Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Encountering crash in 'Hello' test case #1296

Closed
pjcollins opened this issue Feb 12, 2018 · 4 comments · Fixed by #1301
Closed

Encountering crash in 'Hello' test case #1296

pjcollins opened this issue Feb 12, 2018 · 4 comments · Fixed by #1301
Labels
Area: App Runtime Issues in `libmonodroid.so`.
Milestone

Comments

@pjcollins
Copy link
Member

Steps to Reproduce

  1. Install and run the provided .apk file on device.

Mono.Samples.HelloTests-Signed.zip

Expected Behavior

Application launches and the series of test cases which execute at startup will run to completion.

Actual Behavior

Application crashes on startup.

Version Information

Operating System: Microsoft Windows 10 Enterprise 64-bit
Product Name: Xamarin.Android
Branch Name: mono-2017-12
Build Revision: d386c85d46cf4d3b877d6610de14ee2fb84c1b69
[Device] Manufacturer: Google, Model: Pixel XL, Version: 7.1.1, ABI: arm64-v8a, Timezone: America/New_York

Log File

http://xqa.blob.core.windows.net/gist/log-b6d91c4d05284b508516c98e9868492f.txt

@pjcollins pjcollins added the Area: App Runtime Issues in `libmonodroid.so`. label Feb 12, 2018
@pjcollins pjcollins added this to the d15-7 milestone Feb 12, 2018
@jonpryor
Copy link
Member

Can you please provide the diagnostic build log? Also, in the obj/... tree, what are the contents of main.xml?

It looks like we’re not properly translating Mono.Samples.Hello.MyFragment into md5whatever.MyFragment, which is why Android can’t find it. Specifically, if I use apktool d to decode the contents of Mono.Samples.HelloTests-Signed.apk, res/layout-v11/main.xml contains:

<fragment android:name="Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null" android:id="@id/csharp_full_assembly" android:layout_width="fill_parent" android:layout_height="wrap_content" />

This is wrong; it should match one of the previous (in the same file!) lines, e.g.:

<fragment android:name="md5c7537e80324c78ebad79edc8b8a9cda5.MyFragment" android:id="@id/csharp_full_assembly" android:layout_width="fill_parent" android:layout_height="wrap_content" />

…but why wouldn’t that be happening? That seems quite odd. :-/

It's plausible that this is fallout from commit e5b1c92 -- last December! -- which wasn't not part of d15-6 (deliberately!). If this is the case, then ~any Xamarin.Android build after 2017-Jan-11 should likewise exhibit this bug, meaning it is not due to the mono/2017-12 bump.

@pjcollins
Copy link
Member Author

@jonpryor I was unable to reproduce this issue earlier today using monodroid/master/de602598, which is why I was curious as to the delta between master and mono-2017-12.

Diagnostic installation output can be found below:
http://xqa.blob.core.windows.net/gist/log-e7fb0ffa8a224d2ea324a1c06bed7196.txt

obj/Release/res/layout/main.xml contains:

<LinearLayout android:orientation="vertical" android:layout_width="fill_parent" android:layout_height="fill_parent" android:id="@+id/NameFixups" xmlns:android="http://schemas.android.com/apk/res/android"><ScrollView android:layout_width="fill_parent" android:layout_height="fill_parent" android:id="@+id/scroll"><LinearLayout android:id="@+id/extra" android:orientation="vertical" android:layout_width="fill_parent" android:layout_height="fill_parent"><ImageView android:id="@+id/image_view" android:layout_width="wrap_content" android:layout_height="wrap_content" android:src="@drawable/case_check" /><md5c7537e80324c78ebad79edc8b8a9cda5.MyView android:id="@+id/csharp_view" android:layout_width="match_parent" android:layout_height="wrap_content" android:text="csharp_view" /><md5c7537e80324c78ebad79edc8b8a9cda5.MyView android:id="@+id/legacy_acw_view" android:layout_width="match_parent" android:layout_height="wrap_content" android:text="legacy_acw_view" /></LinearLayout></ScrollView></LinearLayout>

obj/Release/res/layout-11/main.xml contains:

<LinearLayout android:orientation="vertical" android:layout_width="fill_parent" android:layout_height="fill_parent" android:id="@+id/NameFixups" xmlns:android="http://schemas.android.com/apk/res/android"><ScrollView android:layout_width="fill_parent" android:layout_height="fill_parent" android:id="@+id/scroll"><LinearLayout android:id="@+id/extra" android:orientation="vertical" android:layout_width="fill_parent" android:layout_height="fill_parent"><ImageView android:id="@+id/image_view" android:layout_width="wrap_content" android:layout_height="wrap_content" android:src="@drawable/case_check" /><md5c7537e80324c78ebad79edc8b8a9cda5.MyView android:id="@+id/csharp_view" android:layout_width="match_parent" android:layout_height="wrap_content" android:text="csharp_view" /><md5c7537e80324c78ebad79edc8b8a9cda5.MyView android:id="@+id/legacy_acw_view" android:layout_width="match_parent" android:layout_height="wrap_content" android:text="legacy_acw_view" /><fragment android:name="md5c7537e80324c78ebad79edc8b8a9cda5.MyFragment" android:id="@+id/csharp_simple_fragment" android:layout_width="match_parent" android:layout_height="wrap_content" /><fragment android:name="md5c7537e80324c78ebad79edc8b8a9cda5.MyFragment" android:id="@+id/csharp_legacy_fragment" android:layout_width="match_parent" android:layout_height="wrap_content" /><fragment android:name="md5c7537e80324c78ebad79edc8b8a9cda5.MyFragment" android:id="@+id/csharp_partial_assembly" android:layout_width="match_parent" android:layout_height="wrap_content" /><fragment android:name="Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null" android:id="@+id/csharp_full_assembly" android:layout_width="match_parent" android:layout_height="wrap_content" /></LinearLayout></ScrollView></LinearLayout>

@pjcollins
Copy link
Member Author

I think my inability to reproduce yesterday against master may have been due to an accidental usage of a different version which existed in a side by side installation. I am able to reproduce this against master as well.

