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [ ] I've discussed this with core contributors already. If not checked, I'm ready to accept this work might be rejected in favor of a different grand plan. Issue number where discussion took place: #xxx
## Validation
On a zh-hans system, simplified Chinese hans are used after this patch (above), versus Japanese hans before (below).
## Summary of the Pull Request
This changes the renederer to make sure to not draw the cursor when it is placed outside the viewport. When the window height isn't an exact multiple of a row's height, then there's a little bit of space below the last line we're actually drawing. If cursor is on the line below the viewport, then it can actually get drawn into this space. Since we're not drawing the text for that line, this is a little odd.
This PR fixes the issue by simply ensuring the cursor is in the veiwport before we draw it.
## References
## PR Checklist
* [x] Closes#3166
* [x] I work here
* [ ] Tests added/passed
* [n/a] Requires documentation to be updated
## Detailed Description of the Pull Request / Additional comments
## Validation Steps Performed
Checked with the GDI renderer as well as the DX renderer in conhost to make sure this is fixed for both of them, as well as the Terminal
cmatrix is somewhat of a pathological case for our infrastructure: it
prints out a bunch of green and white characters and then updates them a
million times a second.
It also maintains a column of space between every green character. When
it prints this column, it prints it in "default" or "white". This ends
up making runs of text that look like this:
(def: G=green B=bright white W=white *=matrix char =space)
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
G W G W G W G W G W G W G W G W
As characters trickle in:
G*W G*W G*W G*W G*W G*W G*W B*W
G*W G*W G*W G*W G*W G*W G*W G W
G*W G*W G*W B*W G*W G*W G*W G W
G*W B*W G*W G W G*W G*W G*W G*W
G*W G W G*W G W G*W B*W G*W G*W
B*W G W G*W G W G*W G W B*W G*W
G W G W G*W G W G*W G W G W B*W
G W G W B*W G W G*W G W G W G W
Every one of those color transitions causes us to break up the run of
text and start rendering it again. This impacts GDI, Direct2D *and*
ConPTY. In the example above, there are 120 runs.
The problem is, printing a space doesn't **use** the foreground color!
This commit introduces an optimization. When we're about to break a text
cluster because its attributes changed, we make sure that it's not just
filled with spaces and doesn't differ in any visually-meaningful way
(like underline or strikethrough, considering global invert state).
This lets us optimize both the rendering _and_ the PTY output to look
like this:
G* * * * * * * B*G
G* * * * * * *
G* * * B*G * * *
G* B*G * * * * *
G* * * B*G * *
B*G * * B*G *
G * * B*G
G B*G *
Text will be printed at best line-by-line and at worst only when the
visible properties of the screen actually change. In the example
above, there are only 21 runs.
This speeds up cmatrix remarkably.
Refs #1064
This PR adds support for "Resize with Reflow" to the Terminal. In
conhost, `ResizeWithReflow` is the function that's responsible for
reflowing wrapped lines of text as the buffer gets resized. Now that
#4415 has merged, we can also implement this in the Terminal. Now, when
the Terminal is resized, it will reflow the lines of it's buffer in the
same way that conhost does. This means, the terminal will no longer chop
off the ends of lines as the buffer is too small to represent them.
As a happy side effect of this PR, it also fixed#3490. This was a bug
that plagued me during the investigation into this functionality. The
original #3490 PR, #4354, tried to fix this bug with some heavy conpty
changes. Turns out, that only made things worse, and far more
complicated. When I really got to thinking about it, I realized "conhost
can handle this right, why can't the Terminal?". Turns out, by adding
resize with reflow, I was also able to fix this at the same time.
Conhost does a little bit of math after reflowing to attempt to keep the
viewport in the same relative place after a reflow. By re-using that
logic in the Terminal, I was able to fix#3490.
I also included that big ole test from #3490, because everyone likes
adding 60 test cases in a PR.
