This commit introduces "AtlasEngine", a new text renderer based on DxEngine.
But unlike it, DirectWrite and Direct2D are only used to rasterize glyphs.
Blending and placing these glyphs into the target view is being done using
Direct3D and a simple HLSL shader. Since this new renderer more aggressively
assumes that the text is monospace, it simplifies the implementation:
The viewport is divided into cells, and its data is stored as a simple matrix.
Modifications to this matrix involve only simple pointer arithmetic and is easy
to understand. But just like with DxEngine however, DirectWrite
related code remains extremely complex and hard to understand.
Supported features:
* Basic text rendering with grayscale AA
* Foreground and background colors
* Emojis, including zero width joiners
* Underline, dotted underline, strikethrough
* Custom font axes and features
* Selections
* All cursor styles
* Full alpha support for all colors
* _Should_ work with Windows 7
Unsupported features:
* A more conservative GPU memory usage
The backing texture atlas for glyphs is grow-only and will not shrink.
After 256MB of memory is used up (~20k glyphs) text output
will be broken until the renderer is restarted.
* ClearType
* Remaining gridlines (left, right, top, bottom, double underline)
* Hyperlinks don't get full underlines if hovered in WT
* Softfonts
* Non-default line renditions
Performance:
* Runs at up to native display refresh rate
Unfortunately the frame rate often drops below refresh rate, due us
fighting over the buffer lock with other parts of the application.
* CPU consumption is up to halved compared to DxEngine
AtlasEngine is still highly unoptimized. Glyph hashing
consumes up to a third of the current CPU time.
* No regressions in WT performance
VT parsing and related buffer management takes up most of the CPU time (~85%),
due to which the AtlasEngine can't show any further improvements.
* ~2x improvement in raw text throughput in OpenConsole
compared to DxEngine running at 144 FPS
* ≥10x improvement in colored VT output in WT/OpenConsole
compared to DxEngine running at 144 FPS
This test has some load bearing specifics in it.
In short, it must trigger multiple boost::small_vector growths. The
vector is reused for the lifetime of the console (!) so it must be
forced to act the way the repro requires.
Oh, and the test needs to be run with Application Verifier +Heap.
This validates the fix in c0ab9cb5b5.
The ES folks removed support for implicit binary dependencies in test projects, which required our testmd to change to include its dependency on testnetv5.
Retrieved from https://microsoft.visualstudio.com os.2020 OS official/rs_wdx_dxp_windev a2a74747feda018c347ba787e166e8385bb17e47
Related work items: MSFT-31480819
A bunch of our feature tests don't work reliably in CI. They rely on
creating a new `OpenConsole.exe` window, then running the test _in that
console_. As a part of that test setup, the test runner used to wait a
second to attach to the newly created console. Then the test goes on
it's merry way, assuming the console is ready to go. However, in CI,
that might take more than a second. If it does, then the test would fail
pretty immediately, as soon as it tries to get at the buffer of the new
console.
This PR introduces a little retry loop to the test init. After attaching
to the new console, we'll try and get at the screen buffer. If that
fails, we'll wait a second and try again. We'll try 5 times total,
before bailing entirely. Hopefully, this should mitigate most of the
random CI failures we get in the feature tests.
Closes#8495
Many include statements use forward slashes, while others use backwards
slashes. This is inconsistent formatting. For this reason, I changed the
backward slashes to forward slashes since that is the standard.
This PR provides a faster algorithm for converting 8-bit and 24-bit
colors into the 4-bit legacy values that are required by the Win32
console APIs. It also fixes areas of the code that were incorrectly
using a simple 16-color conversion that didn't handle 8-bit and 24-bit
values.
The faster conversion algorithm should be an improvement for issues #783
and #3950.
One of the main points of this PR was to fix the
`ReadConsoleOutputAttribute` API, which was using a simplified legacy
color conversion (the original `TextAttribute:GetLegacyAttributes`
method), which could only handle values from the 16-color table. RGB
values, and colors from the 256-color table, would be mapped to
completely nonsensical values. This API has now been updated to use the
more correct `Settings::GenerateLegacyAttributes` method.
But there were also a couple of other places in the code that were using
`GetLegacyAttributes` when they really had no reason to be working with
legacy attributes at all. This could result in colors being downgraded
to 4-bit values (often badly, as explained above), when the code was
already perfectly capable of displaying the full 24-bits.
This included the fill colors in the IME composer (in `ConsoleImeInfo`),
and the construction of the highlighting colors in the color
search/selection handler (`Selection::_HandleColorSelection`). I also
got rid of some legacy attribute code in the `Popup` class, which was
originally intended to update colors below the popup when the settings
changed, but actually caused more problems than it solved.
