Once upon a long time ago we decided to mark the properties in the
UITextInputTraits protocol as required in our API definition, because that way
we'd inline these properties in any class that implemented the
UITextInputTraits protocol, which made calling these properties much easier.
At a later point, we implemented better support for protocols, and now we
automatically generate extension methods for such properties (a corresponding
Get/Set method for the get/set property accessors), so we don't need these
inlined properties anymore.
However, removing them would be a breaking change, so we were stuck with these
redundant inlined properties, until .NET came along.
Ref: 0e80570863
Fixes https://github.com/xamarin/xamarin-macios/issues/5831.
Take into account any Bind attributes on optional property getters and setters
on protocols when generating the Get* and Set* extension methods, so that we
use the right selector.
Fixes https://github.com/xamarin/xamarin-macios/issues/12727.
Don't add FileNativeReferences to the main libraries to link with, because we
pass that list of main libraries to the LinkNativeCodeTask, and we're already
passing the FileNativeReferences for a different task parameter.
This means that we end up adding the file native reference twice to the linker
arguments, and that's wasteful. It can also cause problems if those linker
arguments aren't always computed in the same way (once as a relative path,
once as an absolute path for instance).
Fixes https://github.com/xamarin/xamarin-macios/issues/13503.
Once upon a time there was a single VTCompressionSession.Create method, which
was [driving users insane][1] - they had to manually call CFRetain to avoid
crashes! What an abomination!!
Insane users are clearly not happy users, and we wanted happy users, so time
and effort went into creating a solution: a new Create overload was devised
and [implemented][1], taking extreme care to not break our brave and insane
existing users who had to manually call CFRetain. Because the fix would break
existing users - the now extraneous CFRetain would mean that their apps would
leak memory. *A lot* of it! That's bad, so we decided to make sure that didn't
happen.
Of course, dear old Murhpy wanted a say, so the new Create overload didn't do
as intended. In fact, it had the same insane behavior the old Create overload
had! Ops.
But Murphy decided to have even more fun: the changes were so buggy, that they
in fact fixed the old Create overload! Which from now on wouldn't require the
horrendous manual CFRetain calls... and effectively introducing the leak the
fix was trying so hard to not introduce.
Oh dear Murphy.
Of course he had another trick up his sleeve: in our extreme efforts to help
our users, we added an Obsolete attribute that would tell people to use the
new Create overload.
Let that sink in for a moment: we had an Obsolete attribute on a function that
was (now) perfectly fine, telling users to use a function that was broken.
To get the correct behavior, users would now have to to remove their manual
CFRetain calls, and ignore the obsolete warning on the old (and correct)
Create overload which told users to use the new (and buggy) Create overload.
In other words: still insanity, just a slightly different flavor.
Murphy had a field day!
Time went by, and eventually a sane enough user [reported the insanity to
us][2]. Even better: the user actually provided a fix! Truly, we have some
amazing users.
Unfortunately, the user didn't have access to our code history, and thus was
obviously not able to see the whole picture, and the fix ended up being
incorrect.
Unrelated lesson learned: don't forget your history, otherwise you'll end up
repeating mistakes from the past.
So now came the problem: how to fix all the APIs? In a way that didn't make
our users' existing apps just suddenly start crashing or leaking?
There really was no way, so nothing really happened for quite a while.
Then, an opportunity presented itself: we'd be able to do [widespread breaking
changes][3].
So, hoping that Murphy stays away this sunny winter day, I'm changing both the
new and the old Create overloads to do the right thing. But only in .NET,
where we can do breaking changes! Or at least that's my intention. I've tried
to stave off our dear old friend by adding his arch enemy: unit tests. Which,
of course, Murphy couldn't stay away from, but it seems adding a few
Thread.Sleep calls makes him bored enough to stay away. Hopefully for good...
[1]: 66c50b9a17
[2]: https://github.com/xamarin/xamarin-macios/pull/2070
[3]: https://github.com/xamarin/xamarin-macios/issues/13087
* Capture evaluation output, and write it all to the terminal when something
goes wrong. This way we can see the entire output next to the failure (often
there's a lot of stuff written to the terminal from different threads, and
this way we get all that matters written together).
* Only evaluate one project at a time, to avoid overloading the machine.
* Only execute `git ls-files` once at a time, to avoid overloading the machine.
* Bump evaluation timeout to 5 minutes.
* Also increase the time for git to list files in a directory to 60 seconds.
Hopefully this will fix errors like this:
* `Unable to evaluate the property OutputPath, build failed with exit code 0. Timed out: True`
* `System.Exception: Failed to list the files in the directory /Users/builder/azdo/_work/1/s/xamarin-macios/tests/xharness/tmp-test-dir/monotouch-test3403 (TimedOut: True ExitCode: 0)`
This fixes a problem where we'd build the same project reference from
dotnet-shared.csproj in parallel, and each build would stomp on eachother
(because we'll now clone the project references in dotnet-shared.csproj).