@pjcollins pjcollins changed the title [Mono 2017-12] Encountering crash in 'Hello' test case Encountering crash in 'Hello' test case Feb 13, 2018
jonpryor added a commit to jonpryor/xamarin-android that referenced this issue Feb 14, 2018
Fixes: dotnet#1296

Xamarin.Android attempts to expose the Java-based Android API as a
".NET feeling" API. This takes many forms, such as prefixing interface
names with `I`, mapping `get` and `set` methods to properties, mapping
listener interfaces to events, and PascalCasing method names.

This also affects Java package names and C# namespaces.

When creating the `.apk` file, we philosophically need to go the
opposite direction: PascalCased members need to be mapped to
"something appropriate" within Java. For example, many Android
Resource ids *must* be all lowercase, and Android doesn't support
package names starting with an uppercase letter in all circumstances.

At one point, we tried mapping C# PascalCased namespaces to camelCased
namespaces, so e.g. `MyExampleNamespace` became the
`myExampleNamesapce` Java package within Java Callable Wrappers.

This turned out to be a Terrible Mistake, particularly on
case-sensitive filesystems, as if casing was *inconsistent*
(`MyExampleNamespace` vs `MyExamplenamespace`), files might not be
packaged.

By Mono for Android 1.0, we settled on just lowercasing the namespace
name to produce Java package names within Android Callable Wrappers.

This was not without it's own problems; in particular, the assembly
name wasn't involved, so if the "same" type (namespace + type) were
present in two different assemblies and we needed to generate Android
Callable Wrappers, they'd "collide," the build would fail, and we'd
have some unhappy customers.

This was later addressed in Xamarin.Android 5.1 by changing the Java
package name generation algorithm to be an MD5SUM of the assembly name
and namespace name, thus allowing types to have assembly identity.
(This was not without its own problems.)

Then it gets slightly more complicated: Android allows type names to
appear in various locations, such as in layout View XML. These don't
allow "just" using the type name; the package name is required for
types outside the `android.widget` Java package.

Initially, we did nothing, so developers had to directly use the
Android Callable Wrapper names:

	<myexamplenamespace.MyCustomCSharpView android:id="@+id/yay_csharp" .../>
	<fragment android:name="myexamplenamespace.MyCustomCSharpFragment" ... />

With the change in Xamarin.Android 5.1, *this couldn't work*; those
types didn't exist anymore.

To square this circle, we processed the resource files to "fixup"
identifiers and replace them with the actual Android Callable Wrapper
names. We'd replace any/all of:

	MyExampleNamespace.MyCustomCSharpView                           // Full name
	MyExampleNamespace.MyCustomCSharpView, MyAssembly               // Partial assembly-qualified name
	MyExampleNamespace.MyCustomCSharpView, MyAssembly, Version=...  // Full assembly qualified name
	myexamplenamespace.MyCustomCSharpView                           // compatibility name

with the appropriate md5'd Android Callable Wrapper name.

Brilliant as this was, there was a problem: [Bug #61073][61073].
If the assembly had a wildcard in the assembly version:

[61073]: https://bugzilla.xamarin.com/show_bug.cgi?id=61073

	[assembly: AssemblyVersion ("1.0.0.*")]

then the "Full assembly qualified name" value would change on *every
build*, which had numerious unintended knock-on effects.

This was fixed in commit e5b1c92, which worked largely by just
killing the Full assembly qualified name version entirely.
Xamarin.Android doesn't support embedding two different versions of
the same assembly, so this was considered to be fine.

...except for one compatibility case: `<fragment/>`s can contain
~arbitrary strings, and we support replacing the entire Full assembly
qualified name within them:

	<fragment
	    android:name="MyExampleNamespace.MyCustomCSharpFragment, MyAssembly, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null"
			...
	/>

However, in a post e5b1c92 world, the above now fails to load on the
device, because it's *not* being appropriately fixed up!

	FATAL EXCEPTION: main
	Process: Mono.Samples.HelloTests, PID: 22977
	java.lang.RuntimeException: Unable to start activity ComponentInfo{Mono.Samples.HelloTests/mono.samples.HelloApp}: android.view.InflateException: Binary XML file line #1: Binary XML file line #1: Error inflating class fragment
		...
	Caused by: android.view.InflateException: Binary XML file line #1: Binary XML file line #1: Error inflating class fragment
	Caused by: android.view.InflateException: Binary XML file line #1: Error inflating class fragment
	Caused by: android.app.Fragment$InstantiationException: Unable to instantiate fragment Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null: make sure class name exists, is public, and has an empty constructor that is public
		...
	Caused by: java.lang.ClassNotFoundException: Didn't find class "Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null" on path: DexPathList[[zip file "/data/app/Mono.Samples.HelloTests-1/base.apk"],nativeLibraryDirectories=[/data/app/Mono.Samples.HelloTests-1/lib/arm64, /system/fake-libs64, /data/app/Mono.Samples.HelloTests-1/base.apk!/lib/arm64-v8a, /system/lib64, /vendor/lib64]]
		...

The problem is that what Android "sees" *should* be

	<fragment
	    android:name="md5whatever.MyCustomCSharpFragment"
			...
	/>

where `md5whatever.MyCustomCSharpFragment` *is* a valid Java type that
Android is able to load successfully, but because of e5b1c92 this
replacement was removed.

The fix: simplify the Full assembly-qualified name case to an already
supported example. If the `//fragment/@android:name` value contains a
`,`, assume it's an assembly qualified name and compute the Full name
from it, by stripping off the comma and everything after it, then use
the Full name to lookup the correct Android Callable Wrapper type.
jonpryor added a commit to jonpryor/xamarin-android that referenced this issue Feb 14, 2018
While trying to repro Issue dotnet#1296, I had a brilliant idea: let's use
the new CodeBehind support from 7c31899!

Everything then fell apart.

For starters, Issue dotnet#1296 deals with Fragments, and 7c31899 didn't
support the presence of `<fragment/>` within Layout files.