## References
* #4200 - this scenario
* #405/#4415 - conpty emits wrapped lines, which was needed for this PR
* #4403 - delayed EOL wrapping via conpty, which was also needed for
this
* #4354 - we don't speak of this PR anymore
## PR Checklist
* [x] Closes#1465
* [x] Closes#3490
* [x] Closes#4771
* [x] Tests added/passed
## EDIT: Changes to this PR on 5 March 2020
I learned more since my original version of this PR. I wrote that in
January, and despite my notes that say it was totally working, it
_really_ wasn't.
Part of the hard problem, as mentioned in #3490, is that the Terminal
might request a resize to (W, H-1), and while conpty is preparing that
frame, or before the terminal has received that frame, the Terminal
resizes to (W, H-2). Now, there aren't enough lines in the terminal
buffer to catch all the lines that conpty is about to emit. When that
happens, lines get duplicated in the buffer. From a UX perspective, this
certainly looks a lot worse than a couple lost lines. It looks like
utter chaos.
So I've introduced a new mode to conpty to try and counteract this
behavior. This behavior I'm calling "quirky resize". The **TL;DR** of
quirky resize mode is that conpty won't emit the entire buffer on a
resize, and will trust that the terminal is prepared to reflow it's
buffer on it's own.
This will enable the quirky resize behavior for applications that are
prepared for it. The "quirky resize" is "don't `InvalidateAll` when the
terminal resizes". This is added as a quirk as to not regress other
terminal applications that aren't prepared for this behavior
(gnome-terminal, conhost in particular). For those kinds of terminals,
when the buffer is resized, it's just going to lose lines. That's what
currently happens for them.
When the quirk is enabled, conpty won't repaint the entire buffer. This
gets around the "duplicated lines" issue that requesting multiple
resizes in a row can cause. However, for these terminals that are
unprepared, the conpty cursor might end up in the wrong position after a
quirky resize.
The case in point is maximizing the terminal. For maximizing
(height->50) from a buffer that's 30 lines tall, with the cursor on
y=30, this is what happens:
* With the quirk disabled, conpty reprints the entire buffer. This is
60 lines that get printed. This ends up blowing away about 20 lines
of scrollback history, as the terminal app would have tried to keep
the text pinned to the bottom of the window. The term. app moved the
viewport up 20 lines, and then the 50 lines of conpty output (30
lines of text, and 20 blank lines at the bottom) overwrote the lines
from the scrollback. This is bad, but not immediately obvious, and
is **what currently happens**.
* With the quirk enabled, conpty doesn't emit any lines, but the
actual content of the window is still only in the top 30 lines.
However, the terminal app has still moved 20 lines down from the
scrollback back into the viewport. So the terminal's cursor is at
y=50 now, but conpty's is at 30. This means that the terminal and
conpty are out of sync, and there's not a good way of re-syncing
these. It's very possible (trivial in `powershell`) that the new
output will jump up to y=30 override the existing output in the
terminal buffer.
The Windows Terminal is already prepared for this quirky behavior, so it
doesn't keep the output at the bottom of the window. It shifts it's
viewport down to match what conpty things the buffer looks like.
What happens when we have passthrough mode and WT is like "I would like
quirky resize"? I guess things will just work fine, cause there won't be
a buffer behind the passthrough app that the terminal cares about. Sure,
in the passthrough case the Terminal could _not_ quirky resize, but the
quirky resize won't be wrong.
When ConPTY encounters a string we don't understand, immediately flush the frame.
## References
This PR supersedes #2665. This solution is much simpler than what was proposed in that PR.
As mentioned in #2665: "This might have some long-term consequences for #1173."
## PR Checklist
* [x] Closes#2011
* [x] Closes#4106
* [x] I work here
* [x] Tests added/passed
## Summary of the Pull Request
Allows VT engine methods that print formatted strings (cursor movements, color changes, etc.) to provide a guess at the max buffer size required eliminating the double-call for formatting in the common case.
## PR Checklist
* [x] Found while working on #778
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed
* [x] Am core contributor.