This als required updating project files to use MSBuildThisFileDirectory
instead of MSBuildProjectDirectory, which makes it easier for xharness to
inline/process these files, because MSBuildThisFileDirectory is easy to know
when processing a file, while MSBuildProjectDirectory depends on the calling
project, which complicates matters significantly.
A fix in MonoTouch.Dialog was also required.
New commits in migueldeicaza/MonoTouch.Dialog:
* migueldeicaza/MonoTouch.Dialog@59fbf5b [dotnet] Shared project files don't need the DefaultTargets/ToolsVersion/xmlns attributes.
Diff: 4d0e0a9a5f..59fbf5bb1b
Fixes https://github.com/xamarin/maccore/issues/2527.
If NSFunctionKey isn't in Mac Catalyst in legacy Xamarin, it shouldn't be in
.NET either, so adjust the conditional logic accordingly.
Also make the NSFunctionKey enum a non-native enum in .NET, like it's in the
headers.
Remove Runtime.Arch and ObjCRuntime.Arch from Mac Catalyst, because they don't
apply for a Mac Catalyst app (which is neither a simulator environment, nor a
device environment).
This means that code using these APIs will have to be re-evaluated to
determine what's the correct behavior for Mac Catalyst.
Also update our tests accordingly.
Fixes https://github.com/xamarin/xamarin-macios/issues/10312.
* Change dotnet-linker to only care about whether we're actually trimming anything or not.
* Allow LinkMde/MtouchLink to not be set if TrimMode is set.
* Detect if any assemblies are linked or not by checking the global TrimMode
property + any TrimMode properties on assemblies.
Fixes https://github.com/xamarin/xamarin-macios/issues/13518.
named 'Info.plist', and assume that's the app manifest.
That doesn't quite work when we end up with multiple 'Info.plist' entries in any
of those item groups (one example being a framework as a BundleResource - all frameworks
have an Info.plist, and there's no good way to distinguish what the developer's intention
was).
So:
1. Implement a 'AppManifestDetectionEnabled' property to disable automatic app manifest
detection.
2. Add a public 'AppBundleManifest' property that specifies the app manifest
(this is just a renamed version of our previously private '_AppManifest' property).
This makes it possible for app developers to:
* Disable automatic app manifest detection.
* Still have an app manifest by specifying it manually.
* Disable automatic app manifest detection, but also not specify an app manifest
manually (so no custom app manifest at all).
Also:
* Rename '_AppBundleManifest' to '_AppBundleManifestPath' to make it less confusing
with the new 'AppBundleManifest' property.
Symlinks to directories are treated the same as other symlinks (as files), not
as directories. This way we don't end up re-creating a directory hierarchy
when we only have to create a symlink.
* Change all XAMCORE_4_0 conditions to NET conditions.
* Add numerous Release attributes that xtro started complaining about.
* Misc other minor API changes/updates.
We are in a situation where:
1. .NET MAUI is still in preview
2. We need dotnet/runtime fixes for MAUI, but we don't necessarily want all fixes to go into the .NET 6 service release.
The solution is to simplify use different builds/packs from dotnet/runtime.
Ref: https://github.com/xamarin/xamarin-android/pull/6542
* [tools] Unify Application.link_flags and Application.gcc_flags from mtouch and mmp into Application.CustomLinkFlags.
* [tests] Update mtouch tests according to mtouch changes.
* [ObjCRuntime] Remove deprecated availability attribute API from .NET.
They're quite useful for binding code though, so instead of removing them completely,
make them binding-only attributes (like numerous other binding attributes we have)
for .NET.
* [src] Remove removed attributes from the list of attributes that should be removed by the linker.
* [tests] Update tests to not use the old attributes for .NET.
There can't be any files in the root directory of the app bundle for macOS and
Mac Catalyst, otherwise code signing will fail. The problem is that Mono will
create a crash report in the current directory if the process crashes, and the
current directory is the root directory of the app bundle, which means that if
running an app crashes, the next build will likely fail because of the crash
report.
We had logic to detect this and remove any crash reports, but our crash report
detection pattern wasn't good enough and let some files through. This PR
updates that pattern, and also improves the code to report warnings for any
other files in the app bundle's root directory.
* Remove System.nint and System.nuint from .NET
* Add support for C#'s nint/nuint types to the generator.
* Accept IntPtr/UIntPtr as target types for BindAs attributes for NSNumber conversions.
* Fix a few APIs to take/return NativeHandle instead of IntPtr.
Fixes https://github.com/xamarin/xamarin-macios/issues/10508.
* [monotouch-test] Ignore a few tests in non-ARM64 simulators.
Some tests fail when running on an M1, but in a x64_86 mode, so just ignore
those unless we're running on ARM64 (this will currently exclude them on
x86_64 hardware too, but that'll eventually not be a problem anymore when
there's no more x86_64 hardware, and just checking for ARM64 is easier than
checking for x86_64 mode on an ARM64 CPU).
* Make more legacy projects unsafe.
Rolf Bjarne Kvinge
TJ Lambert
Manuel de la Pena
Přemek Vysoký