Plumb support for that, by adding:

	partial class MainActivity {
	  partial void OnLayoutFragmentNotFound<T> (int resourceId, ref T type)
	    where T : global::Android.App.Fragment;
	}

Another issue was noticed as well: all the line numbers in the `#line`
pragmas were always `1`. This was narrowed down to a bug in
`GetLineInfo()`.

Then there's an issue with where I was using the CodeBehind file:
within the `src/Mono.Android/Test` project, which is in the
`Xamarin.Android.RuntimeTests` namespace. This prompted all manner of
namespace resolution failures:

	error CS0234: The type or namespace name 'App' does not exist in the namespace 'Xamarin.Android' (are you missing an assembly reference?)
	...

Fix this by using `CodeTypeReferenceOptions.GlobalReference` as much
as is practical (which isn't enough), and by "kludging" up the
`CodeNamespaceImport` construction so that we instead emit:

	using global::System;

Finally, the generated codebehind had some weird nesting going on:

	partial class MainActivity {
	  public __first_text_view_Views first_text_view {get;}
	  public sealed partial class __first_text_view_Views {
	    public __second_text_view_Views second_text_view {get;}
	    public sealed partial class __second_text_view_Views {
	      public sealed partial class __csharp_simple_fragment_Views {
	        public sealed partial class __csharp_partial_assembly_Views {
	        }
	      }
	    }
	  }
	}

It was bizarre *and* unusable: the `second_text_view` property --
corresponding to `<TextView android:id="@+id/second_text_view" />` --
was *nested within* a generated `MainActivity.__first_text_view_Views`
type, and I'm not sure why that existed. Furthermore, it *can't* work,
as the generated
`MainActivity.__first_text_view_Views.__CreateClass___second_text_view_Views()`
method depends on a constructor which doesn't exist:

	private @__second_text_view_Views @__CreateClass___second_text_view_Views() {
	  // There *is* a __second_text_view_Views(Activity) constructor,
	  // but not __second_text_view_Views(__first_text_view_Views).
	  return new @__second_text_view_Views(this);
	}

I "solved" (?) this by updating `LoadWidgets()` so that it always used
`widgetRoot` for all recursive calls to `LoadWidgets()`, so that there
is only ever one root. This removed all the nested types.
dellis1972 pushed a commit that referenced this issue Feb 14, 2018
Fixes: #1296

Xamarin.Android attempts to expose the Java-based Android API as a
".NET feeling" API. This takes many forms, such as prefixing interface
names with `I`, mapping `get` and `set` methods to properties, mapping
listener interfaces to events, and PascalCasing method names.

This also affects Java package names and C# namespaces.

When creating the `.apk` file, we philosophically need to go the
opposite direction: PascalCased members need to be mapped to
"something appropriate" within Java. For example, many Android
Resource ids *must* be all lowercase, and Android doesn't support
package names starting with an uppercase letter in all circumstances.

At one point, we tried mapping C# PascalCased namespaces to camelCased
namespaces, so e.g. `MyExampleNamespace` became the
`myExampleNamesapce` Java package within Java Callable Wrappers.

This turned out to be a Terrible Mistake, particularly on
case-sensitive filesystems, as if casing was *inconsistent*
(`MyExampleNamespace` vs `MyExamplenamespace`), files might not be
packaged.

By Mono for Android 1.0, we settled on just lowercasing the namespace
name to produce Java package names within Android Callable Wrappers.

This was not without it's own problems; in particular, the assembly
name wasn't involved, so if the "same" type (namespace + type) were
present in two different assemblies and we needed to generate Android
Callable Wrappers, they'd "collide," the build would fail, and we'd
have some unhappy customers.

This was later addressed in Xamarin.Android 5.1 by changing the Java
package name generation algorithm to be an MD5SUM of the assembly name
and namespace name, thus allowing types to have assembly identity.
(This was not without its own problems.)

Then it gets slightly more complicated: Android allows type names to
appear in various locations, such as in layout View XML. These don't
allow "just" using the type name; the package name is required for
types outside the `android.widget` Java package.

Initially, we did nothing, so developers had to directly use the
Android Callable Wrapper names:

	<myexamplenamespace.MyCustomCSharpView android:id="@+id/yay_csharp" .../>
	<fragment android:name="myexamplenamespace.MyCustomCSharpFragment" ... />

With the change in Xamarin.Android 5.1, *this couldn't work*; those
types didn't exist anymore.

To square this circle, we processed the resource files to "fixup"
identifiers and replace them with the actual Android Callable Wrapper
names. We'd replace any/all of:

	MyExampleNamespace.MyCustomCSharpView                           // Full name
	MyExampleNamespace.MyCustomCSharpView, MyAssembly               // Partial assembly-qualified name
	MyExampleNamespace.MyCustomCSharpView, MyAssembly, Version=...  // Full assembly qualified name
	myexamplenamespace.MyCustomCSharpView                           // compatibility name

with the appropriate md5'd Android Callable Wrapper name.

Brilliant as this was, there was a problem: [Bug #61073][61073].
If the assembly had a wildcard in the assembly version:

[61073]: https://bugzilla.xamarin.com/show_bug.cgi?id=61073

	[assembly: AssemblyVersion ("1.0.0.*")]

then the "Full assembly qualified name" value would change on *every
build*, which had numerious unintended knock-on effects.

This was fixed in commit e5b1c92, which worked largely by just
killing the Full assembly qualified name version entirely.
Xamarin.Android doesn't support embedding two different versions of
the same assembly, so this was considered to be fine.

...except for one compatibility case: `<fragment/>`s can contain
~arbitrary strings, and we support replacing the entire Full assembly
qualified name within them:

	<fragment
	    android:name="MyExampleNamespace.MyCustomCSharpFragment, MyAssembly, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null"
			...
	/>

However, in a post e5b1c92 world, the above now fails to load on the
device, because it's *not* being appropriately fixed up!