## Detailed Description of the Pull Request / Additional comments
- The most common case for VT rendering is formatting a few numbers into a sequence. For the most part, we can already tell the maximum length that the string could be based on the number of substitutions and the size of the parameters.
- The existing formatting method would always double-call. It would first call for how long the string was going to be post-formatting, allocate that memory, then call again and fill it up. This cost two full times of running through the string to find a length we probably already knew for the most part.
- Now if a size is provided, we allocate that first and attempt the "second pass" of formatting directly into the buffer. This saves the count step in the common case.
- If this fails, we fall back and do the two-pass method (which theoretically means the bad case is now 3 passes.)
- The next biggest waste of time in this method was allocating and freeing strings for every format pass. Due to the nature of the VT renderer, many things need to be formatted this way. I've now instead moved the format method to hold a static string that really only grows over the course of the session for all of these format operations. I expect a majority of the time, it will only be consuming approximately 5-15 length of a std::string of memory space. I cannot currently see a circumstance where it would use more than that, but I'm consciously trading memory usage when running as a PTY for overall runtime performance here.
## Validation Steps Performed
- Ran the thing manually and checked it out with wsl and cmatrix and Powershell and such attached to Terminal
- Wrote and ran automated tests on formatting method
## Summary of the Pull Request
- Changes the `IRenderEngine` interface to return a vector of values instead of just a single one. Engines that want to report one still can. Engines that want to report multiple smaller ones will be able to do so going forward.
## PR Checklist
* [x] In support of differential rendering #778
* [x] I work here.
* [x] Manually tested it still works.
* [x] Am core contributor.
## Detailed Description of the Pull Request / Additional comments
- Some of my ideas for the `DxEngine` require the ability to specify multiple smaller rectangles instead of one giant one, specifically to mitigate the case where someone refreshes just one cell in two opposite corners of the display (which currently coalesces into refreshing the entire display.)
- This is pulled out into an individual PR to make it easier to review that concept changing.
## Validation Steps Performed
- Ran the Terminal
With certain font faces at certain sizes, the advances seem to be
slightly more than the pixel grid; Cascadia Code at 13pt (though, 200%
scale) had an advance of 10.000001.
This commit makes it so that anything sub-1/100 of a cell won't make us
break up runs, because doing so results in suboptimal rendering.
Fixes#4806.
## Summary of the Pull Request
Adjusts column padding code in `CustomTextLayout` to only pad out for surrogate pairs, not anything that reports two columns.
## References
- See also #4747
## PR Checklist
* [x] Closes#4780
* [x] I work here.
* [x] Manual tests.
* [x] No doc, this fixes code to match comment. Oops.
* [x] Am core contributor. Also discussed with @leonMSFT.
## Detailed Description of the Pull Request / Additional comments
For surrogate pairs like high Unicode emoji, we receive two wchar_ts but only one column count (which is usually 2 because emoji are usually inscribed in the full-width squares.) To compensate for this, I added in a little padding function at the top of the `CustomTextLayout` construction that adds a column of 0 aligned with the second half of a surrogate pair so the text-to-glyph mapping lines up correctly.
Unfortunately, I made a mistake while either responding to PR feedback in #4747 or in the first place and I made it pad out extra 0 columns based on the FIRST column count, not based on whether or not there is a trailing surrogate pair. The correct thing to do is to pad it out based on the LENGTH of text associated with the given column count. This means that full width characters which can be represented in one wchar_t, like those coming from the IME in most cases (U+5C41 for example) will have a column count of 2. This is perfectly correct for mapping text-to-glyphs and doesn't need a 0 added after it. A house emoji (U+1F3E0) comes in as two wchar_ts (0xD83C 0xDFE0) with the column count of 2. To ensure that the arrays are aligned, the 2 matches up with the 0xD83C but the 0xDFE0 needs a 0 on it so it will be skipped over. (Don't worry, because it's a surrogate, it's naturally consumed correctly by the glyph mapper.)
The effect was that every OTHER character inserted by the IME was scaled to 0 size (as an advance of 0 was expected for 0 columns).