The other major goal of this PR was to improve the performance of the
`GenerateLegacyAttributes` method, since the existing implementation
could be quite slow when dealing with RGB values.
The simple cases are handled much the same as they were before. For an
`IsDefault` color, we get the default index from the
`Settings::_wFillAttribute` field. For an `IsIndex16` color, the index
can just be returned as is.
For an `IsRgb` color, the RGB components are compressed down to 8 bits
(3 red, 3 green, 2 blue), simply by dropping the least significant bits.
This 8-bit value is then used to lookup a representative 16-color value
from a hard-coded table. An `IsIndex256` color is also converted with a
lookup table, just using the existing 8-bit index.
The RGB mapping table was calculated by taking each compressed 8-bit
color, and picking a entry from the _Campbell_ palette that best
approximated that color. This was done by looking at a range of 24-bit
colors that mapped to the 8-bit value, finding the best _Campbell_ match
for each of them (using a [CIEDE2000] color difference calculation), and
then the most common match became the index that the 8-bit value would
map to.
The 256-color table was just a simpler version of this process. For each
entry in the table, we take the default RGB palette value, and find it's
closest match in the _Campbell_ palette.
Because these tables are hard-coded, the results won't adjust to changes
in the palette. However, they should still produce reasonable results
for palettes that follow the standard ANSI color range. And since
they're only a very loose approximation of the colors anyway, the exact
value really isn't that important.
That said, I have tried to make sure that if you take an RGB value for a
particular index in a reasonable color scheme, then the legacy color
mapped from that value should ideally match the same index. This will
never be possible for all color schemes, but I have tweaked a few of the
table entries to improve the results for some of the common schemes.
One other point worth making regarding the hard-coded tables: even if we
wanted to take the active palette into account, that wouldn't actually
be possible over a conpty connection, because we can't easily know what
color scheme the client application is using. At least this way the
results in conhost are guaranteed to be the same as in the Windows
Terminal.
[CIEDE2000]: https://en.wikipedia.org/wiki/Color_difference#CIEDE2000
## Validation Steps Performed
This code still passes the `TextAttributeTests` that check the basic
`GetLegacyAttribute` behaviour and verify the all legacy attributes
roundtrip correctly. However, some of the values in the `RgbColorTests`
had to be updated, since we're now intentionally returning different
values as a result of the changes to the RGB conversion algorithm.
I haven't added additional unit tests, but I have done a lot of manual
testing to see how well the new algorithm works with a range of colors
and a variety of different color schemes. It's not perfect in every
situation, but I think it works well enough for the purpose it serves.
I've also confirmed that the issues reported in #5940 and #6247 are now
fixed by these changes.
Closes#5940Closes#6247
## Summary of the Pull Request
Despite being specified as `noexcept`, `FillConsoleOutputCharacterA` emits an exception when a call to `ConvetToW` is made with an argument character which can't be converted. This PR fixes this throw, by wrapping `ConvertToW` in a try-catch_return.
## PR Checklist
* [x] Closes#4258
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed: thanks @miniksa
## Detailed Description of the Pull Request / Additional comments
Following the semantics of other `FillConsoleOutputCharacter*` the output param `cellsModified` is set to `0`. The try-catch_return is also what other functions of this family perform in case of errors.
## Validation Steps Performed
Original repro no longer crashes.
Generated by https://github.com/jsoref/spelling `f`; to maintain your repo, please consider `fchurn`
I generally try to ignore upstream bits. I've accidentally included some items from the `deps/` directory. I expect someone will give me a list of items to drop, I'm happy to drop whole files/directories, or to split the PR into multiple items (E.g. comments/locals/public).
Closes#4294
Moves the tests from using the `vstest.console.exe` route to just using `te.exe`.
PROs:
- `te.exe` is significantly faster for running tests because the TAEF/VSTest adapter isn't great.
- Running through `te.exe` is closer to what our developers are doing on their dev boxes
- `te.exe` is how they run in the Windows gates.
- `te.exe` doesn't seem to have the sporadic `0x6` error code thrown during the tests where somehow the console handles get lost
- `te.exe` doesn't seem to repro the other intermittent issues that we have been having that are inscrutable.
- Fewer processes in the tree (te is running anyway under `vstest.console.exe`, just indirected a lot
- The log outputs scroll live with all our logging messages instead of suppressing everything until there's a failure
- The log output is actually in the order things are happening versus vstest.
CONs:
- No more code coverage.
- No more test records in the ADO build/test panel.
- Tests really won't work inside Visual Studio at all.
- The log files are really big now
- Testing is not a test task anymore, just another script.