	FATAL EXCEPTION: main
	Process: Mono.Samples.HelloTests, PID: 22977
	java.lang.RuntimeException: Unable to start activity ComponentInfo{Mono.Samples.HelloTests/mono.samples.HelloApp}: android.view.InflateException: Binary XML file line #1: Binary XML file line #1: Error inflating class fragment
		...
	Caused by: android.view.InflateException: Binary XML file line #1: Binary XML file line #1: Error inflating class fragment
	Caused by: android.view.InflateException: Binary XML file line #1: Error inflating class fragment
	Caused by: android.app.Fragment$InstantiationException: Unable to instantiate fragment Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null: make sure class name exists, is public, and has an empty constructor that is public
		...
	Caused by: java.lang.ClassNotFoundException: Didn't find class "Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null" on path: DexPathList[[zip file "/data/app/Mono.Samples.HelloTests-1/base.apk"],nativeLibraryDirectories=[/data/app/Mono.Samples.HelloTests-1/lib/arm64, /system/fake-libs64, /data/app/Mono.Samples.HelloTests-1/base.apk!/lib/arm64-v8a, /system/lib64, /vendor/lib64]]
		...

The problem is that what Android "sees" *should* be

	<fragment
	    android:name="md5whatever.MyCustomCSharpFragment"
			...
	/>

where `md5whatever.MyCustomCSharpFragment` *is* a valid Java type that
Android is able to load successfully, but because of e5b1c92 this
replacement was removed.

The fix: simplify the Full assembly-qualified name case to an already
supported example. If the `//fragment/@android:name` value contains a
`,`, assume it's an assembly qualified name and compute the Full name
from it, by stripping off the comma and everything after it, then use
the Full name to lookup the correct Android Callable Wrapper type.
jonpryor added a commit to jonpryor/xamarin-android that referenced this issue Feb 14, 2018
While trying to repro Issue dotnet#1296, I had a brilliant idea: let's use
the new CodeBehind support from 7c31899!

Everything then fell apart.

For starters, Issue dotnet#1296 deals with Fragments, and 7c31899 didn't
support the presence of `<fragment/>` within Layout files.

Plumb support for that, by adding:

	partial class MainActivity {
	  partial void OnLayoutFragmentNotFound<T> (int resourceId, ref T type)
	    where T : global::Android.App.Fragment;
	}

Another issue was noticed as well: all the line numbers in the `#line`
pragmas were always `1`. This was narrowed down to a bug in
`GetLineInfo()`.

Then there's an issue with where I was using the CodeBehind file:
within the `src/Mono.Android/Test` project, which is in the
`Xamarin.Android.RuntimeTests` namespace. This prompted all manner of
namespace resolution failures:

	error CS0234: The type or namespace name 'App' does not exist in the namespace 'Xamarin.Android' (are you missing an assembly reference?)
	...

Fix this by using `CodeTypeReferenceOptions.GlobalReference` as much
as is practical (which isn't enough), and by "kludging" up the
`CodeNamespaceImport` construction so that we instead emit:

	using global::System;

Finally, the generated codebehind had some weird nesting going on:

	partial class MainActivity {
	  public __first_text_view_Views first_text_view {get;}
	  public sealed partial class __first_text_view_Views {
	    public __second_text_view_Views second_text_view {get;}
	    public sealed partial class __second_text_view_Views {
	      public sealed partial class __csharp_simple_fragment_Views {
	        public sealed partial class __csharp_partial_assembly_Views {
	        }
	      }
	    }
	  }
	}

It was bizarre *and* unusable: the `second_text_view` property --
corresponding to `<TextView android:id="@+id/second_text_view" />` --
was *nested within* a generated `MainActivity.__first_text_view_Views`
type, and I'm not sure why that existed. Furthermore, it *can't* work,
as the generated
`MainActivity.__first_text_view_Views.__CreateClass___second_text_view_Views()`
method depends on a constructor which doesn't exist:

	private @__second_text_view_Views @__CreateClass___second_text_view_Views() {
	  // There *is* a __second_text_view_Views(Activity) constructor,
	  // but not __second_text_view_Views(__first_text_view_Views).
	  return new @__second_text_view_Views(this);
	}

I "solved" (?) this by updating `LoadWidgets()` so that it always used
`widgetRoot` for all recursive calls to `LoadWidgets()`, so that there
is only ever one root. This removed all the nested types.
baulig pushed a commit that referenced this issue Feb 21, 2018
commit d3ad544
Author: Dean Ellis <dellis1972@googlemail.com>
Date:   Wed Feb 21 16:36:10 2018 +0000

    [Xamarin.Android.Build.Tasks] Update `CalculateProjectDependencies` to match spec (#1323)

    The `GetAndroidDependencies` target (f9b2c97) is intended to be
    called by an IDE on project-load, so that the IDE can install project
    dependencies "in the background," and ideally have the dependencies
    installed before the project is actually built.

    As such, the semantics of the `@(AndroidDependency)` output item
    group need to be agreed upon by xamarin-android, the IDEs, and
    whatever the IDEs use to actually install the dependencies. The
    `%(AndroidDependency.Identity)` values need to be well-defined
    and consistent across these three different libraries.

    In commit f9b2c97, certain values would have a version number
    "embedded" within `%(AndroidDependency.Identity)`, using a `;` to
    separate the name from the version, e.g. `build-tools;26.0.1`.
    Use of `;` was chosen because that made it easier for the package
    installation code to use (e.g. it would be identical to what the
    Android SDK `sdkmanager` utility expects).

    Unfortunately, the use of `;` was counter to the [spec][spec], and
    also looks "weird" in the context of MSBuild, as `;` is *also* the
    item group separator.