The fix restores it so those characters don't have an associated count and aren't scaled.
## Validation Steps Performed
- Opened it up
- Put in the house emoji like #4747 (U+1f3e0)
- Put in some characters with simplified Chinese IME (fixed now)
- Put in the utf83.txt sample text used in #4747
## Summary of the Pull Request
- Improves the correction of the scaling and spacing that is applied to
glyphs if they are too large or too small for the number of columns that
the text buffer is expecting
## References
- Supersedes #4438
Co-authored-by: Mili (Yi) Zhang <milizhang@gmail.com>
- Related to #4704 (#4731)
## PR Checklist
* [x] Closes#696
* [x] Closes#4375
* [x] Closes#4708
* [x] Closes a crash that @DHowett-MSFT complained about with
`"x" * ($Host.UI.RawUI.BufferSize.Width - 1) + "`u{241b}"`
* [x] Eliminates an exception getting thrown with the U+1F3E0 emoji in
`_CorrectGlyphRun`
* [x] Corrects several graphical issues that occurred after #4731 was
merged to master (weird repeats and splits of runs)
* [x] I work here.
* [x] Tested manually versus given scenarios.
* [x] Documentation written into comments in the code.
* [x] I'm a core contributor.
## Detailed Description of the Pull Request / Additional comments
- The `_CorrectGlyphRun` function now walks through and uses the
`_glyphClusters` map to determine the text span and glyph span for each
cluster so it can be considered as a single unit for scaling.
- The total number of columns expected across the entire cluster
text/glyph unit is considered for the available spacing for drawing
- The total glyph advances are summed to see how much space they will
take
- If more space than necessary to draw, all glyphs in the cluster are
offset into the center and the extra space is padded onto the advance of
the last glyph in the range.
- If less space than necessary to draw, the entire cluster is marked for
shrinking as a single unit by providing the initial text index and
length (that is back-mapped out of the glyph run) up to the parent
function so it can use the `_SetCurrentRun` and `_SplitCurrentRun`
existing functions (which operate on text) to split the run into pieces
and only scale the one glyph cluster, not things next to it as well.
- The scale factor chosen for shrinking is now based on the proportion
of the advances instead of going through some font math wizardry
- The parent that calls the run splitting functions now checks to not
attempt to split off text after the cluster if it's already at the end.
This was @DHowett-MSFT's crash.
- The split run function has been corrected to fix the `glyphStart`
position of the back half (it failed to `+=` instead of `=` which
resulted in duplicated text, sometimes).
- Surrogate pair emoji were not allocating an appropriate number of
`_textClusterColumns`. The constructor has been updated such that the
trailing half of surrogate pairs gets a 0 column width (as the lead is
marked appropriately by the `GetColumns()` function). This was the
exception thrown.
- The `_glyphScaleCorrections` array stored up over the calls to
`_CorrectGlyphRun` now uses a struct `ScaleCorrection` as we're up to 3
values.
- The `ScaleCorrection` values are named to clearly indicate they're in
relation to the original text span, not the glyph spans.
- The values that are used to construct `ScaleCorrection`s within
`_CorrectGlyphRun` have been double checked and corrected to not
accidentally use glyph index/counts when text index/counts are what's
required.
## Validation Steps Performed
- Tested the utf82.txt file from one of the linked bugs. Looked
specifically at Burmese through Thai to ensure restoration (for the most
part) of the behavior
- Ensured that U+1f3e0 emoji (🏠) continues to draw correctly
- Checked Fixedsys Excelsior font to ensure it's not shrinking the line
with its ligatures
- Checked ligatureness of Cascadia Code font
- Checked combining characters U+0300-U+0304 with a capital A
## Summary of the Pull Request
Fix a bug where the `Renderer::PaintFrame` method:
1. is not called until the next `RenderThread::NotifyThread` call but needs to be called because there the terminal was updated (theoretical bug)
2. is called twice but needs to be called only once (verified bug)
## References
The bug was introduced by #3511.