Refuting each CON:
- We didn't read the code coverage numbers
- We didn't look at the ADO test panel results or build-over-build velocities
- Tests didn't really work inside Visual Studio anyway unless you did the right incantations under the full moon.
- We could tone down the logging if we wanted at either the te.exe execution time (with a switch) or by declaring properties in the tests/classes/modules that are very verbose to not log unless it fails.
- I don't think anyone cares how they get run as long as they do.
<!-- 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
It's 2020 now. It's *about* time that we move on from 1990's macros.
<!-- 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
* [X] Closes#4152
* [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
Remove global namespaced min/max and replace it with STL min/max.
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
Run it.
* Change to use App CRT in preparation for universal.
* Try to make project build again by setting winconpty to static lib so it'll use the CRT inside TerminalConnection (or its other consumers) instead of linking its own.
* Remove test for conpty dll, it's a lib now. Add additional commentary on how CRT linking works for future reference. I'm sure this will come up again.
* use the _apiset variant until proven otherwise to match the existing one.
* Clarification in the comments for linking.
This commit cleans up and deduplicates all of the common build
preamble/postamble across exe, dll, lib and c++/winrt projects.
The following specific changes have been made:
* All projects now define their ConfigurationType
* All projects now set all their properties *before* including a common
build file (or any other build files)
* cppwinrt.pre and cppwinrt.post now delegate most of their
configuration to common.pre and common.post
* (becuase of the above,) all build options are conserved between
console and c++/winrt components, including specific warnings and
preprocessor definitions.
* More properties that are configurable per-project are now
conditioned so the common props don't override them.
* The exe, dll, exe.or.dll, and lib postincludes have been merged into
pre or post and switched based on condition as required
* Shared items (-shared, -common) are now explicitly vcxitems instead of
vcxproj files.
* The link line is now manipulated after Microsoft.Cpp sets it, so the
libraries we specify "win". All console things link first against
onecore_apiset.lib.
* Fix all compilation errors caused by build unification
* Move CascadiaPackage's resources into a separate item file
Fixes#922.
* Ensure rights check and increments pass before assigning an object to a handle (since deallocation of handles will automatically attempt to cleanup).
EraseInLine calls `FillConsoleOutputCharacterW()`. In filling the row with
chars, we were setting the wrap flag. We need to specifically not do this on
ANY _FILL_ operation. Now a fill operation UNSETS the wrap flag if we fill to
the end of the line.
Originally, we had a boolean `setWrap` that would mean...
- **true**: if writing to the end of the row, SET the wrap value to true
- **false**: if writing to the end of the row, DON'T CHANGE the wrap value
Now we're making this bool a std::optional to allow for a ternary state. This
allows for us to handle the following cases completely. Refer to the table
below:
,- current wrap value
| ,- are we filling the last cell in the row?
| | ,- new wrap value
| | | ,- comments
|-- |-- |-- |
| 0 | 0 | 0 |
| 0 | 1 | 0 |
| 0 | 1 | 1 | THIS CASE WAS HANDLED CORRECTLY
| 1 | 0 | 0 | THIS CASE WAS UNHANDLED
| 1 | 0 | 1 |
| 1 | 1 | 1 |
To handle that special case (1-0-0), we need to UNSET the wrap. So now, we have
~setWrap~ `wrap` mean the following:
- **true**: if writing to the end of the row, SET the wrap value to TRUE
- **false**: if writing to the end of the row, SET the wrap value to FALSE
- **nullopt**: leave the wrap value as it is
Closes#1126
Adjusts the startup and shutdown behavior of most threads in the console host to alleviate race conditions that are either exacerbated or introduced by the VT PTY threads.
This pull request introduces a copy of the code from kernel32.dll that
implements CreatePseudoConsole, ClosePseudoConsole and
ResizePseudoConsole. Apart from some light modifications to fit into the
infrastructure in this project and support launching OpenConsole.exe, it
is intended to be 1:1 with the code that ships in Windows.
Any guideline violations in this code are likely intentional. Since this
was built into kernel32, it uses the STL only _very sparingly._
Consumers of this library must make sure that conpty.lib lives earlier
in the link line than onecoreuap_apiset, onecoreuap, onecore_apiset,
onecore or kernel32.
Refs #1130.
[Git2Git] Git Train: Merge of building/rs_onecore_dep_uxp/190820-1847 into official/rs_onecore_dep_uxp Retrieved from https://microsoft.visualstudio.com os OS official/rs_onecore_dep_uxp 73e964d4046c37df3030970cae1ae32e83103fb5
(cherry picked from commit 8c63dff982093db1af7e2bb46b49af884dfec0c5)
Flushes the input queue on RawReadUnpacksCoalescedInputRecords test to ensure that other tests cannot cause failure by leaving extraneous input records behind after they run.
