This should make it more obvious from the error what's gone wrong if we see a repeat of #19096 (previously the error was something like "Bad state: future already completed").
* Rename util -> test_utils
* Rename flutter_test_driver -> test_driver
* Switch testWithContext -> test
* Remove unused import
* Move test project into a class to make it easier to have multiple of these
Each "TestProject" class can contain its files and things like named breakpoint locations.
* Split expression evaluation tests into own file
* Include last response in error messages
* Update expectations based on current bugs
* Fix async-ness in tests
* Fix incorrect expectation in test
* Fix incorrect evaluations
* Remove skips for tests that are now passing on master
* Expect pass on Linux
🤷♂️
* Call the code
* Skip expression evaluation tests on Windows
* Skip whole group, not just one test
* Remove duplicated method from merge
* Fix misplaced close of group
* Remove code that was duplicated from test we copied
Not sure how this ended up in here?
* Re-fix typo
A significant amonut of code uses defaultTargetPlatform, so it's
currently impossible to run most apps on Windows. This adds a mapping
from Windows as a host to Android as a target, paralleling the
macOS->iOS and Linux->Android mappings.
This allows more use of Windows as a host platform (e.g., for testing,
as in issue #17768).
This will let us re-use that code from fuchsia_tester.dart.
Tested by running the stocks example tests with coverage collection before and after.
.lcov files are identical.
* Adding a setup function for hot restart that can be overwritten to perform any necessary initialization before a restart is started.
* Adding a test for setupHotRestart
* Removing local pubspec change
* Fixing some type annotation errors
* Responding to comments
* fixed segmented control golden test
* fixed segmented control golden test
* added cursorWidth, cursorRadius
* added default value for cursorWidth based on Apple specs
* test default cursorWidth
* removed cursorHeight stuff
* added functionality to keep cursor from blinking
* cursor width and radius is configurable + tests
* changed goldens repo version in goldens.version
* working version of configurable cursor (erased debugKeepCursorOn)
* minor changes
* docs
* changed textfield test that was failing due to new default cursorwidth
* added default value of cursorwidth in RenderEditable
* only run golden file tests on Mac
* cursor tests
* the tests are actually there now
* weak warning fixed
* switching to Linux
* changed default cursorWidth: 2.0 -> 1.0
* assorted changes, including changing text field test
* re-paint -> re-layout when changing cursorWidth
* Upgrade to current LTS version of Node
The version of NPM with Node 6 is over a year old and frequently hangs trying to install Firebase tools.
* Temporarily comment out this condition so the install runs on Travis
* Revert "Temporarily comment out this condition so the install runs on Travis"
This reverts commit 74db9366b4b1b094369388ef50fb26f3b7fd3104.
* Pin to nodejs v8.1
* Include stdout/stderr in failure messages
Sometimes some of these tests unexpectedly fail with a non-zero exit code. This ensures stdout/stderr is included in the test failure message when this happens so that we can track down the issue.
* Remove redundant info about exit code
* Remove unnecessary indenting
The [markers] make it fairly clear so this just makes the test code noisy.
This logic is described in the test as looking for a scroll ending
very close to a new page, but in fact its behavior is more like
"very close to a page to the right": if we're not very, very close
to any page, it will pick the page to the left, not an old page.
There's no reason this should be left-right asymmetrical.
Instead, pick the nearest page.
In practice, the case where this makes a difference never arises when
the scroll runs undisturbed to completion; but when the user taps on
the page to hold or drag, the scroll will be interrupted before it
gets within tolerance of a particular page, and this case does arise.
This fixes a glitch that is hard to trigger without time dilation,
but is quite conspicuous with it:
* Open a tab view with at least 4 tabs, e.g. the Buttons screen
of the gallery (with "Animate Slowly" on.)
* Starting at tab 0, tap tab 2.
* When the animation is nearly complete, tap the page a couple
of times, as if to drag it around to scroll. Then let the
page view settle ballistically toward page 2.
* Before it finishes, tap tab 3.
* Suddenly page 1 fills the view, replacing page 2, before we
scroll from there to page 3.
With this fix, the animation in the last step moves smoothly from
where we are when it starts onward to page 3.