## PR Checklist
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
## Detailed Description of the Pull Request / Additional comments
### Before
#### First bug
In the original code, `_fNextFrameRequested` is set to `true` in render thread because `std::atomic_flag::test_and_set` is called.
This is wrong because it means that the render thread will render the terminal again even if there is no change after the last render.
I think the the goal was to load the boolean value for `_fNextFrameRequested` to check whether the thread should sleep or not.
The problem is that there is no method on `std::atomic_flag` to load its boolean value. I guess what happened was that the "solution" that was found was to use `std::atomic_flag::test_and_set`, followed by `std::atomic_flag::clear` if the value was `false` originally (if `std::atomic_flag::test_and_set` returned `false`) to restore the original value. I guess that this was believed to be equivalent to just a simple load, without doing any change to the value because it restores it at the end.
But it's not: this is dangerous because if the value is changed to `true` between the call to `std::atomic_flag::test_and_set` and the call to `std::atomic_flag::clear`, then the value ends up being `false` at the end which is wrong because we don't want to change it! And if that value ends up being `false`, it means that we miss a render because we will wait on `_hEvent` during the next iteration on the render thread.
Well actually, here, this not even a problem because when that code is ran, `_fPainting` is `false` which means that the other thread that modifies the `_fNextFrameRequested` value through `RenderThread::NotifyPaint` will not actually modify `_fNextFrameRequested` but rather call `SetEvent` (see the method's body).
But wait! There is a problem there too! `std::atomic_flag::test_and_set` is called for `_fPainting` which sets its value to `true`. It was probably unintended. So actually, the next call to `RenderThread::NotifyPaint` _will_ end up modifying `_fNextFrameRequested` which means that the data race I was talking about _might_ happen!
#### Second bug
Let's go back a little bit in my explanation. I was talking about the fact that:
> I guess what happened was that the "solution" that was found was to use `std::atomic_flag::test_and_set`, followed by `std::atomic_flag::clear` if the value was `false` originally (if `std::atomic_flag::test_and_set` returned `false`) to restore the original value.
The problem is that the reverse was done in the implementation: `std::atomic_flag::clear` is called if the value was _`true`_ originally!
So at this point, if the value of `_fNextFrameRequested` was `false`, then `std::atomic_flag::test_and_set` sets its is set to `true` and returns `false`. So for the next iteration, `_fNextFrameRequested` is `true` and the render thread will re-render but that was not needed.
### After
I used `std::atomic<bool>` instead of `std::atomic_flag` for `_fNextFrameRequested` and the other atomic field because it has a `load` and a `store` method so we can actually load the value without changing it.
I also replaced `_fPainting` by `_fWaiting`, which is basically the opposite of `_fPainting` but stays `true` for a little shorter than `_fPainting` would stay `false`. Indeed, I think that it makes more sense to directly wrap/scope _just_ the call to `WaitForSingleObject` by setting my atomic variable to `true` _just_ before and to `false` _just_ after because:
* It makes more sense while you're reading the code: it's easier IMO to understand what the purpose of `_fWaiting` is (that is, to call `SetEvent` from `RenderThread::NotifyPaint` if it's `true`).
* It's probably a tiny bit better for performance because it will become `true` for a little shorter which means less calls to `SetEvent`.
#### Warning
I don't really understand [std::memory_order](https://en.cppreference.com/w/cpp/atomic/memory_order)s.
So I used the default one (`std::memory_order_seq_cst`) which is the safest.
I believe that if no read or write are reordered in the two threads (`RenderThread::NotifyPaint` and `RenderThread::_ThreadProc`), then the code I wrote will behave correctly.
I think that `std::memory_order_seq_cst` enforces that so it should be fine, but I'm not sure.
## Validation Steps Performed
**I tried to reproduce the second bug that I described in the first section of this PR.**
I put a breakpoint on `RenderThread::NotifyPaint` and on `Renderer::PaintFrame`. Initially they are disabled. Then I ran the terminal in Release mode, waited a bit for the prompt to display and the cursor to start blinking. Then I enabled the breakpoints.