    Change the version separator to instead be a `/`. This "looks" better
    and is also consistent with the [installer manifest][manifest] schema
    that we will use, wherein `%(AndroidDependency.Identity)` matches a
    `//xamarin-android/*/@filesystem-path` attribute value:

    	<xamarin-android sdk-version="0.0.0" generated-on="Tue, 20 Feb 2018 21:16:01 GMT">
    	  <platform-tools filesystem-path="platform-tools" path="platform-tools" revision="27.0.1" manifest-url="https://dl.google.com/android/repository/repository2-1.xml" description="Android SDK Platform-Tools" obsolete="False" preview="False" license="android-sdk-license" original-type="generic:genericDetailsType">
    	    <!-- ... -->
    	  </platform-tools>
    	  <build-tool filesystem-path="build-tools/27.0.3" path="build-tools;27.0.3" revision="27.0.3" manifest-url="https://dl.google.com/android/repository/repository2-1.xml" description="Android SDK Build-Tools 27.0.3" obsolete="False" preview="False" license="android-sdk-license" original-type="generic:genericDetailsType">
    	  </build-tool>
    	  <!-- ... -->
    	</xamarin-android>

    `@(AndroidDependency)` could be equivalent to:

    	<ItemGroup>
    	  <AndroidDependency Include="build-tools/27.0.3">
    	    <Version>27.0.3</Version>
    	  </AndroidDependency>
    	  <AndroidDependency Include="platform-tools">
    	    <Version>$(AndroidSdkPlatformToolsVersion)</Version>
    	  </AndroidDependency>
    	  <AndroidDependency Include="platforms/android-27">
    	    <Version>27</Version>
    	  </AndroidDependency>
    	  <AndroidDependency Include="tools">
    	    <Version>$(AndroidSdkToolsVersion)</Version>
    	  </AndroidDependency>
    	</ItemGroup>

    [manifest]: https://gist.github.com/dellis1972/08ba76cc19cdce3dec89c68684664299
    [spec]: https://microsoft-my.sharepoint.com/:w:/r/personal/mhutch_microsoft_com/_layouts/15/WopiFrame.aspx?sourcedoc=%7B0436dd38-c9ff-4cf2-b33c-ee4515b68546%7D&action=edit&wdPid=64869a58

commit e489aae
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Wed Feb 21 10:41:32 2018 -0500

    Bump to mono/2017-12/de4729fa (#1325)

    Context: #1078
    Context: mono/mono#7145

    The primary impetus is that mono/de4729fa contains "ignore"
    improvements to `mono-api-html`, which will be needed in order to
    complete PR #1078.

commit f7b24f7
Author: Dean Ellis <dellis1972@googlemail.com>
Date:   Tue Feb 20 16:25:23 2018 +0000

    [Xamarin.Android.Build.Tests] Add tests for <ResolveSdks/> (#1321)

    This commit adds a basic unit test for the `<ResolveSdks/>` task.

commit d207275
Author: Atsushi Eno <atsushieno@gmail.com>
Date:   Tue Feb 20 00:37:10 2018 +0900

    [msbuild] Fix <Import>-ed filename, for case sensitivity. (#1322)

    There is a trivial difference between `msbuild` and `xbuild` - `msbuild`
    seems to respect (or ignorant of) the fact that filenames that are different
    in case are actually different on case-sensitive filesystem.

    Thus with `msbuild`, our binding projects don't build due to "missing"
    Microsoft.CSharp.Targets (there are only Microsoft.CSharp.targets even in
    `xbuild` support files in mono).

    This trivial fix should make it work.

commit 49ecdda
Author: Ludovic Henry <luhenry@microsoft.com>
Date:   Fri Feb 16 17:09:43 2018 -0500

    Bump to mono/2017-12/0d51a806 (#1318)

    To simplify integrating mono's SDKs feature.

commit 8419f81
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Fri Feb 16 16:07:12 2018 -0500

    Bump $(ProductVersion) to 8.3.99

    (Commercial) Xamarin.Android v8.3 is being tracked in the
    [xamarin-android/d15-7][0] branch, which was branched from
    commit 70ccc83.

    Bump `$(ProductVersion)` to 8.3.99 to track development progress of
    the *next* version of Xamarin.Android.

    [0]: https://github.com/xamarin/xamarin-android/commits/d15-7

commit 70ccc83
Author: Matt Sylvia <msylvia@nukefile.net>
Date:   Fri Feb 16 11:56:35 2018 -0500

    Bump to Java.Interop/master/15cf8c1, mono/2017-12/fbc4f05 (#1317)

    Context: mono/mono#6948

    Harmonizes on cecil/mono-2017-12/dfee11e.

commit c960a6f
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Fri Feb 16 10:43:11 2018 -0500

    Bump to xamarin-android-api-compatibility/d15-7/dc5f0c2e

    Fixes an [ABI break][0] in 72ef39b:

    [0]: https://jenkins.mono-project.com/view/Xamarin.Android/job/xamarin-android/877/

    	<h3>Type Changed: Microsoft.SqlServer.Server.SqlDataRecord</h3>
    	<p>Removed method:</p>
    	<pre>
    	  <span class='removed removed-method breaking' data-is-breaking>public virtual System.Data.IDataReader GetData (int);</span>
    	</pre>

    	<h3>Type Changed: System.Data.SqlClient.SqlConnectionStringBuilder</h3>
    	<p>Removed property:</p>
    	<pre>
    	  <span class='removed removed-property breaking' data-is-breaking>public bool IsFixedSize { get; }</span>
    	</pre>

    	</div> <!-- end type SqlConnectionStringBuilder -->
    	<!-- start type SqlParameterCollection --> <div>
    	<h3>Type Changed: System.Data.SqlClient.SqlParameterCollection</h3>
    	<p>Removed properties:</p>
    	<pre>
    	  <span class='removed removed-property breaking' data-is-breaking>public bool IsFixedSize { get; }</span>
    	  <span class='removed removed-property breaking' data-is-breaking>public bool IsReadOnly { get; }</span>
    	  <span class='removed removed-property breaking' data-is-breaking>public bool IsSynchronized { get; }</span>
    	</pre>

    This doesn't fix the "ABI break"; there is no ABI break of consequence.
    This fixes the *reporting* of the ABI break.