This only failed in the core operating system gate tests. This is because those tests run a subset of the complete test suite (subtracting the ones that do not make sense in a core environment). Apparently one of the tests that was skipped that normally runs prior to the UnpacksCoalesced test ensured that the input queue was clean enough for this test to succeed. But in the core environment, the test that ran prior left stuff behind.
To resolve this, I'm making the Coalesced test more resilient by cleaning out the queue prior to performing its operations.
(Also, bonus, I'm fixing the typo in the name Coalesced.)
This is less complicated/expensive than tracking down the tests that are leaving garbage behind, should prevent issues in the future related to ordering (since the tests run alphabetically, by default), and isn't as expensive as running the test in isolation (with its own conhost stood up for just the one test.)
Validated by running te.exe Microsoft.Console.Host.FeatureTests.dll /name:*InputTests* against a core operating system variant. Prior to change, this test failed. After the change, this test succeeded.
This will be automatically double-checked by the gates run after check-in.
Almost all functions in the Windows API that open or create objects and return HANDLEs to them return null on failure; only a few (mostly to do with the file system) return INVALID_HANDLE_VALUE on failure. This PR scrubs the repo of a few, but not necessarily all, cases where INVALID_HANDLE_VALUE was mistakenly used or tested against instead of null. In particular, it fixes 2 cases reported in issue #427 where the return value of CreateThread() was compared against INVALID_HANDLE_VALUE against null, causing the error handling code to run at the wrong time.
There are a lot of other uses of INVALID_HANDLE_VALUE I found that looked questionable, but which I left alone. Most of these were used to initialize HANDLE-typed variables and as a sentinel to see if those variables remained unset to a "real" value.
Fixes#427
* Removed using namespace directive from header files and put these in cpp files where they are used
* Fixed tabbing issues by replacing them with spaces.
Also regrouped the using directives.
* Update src/host/exemain.cpp
Co-Authored-By: Mike Griese <migrie@microsoft.com>
* Update src/interactivity/win32/find.cpp
Co-Authored-By: Mike Griese <migrie@microsoft.com>
This encompasses a handful of problems with column counting.
The Terminal project didn't set a fallback column counter. Oops. I've fixed this to use the `DxEngine` as the fallback.
The `DxEngine` didn't implement its fallback method. Oops. I've fixed this to use the `CustomTextLayout` to figure out the advances based on the same font and fallback pattern as the real final layout, just without "rounding" it into cells yet.
- `CustomTextLayout` has been updated to move the advance-correction into a separate phase from glyph shaping. Previously, we corrected the advances to nice round cell counts during shaping, which is fine for drawing, but hard for column count analysis.
- Now that there are separate phases, an `Analyze` method was added to the `CustomTextLayout` which just performs the text analysis steps and the glyph shaping, but no advance correction to column boundaries nor actual drawing.
I've taken the caching code that I was working on to improve chafa, and I've brought it into this. Now that we're doing a lot of fallback and heavy lifting in terms of analysis via the layout, we should cache the results until the font changes.
I've adjusted how column counting is done overall. It's always been in these phases:
1. We used a quick-lookup of ranges of characters we knew to rapidly decide `Narrow`, `Wide` or `Invalid` (a.k.a. "I dunno")
2. If it was `Invalid`, we consulted a table based off of the Unicode standard that has either `Narrow`, `Wide`, or `Ambiguous` as a result.
3. If it's still `Ambiguous`, we consult a render engine fallback (usually GDI or now DX) to see how many columns it would take.
4. If we still don't know, then it's `Wide` to be safe.
- I've added an additional flow here. The quick-lookup can now return `Ambiguous` off the bat for some glyph characters in the x2000-x3000 range that used to just be simple shapes but have been retroactively recategorized as emoji and are frequently now using full width color glyphs.
- This new state causes the lookup to go immediately to the render engine if it is available instead of consulting the Unicode standard table first because the half/fullwidth table doesn't appear to have been updated for this nuance to reclass these characters as ambiguous, but we'd like to keep that table as a "generated from the spec" sort of table and keep our exceptions in the "quick lookup" function.
I have confirmed the following things "just work" now:
- The windows logo flag from the demo. (⚫⚪💖✅🌌😊)
- The dotted chart on the side of crossterm demo (•)
- The powerline characters that make arrows with the Consolas patched font (██)
- An accented é
- The warning and checkmark symbols appearing same size as the X. (✔⚠🔥)
Related work items: #21167256, #21237515, #21243859, #21274645, #21296827