### Before
Each `RenderThread::NotifyPaint` is followed by 2 `Renderer::PaintFrame` calls. ❌
### After
Each `RenderThread::NotifyPaint` is followed by 1 `Renderer::PaintFrame` call. ✔️
## Summary of the Pull Request
Changes how conpty emits text to preserve line-wrap state, and additionally adds rudimentary support to the Windows Terminal for wrapped lines.
## References
* Does _not_ fix (!) #3088, but that might be lower down in conhost. This makes wt behave like conhost, so at least there's that
* Still needs a proper deferred EOL wrap implementation in #780, which is left as a todo
* #4200 is the mega bucket with all this work
* MSFT:16485846 was the first attempt at this task, which caused the regression MSFT:18123777 so we backed it out.
* #4403 - I made sure this worked with that PR before I even sent #4403
## PR Checklist
* [x] Closes#405
* [x] Closes#3367
* [x] I work here
* [x] Tests added/passed
* [n/a] Requires documentation to be updated
## Detailed Description of the Pull Request / Additional comments
I started with the following implementation:
When conpty is about to write the last column, note that we wrapped this line here. If the next character the vt renderer is told to paint get is supposed to be at the start of the following line, then we know that the previous line had wrapped, so we _won't_ emit the usual `\r\n` here, and we'll just continue emitting text.
However, this isn't _exactly_ right - if someone fills the row _exactly_ with text, the information that's available to the vt renderer isn't enough to know for sure if this line broke or not. It is possible for the client to write a full line of text, with a `\n` at the end, to manually break the line. So, I had to also add the `lineWrapped` param to the `IRenderEngine` interface, which is about half the files in this changelist.
## Validation Steps Performed
* Ran tests
* Checked how the Windows Terminal behaves with these changes
* Made sure that conhost/inception and gnome-terminal both act as you'd expect with wrapped lines from conpty
## Summary of the Pull Request
- Surrogate pairs are being split in half with the run splitting check.
## References
- Related to #4708 but not going to fix it.
## PR Checklist
* [x] Closes#4704
* [x] I work here.
* [x] I am a core contributor.
## Detailed Description of the Pull Request / Additional comments
- The adjustment of the run heights in the correction function reports back a text index and a scaling factor. However, I didn't remember at the time that the text is being stored as UTF-16. So the index given can be pointing to the high surrogate of a pair. Thus adding 1 to split "after" the text character, then backing up by 1 isn't valid in if the index given was for a high surrogate.
The quick fix is to advance by two if it's a high surrogate and one otherwise.
## Validation Steps Performed
- Used the sample code from #4704 to print the house emoji in various situations into the buffer.
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
- Scale the retro terminal effects (#3468) scan lines with the screen's DPI.
- Remove artifacts from sampling wrap around.
Before & after, with my display scale set to 350%:
Before & after showing artifact removal, with my display scale set to 100%, and image enlarged to 400%:
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#4362
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [ ] I've discussed this with core contributors already. If not checked, I'm ready to accept this work might be rejected in favor of a different grand plan. Issue number where discussion took place: #xxx
<!-- Provide a more detailed description of the PR, other things fixed or any additional comments/features here -->
## Detailed Description of the Pull Request / Additional comments
Adds a constant buffer, which could be used for other settings for the retro terminal pixel shader.
I haven't touched C++ in over a decade before this change, and this is the first time I've played with DirectX, so please assume my code isn't exactly best practice. 🙂
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
- Changed display scale with experimental.retroTerminalEffect enabled, enjoyed scan lines on high resolution monitors.
- Enabled experimental.retroTerminalEffect, turned the setting off, changed display scale. Retro tabs still scale scan lines.
## Summary of the Pull Request
Adjusts `DrawGlyphRun` method inside DirectX renderer to restrict text
to be clipped within the boundaries of the row.
## PR Checklist
* [x] Closes#1703
* [x] I work here.