    See also: xamarin/xamarin-android-api-compatibility@5be4d43

commit 72ef39b
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Thu Feb 15 20:32:20 2018 -0500

    Bump to xamarin-android-api-compatibility/d15-7/9d71e200

    Context: #1303

    Fixes an [ABI BREAK][0] from b2ca17a:

    	<h3>Type Changed: Android.Widget.ListView</h3>
    	<p>Modified fields:</p>
    	<pre>
    	<div data-is-breaking>	public const <span class='removed removed-inline removed-breaking-inline'>int</span> <span class='added '>ChoiceMode</span> ChoiceModeMultiple = 2;
    	</div><div data-is-breaking>	public const <span class='removed removed-inline removed-breaking-inline'>int</span> <span class='added '>ChoiceMode</span> ChoiceModeNone = 0;
    	</div><div data-is-breaking>	public const <span class='removed removed-inline removed-breaking-inline'>int</span> <span class='added '>ChoiceMode</span> ChoiceModeSingle = 1;
    	</div></pre>

    The problem is that [xamarin-android-api-compatibility PR #13][ac-13]
    is part of PR #1303, but #1303 hasn't been merged yet. As such, it had
    no need to be in the xamarin-android-api-compatibility/d15-7 branch.

    It has been reverted, thus fixing the above ABI break.

    [0]: https://jenkins.mono-project.com/view/Xamarin.Android/job/xamarin-android/876/console
    [ac-13]: xamarin/xamarin-android-api-compatibility#13

commit d0cb3ce
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Thu Feb 15 14:12:30 2018 -0500

    Bump to mono/2017-12/2ad3f0bf

    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60088

commit b2ca17a
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Thu Feb 15 12:38:12 2018 -0500

    Bump to xamarin-android-api-compatibility/d15-7/10df9b67 (#1314)

    Context: 34f437d
    Context: https://jenkins.mono-project.com/view/Xamarin.Android/job/xamarin-android/873/

    The bump to mono/2017-12 in commit 34f437d included ABI "breakage" in
    `System.Data.dll` which wasn't *actual* breakage.

    Update xamarin-android-api-compatibility to so that we don't report
    warnings for this "breakage."

commit 3a35c00
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Thu Feb 15 12:36:24 2018 -0500

    Bump to Java.Interop/master/0841c32f

    Bumps to cecil/master/dfee11e

    For d15-7 harmonization.

commit 92949f4
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Thu Feb 15 12:27:55 2018 -0500

    [mono-runtimes] Fix bundle usage

    Commit 34f437d inadvertently broke mono bundle usage. For example,
    the [first commit after the mono bump][xa-01be] *should* have taken
    around 3h to build. Instead, it took nearly 7h, because the bundle
    wasn't being used, causing the mono runtimes to be rebuilt:

    [xa-01be]: https://jenkins.mono-project.com/view/Xamarin.Android/job/xamarin-android/874/

    	Target _BuildUnlessCached needs to be built as output file
    	'…/xamarin-android//bin/Debug/lib/xamarin.android//../../bcl-tests/System.Runtime.CompilerServices.Unsafe.pdb'
    	does not exist.

    Indeed, `System.Runtime.CompilerServices.Usafe.pdb` *doesn't* exist,
    nor should it, as it's compiled from IL directly.

    Fix mono bundle usage by updating `@(_BclTestAssemblyDestination)` to
    exclude `System.Runtime.CompilerServices.Unsafe.pdb`. This allows the
    `_BuildUnlessCached` target to properly recognize that all required
    files *do* exist, thus skipping the mono-runtimes build (as desired).

commit 01be8ac
Author: Dean Ellis <dellis1972@googlemail.com>
Date:   Thu Feb 15 02:38:56 2018 +0000

    [Xamarin.Android.Build.Tasks] Stop designtime designer being deleted. (#1307)

    Fixes: #1286

    We have a number of problems with our DesignTime build system. The
    main one which this PR addresses is the designer file is deleted
    by `IncrementalClean` AND `CoreClean`. This them completely messes
    up the DesignTime system since it can no longer find the file.

    So what we should be doing is making sure we don't tell the build
    system about the designer file ;-). We do this by not writing the
    path to the `$(CleanFile)`.

    ~~ Some background on the [DesignTime build system][2] ~~

    The primary entry point is the `CoreCompile` target, with occasional
    calls to the `RefreshReferences` target. In theory
    [`$(DesignTimeBuild)`][1] should be set to `true`; however, this is
    not always the case which is why we have the `_SetupForDesignTimeBuild`
    target which hooks into the `CoreCompile` target. Additionally,
    `$(BuildingProject)` should be False. ([Additional information][2]).
    However this document is not seem to be 100% accurate since the IDE's
    (namely Visual Studio) do not always set the required properties. One
    of the key requirements for design time builds seems to be that they
    are independent of the main builds. The files it uses should not be
    removed since the `CoreCompile` target does not seem to be called
    consistently. For example, if the designer file is removed and the
    `RefreshReferences` target is called, the design time build may well
    fail because the targets required to re-generate the designer file
    are not called.

    Now for the problem: the `IncrementalClean` target runs between builds.
    It seems to want to remove files from the previous build. I suspect the
    purpose is to remove files that are no longer needed. (For example an
    assembly which is no longer referenced.) The problem is that we have
    a file (the design time `Resource.designer.cs`) which is not built as
    part of the normal build. But this file was still being written to
    `$(CleanFile)` which is the driver for the `IncrementalClean` process.
    As a result, the file gets removed because its not in the list of
    current `@(FileWrites)`. Not writing the name of this file to
    `$(CleanFile)` seems to hide it from `IncrementalClean`.

    There was also another problem with the `CoreClean` target. This was
    also removing the designer files. Again this was down to the fact
    that the file was listed in `$(CleanFile)`. However our older
    workarounds did not work since they relied on `IncrementalClean`
    running. For a `Clean` target invocation it does not. Again the
    solution seems to be Don't Write the file to `$(CleanFile)`, which is
    counterintuitive since we are told to write all files we generate to
    that file.

    [1]:  https://github.com/dotnet/project-system/blob/master/docs/design-time-builds.md#determining-whether-a-target-is-run-in-a-design-time-build
    [2]: https://github.com/dotnet/project-system/blob/master/docs/design-time-builds.md

commit 34f437d
Author: Ludovic Henry <luhenry@microsoft.com>
Date:   Wed Feb 14 14:05:02 2018 -0500

    Bump to mono/2017-12/a0af42ab (#1263)

    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=6339
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=6401
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=8477
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=19503
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=20562
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=31507
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=35661
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=40699
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=45893
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=49308
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=52675
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=53296
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=56003
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=56194
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=57893
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=57938
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=58261
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=58400
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=58411
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=58965
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59080
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59182
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59364
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59393
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59608
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59664
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59881
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59909
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59913
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59916
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59953
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59956
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=59967
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60028
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60029
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60115
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60175
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60216
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60224
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60233
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60238
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60245
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60255
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60267
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60288
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60298
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60317
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60340
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60422
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60435
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60505
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60514
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60539
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60545
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60634
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60680
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60771
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60860
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60865
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=60973
    Context: https://bugzilla.xamarin.com/show_bug.cgi?id=61002

    Bump minimum macOS Mono version to 5.10.0.47.

commit f279ad3
Author: Dean Ellis <dellis1972@googlemail.com>
Date:   Wed Feb 14 15:04:45 2018 +0000

    [Xamarin.Android.Build.Tasks] GetAndroidDependencies calls _SetLatestTargetFrameworkVersion (#1305)

    This is a bug in the `GetAndroidDependencies` target: it does not
    cause `MonoAndroidHelper.RefreshSupportedVersions()` to be invoked.
    As a result if called directly it will generate a
    `NullReferenceException` as `MonoAndroidHelper.SupportedVersions`
    will be `null`.

    The `GetAndroidDependencies` should thus depend upon the
    `_SetLatestTargetFrameworkVersion` target. This will initialize
    `MonoAndroidHelper.SupportedVersions`, and will also resolve the
    correct `$(TargetFrameworkVersion)` if the user has
    `$(AndroidUseLatestPlatformSdk)`=True.

commit 0dee27d
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Wed Feb 14 05:50:23 2018 -0500

    [Xamarin.Android.Build.Tasks] Fragments & Shorter acw-map (#1301)

    Fixes: #1296

    Xamarin.Android attempts to expose the Java-based Android API as a
    ".NET feeling" API. This takes many forms, such as prefixing interface
    names with `I`, mapping `get` and `set` methods to properties, mapping
    listener interfaces to events, and PascalCasing method names.

    This also affects Java package names and C# namespaces.

    When creating the `.apk` file, we philosophically need to go the
    opposite direction: PascalCased members need to be mapped to
    "something appropriate" within Java. For example, many Android
    Resource ids *must* be all lowercase, and Android doesn't support
    package names starting with an uppercase letter in all circumstances.

    At one point, we tried mapping C# PascalCased namespaces to camelCased
    namespaces, so e.g. `MyExampleNamespace` became the
    `myExampleNamesapce` Java package within Java Callable Wrappers.

    This turned out to be a Terrible Mistake, particularly on
    case-sensitive filesystems, as if casing was *inconsistent*
    (`MyExampleNamespace` vs `MyExamplenamespace`), files might not be
    packaged.

    By Mono for Android 1.0, we settled on just lowercasing the namespace
    name to produce Java package names within Android Callable Wrappers.

    This was not without it's own problems; in particular, the assembly
    name wasn't involved, so if the "same" type (namespace + type) were
    present in two different assemblies and we needed to generate Android
    Callable Wrappers, they'd "collide," the build would fail, and we'd
    have some unhappy customers.

    This was later addressed in Xamarin.Android 5.1 by changing the Java
    package name generation algorithm to be an MD5SUM of the assembly name
    and namespace name, thus allowing types to have assembly identity.
    (This was not without its own problems.)

    Then it gets slightly more complicated: Android allows type names to
    appear in various locations, such as in layout View XML. These don't
    allow "just" using the type name; the package name is required for
    types outside the `android.widget` Java package.

    Initially, we did nothing, so developers had to directly use the
    Android Callable Wrapper names:

    	<myexamplenamespace.MyCustomCSharpView android:id="@+id/yay_csharp" .../>
    	<fragment android:name="myexamplenamespace.MyCustomCSharpFragment" ... />

    With the change in Xamarin.Android 5.1, *this couldn't work*; those
    types didn't exist anymore.

    To square this circle, we processed the resource files to "fixup"
    identifiers and replace them with the actual Android Callable Wrapper
    names. We'd replace any/all of:

    	MyExampleNamespace.MyCustomCSharpView                           // Full name
    	MyExampleNamespace.MyCustomCSharpView, MyAssembly               // Partial assembly-qualified name
    	MyExampleNamespace.MyCustomCSharpView, MyAssembly, Version=...  // Full assembly qualified name
    	myexamplenamespace.MyCustomCSharpView                           // compatibility name

    with the appropriate md5'd Android Callable Wrapper name.

    Brilliant as this was, there was a problem: [Bug #61073][61073].
    If the assembly had a wildcard in the assembly version:

    [61073]: https://bugzilla.xamarin.com/show_bug.cgi?id=61073

    	[assembly: AssemblyVersion ("1.0.0.*")]

    then the "Full assembly qualified name" value would change on *every
    build*, which had numerious unintended knock-on effects.

    This was fixed in commit e5b1c92, which worked largely by just
    killing the Full assembly qualified name version entirely.
    Xamarin.Android doesn't support embedding two different versions of
    the same assembly, so this was considered to be fine.

    ...except for one compatibility case: `<fragment/>`s can contain
    ~arbitrary strings, and we support replacing the entire Full assembly
    qualified name within them:

    	<fragment
    	    android:name="MyExampleNamespace.MyCustomCSharpFragment, MyAssembly, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null"
    			...
    	/>

    However, in a post e5b1c92 world, the above now fails to load on the
    device, because it's *not* being appropriately fixed up!