* [x] No tests.
* [x] No docs.
* [x] I am core contributor.
## Detailed Description of the Pull Request / Additional comments
For whatever reason, some of these shade glyphs near U+2591 tend to
extend way above the height of where we expect they should. This didn't
look like a problem in conhost because it clipped every draw inside the
bounds. This therefore applies the same clip logic as people don't
really expect text to pour out of the box.
It could, theoretically, get us into trouble later should someone
attempt zalgo text. But doing zalgo text is more of a silliness that
varies in behavior across rendering platforms anyway.
## Validation Steps Performed
- Ran the old conhost GDI renderer and observed
- Ran the new Terminal DX renderer and observed
- Made the code change
- Observed that the height and approximate display characteristics of
the U+2591 shade and neighboring characters now matches with the conhost
GDI style to stay within its lane.
## Summary of the Pull Request
- Height adjustment of a glyph is now restricted to itself in the DX
renderer instead of applying to the entire run
- ConPTY compensates for drawing the right half of a fullwidth
character. The entire render base has this behavior restored now as
well.
## PR Checklist
* [x] Closes#2191
* [x] I work here
* [x] Tests added/passed
* [x] No doc
* [x] Am core contributor.
## Detailed Description of the Pull Request / Additional comments
Two issues:
1. On the DirectX renderer side, when confronted with shrinking a glyph,
the correction code would apply the shrunken size to the entire run, not
just the potentially individual glyph that needed to be reduced in size.
Unfortunately while adjusting the horizontal X width can be done for
each glyph in a run, the vertical Y height has to be adjusted for an
entire run. So the solution here was to split the individual glyph
needing shrinking out of the run into its own run so it can be shrunk.
2. On the ConPTY side, there was a long standing TODO that was never
completed to deal with a request to draw only the right half of a
two-column character. This meant that when encountering a request for
the right half only, we would transmit the entire full character to be
drawn, left and right halves, struck over the right half position. Now
we correct the cursor back a position (if space) and draw it out so the
right half is struck over where we believe the right half should be (and
the left half is updated as well as a consequence, which should be OK.)
The reason this happens right now is because despite VIM only updating
two cells in the buffer, the differential drawing calculation in the
ConPTY is very simplistic and intersects only rectangles. This means
from the top left most character drawn down to the row/col cursor count
indicator in vim's modeline are redrawn with each character typed. This
catches the line below the edited line in the typing and refreshes it.
But incorrectly.
We need to address making ConPTY smarter about what it draws
incrementally as it's clearly way too chatty. But I plan to do that with
some of the structures I will be creating to solve #778.
## Validation Steps Performed
- Ran the scenario listed in #2191 in vim in the Terminal
- Added unit tests similar to examples given around glyph/text mapping
in runs from Microsoft community page
## Summary of the Pull Request
Adds an ETW provider for tracing out operations inside the DirectX Renderer.
## References
This supports #2191 and #778 and other rendering issues.
## PR Checklist
* [x] I work here
## Detailed Description of the Pull Request / Additional comments
This declares and defines the provider with the a GUID appropriate for the namespace and adds an initial invalidation rectangle method for figuring out what is being drawn mostly to understand what could be differential.
## Validation Steps Performed
Implemented provider.
Opened real-time ETW tracing tool
Ran the Terminal and watched invalidation events appear live
Reduce unnecessary SetEvent CPU overhead by using `std::atomic_flag` to avoid setting kernel events when the painting thread is already awake and running.
## Summary of the Pull Request
- If no one is listening to the ETW provider for the VT Renderer for diagnostic purposes, do not spend time allocating/deleting/formatting strings for presentation in TraceLogging messages.
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes something I noticed while working on the renderer.
* [x] I work here.
* [x] Existing tests should pass
* [x] No doc
* [x] Am core contributor.
## Validation Steps Performed
WPR/WPA
Before: 321/3016 samples on hot path (10.64%)
After: 0/1266 samples on the same path (0%)