    	FATAL EXCEPTION: main
    	Process: Mono.Samples.HelloTests, PID: 22977
    	java.lang.RuntimeException: Unable to start activity ComponentInfo{Mono.Samples.HelloTests/mono.samples.HelloApp}: android.view.InflateException: Binary XML file line #1: Binary XML file line #1: Error inflating class fragment
    		...
    	Caused by: android.view.InflateException: Binary XML file line #1: Binary XML file line #1: Error inflating class fragment
    	Caused by: android.view.InflateException: Binary XML file line #1: Error inflating class fragment
    	Caused by: android.app.Fragment$InstantiationException: Unable to instantiate fragment Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null: make sure class name exists, is public, and has an empty constructor that is public
    		...
    	Caused by: java.lang.ClassNotFoundException: Didn't find class "Mono.Samples.Hello.MyFragment, Hello, Version=1.0.0.0, Culture=neutral, PublicKeyToken=null" on path: DexPathList[[zip file "/data/app/Mono.Samples.HelloTests-1/base.apk"],nativeLibraryDirectories=[/data/app/Mono.Samples.HelloTests-1/lib/arm64, /system/fake-libs64, /data/app/Mono.Samples.HelloTests-1/base.apk!/lib/arm64-v8a, /system/lib64, /vendor/lib64]]
    		...

    The problem is that what Android "sees" *should* be

    	<fragment
    	    android:name="md5whatever.MyCustomCSharpFragment"
    			...
    	/>

    where `md5whatever.MyCustomCSharpFragment` *is* a valid Java type that
    Android is able to load successfully, but because of e5b1c92 this
    replacement was removed.

    The fix: simplify the Full assembly-qualified name case to an already
    supported example. If the `//fragment/@android:name` value contains a
    `,`, assume it's an assembly qualified name and compute the Full name
    from it, by stripping off the comma and everything after it, then use
    the Full name to lookup the correct Android Callable Wrapper type.

commit 5c46ee3
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Wed Feb 14 05:42:22 2018 -0500

    [api-xml-adjuster] Fix rebuilds (#1300)

    Ideally, project rebuilds when nothing has changed should be *fast*.

    `api-xml-adjuster` isn't:

    	$ time (cd build-tools/api-xml-adjuster ; xbuild)
    	real	2m1.084s
    	user	1m55.916s
    	sys	0m8.853s

    	# and the rebuild!
    	$ time (cd build-tools/api-xml-adjuster ; xbuild)
    	real	2m0.824s
    	user	1m56.140s
    	sys	0m8.600s

    A *minimum* two minute+ rebuild -- when *nothing* has changed -- is a
    surefire way to get really annoyed.

    With diagnostic logging, we start to see the culprit:

    	Target _ClassParse needs to be built as input file '@(ApiFileDefinition -> /Volumes/Seagate4TB/work/xamarin-android/build-tools/api-xml-adjuster/../../src/Mono.Android/Profiles/api-27.params.txt)' does not exist.

    This in turn causes `class-parse.exe` and `api-xml-adjuster.exe` to be
    *re-executed* on *every* `android.jar` on *every* build.

    Fix this by correcting the `//Target/@Inputs` and `//Target/@Outputs`
    for the `_ClassParse` and `_AdjustApiXml` tasks. After which,
    no-change rebuilds are *significantly* faster:

    	$ time (cd build-tools/api-xml-adjuster ; xbuild)
    	real	0m5.308s
    	user	0m6.042s
    	sys	0m1.237s

commit 20b4190
Author: Peter Collins <pecolli@microsoft.com>
Date:   Tue Feb 13 12:53:26 2018 -0500

    [BCL-Tests] Remove '-s' from _GrantPermissions (#1299)

    The [semantics of the `$(AdbTarget)` property][adb-target] are the
    same as the [`adb` Target Device option][adb], which permits any of:

    [adb-target]: Documentation/build_process.md
    [adb]: https://developer.android.com/studio/command-line/adb.html#issuingcommands

      * `-d`: Only attached *device*.
      * `-e`: Only attached *emulator*
      * `-s SERIAL_NUMBER`: A specifically named target; needed if there
        is more than one attached device or emulator.

    The problem with commit c4e8165 is that it overrode the
    `$(InstallDependsOnTargets)` property to call the `_GrantPermissions`
    target, which doesn't properly use `$(AdbTarget)`; it instead
    *requires* using the `-s` option, which is inconsistent:

    	adb -s $(AdbTarget) shell pm grant ...

    Remove `-s` from the `adb` invocation so that `$(AdbTarget)` can
    contain e.g. `-d` or `-e`, as is intended & documented.

commit a4ed574
Author: Dean Ellis <dellis1972@googlemail.com>
Date:   Tue Feb 13 16:04:44 2018 +0000

    [Xamarin.Android.Build.Tasks] Only report `ndk-bundle` if required. (#1298)

    We should only list `ndk-bundle` in the `@(AndroidDependencies)` if
    the user is using AOT or MkBundle. Otherwise we will end up
    downloading a ton of stuff we don't need.

    This commit fixes the `CalculateProjectDependencies` to report
    `ndk-bundle` if needed. It also adds a unit test to make sure
    it is not included when it shouldn't be.

commit cb68bc3
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Mon Feb 12 17:58:59 2018 -0500

    Bump to Java.Interop/master/10647a5e (#1295)

    Fixes the `ObjectDisposedException` during process shutdown described
    in commit 0a9d164.

    Moves the `ManagedPeer.Init()` invocation within the `JniRuntime`
    constructor so that a `NullReferenceException` is avoided. (Instead, a
    *different*, more meaningful exception should be thrown.)

    Bumps to cecil/mono-2017-12/1afa0668 for d15-7 harmonization.

commit f9b2c97
Author: Dean Ellis <dellis1972@googlemail.com>
Date:   Mon Feb 12 16:20:35 2018 +0000

    [Xamarin.Android.Build.Tasks] Add GetAndroidDependencies Target (#1290)

    Fixes: #1269

    This commit adds the `GetAndroidDependencies` target to the
    `Xamarin.Android.Common.targets`. Its purpose is to examine the
    various settings in the project and report which Android SDK
    build-tools, platform-tools, etc. are required.

    `GetAndroidDependencies` will output an `@(AndroidDependency)` with
    `%(AndroidDependency.Version)` metadata.

    `@(AndroidDependency)` will contain `%(Identity)` values of:

      * `build-tools;{VERSION}`, where `{VERSION}` and `%(Version)` is
         controlled by the `$(AndroidSdkBuildToolsVersion)` property.
      * `ndk-bundle`: `%(Version)` is controlled by the
        `$(AndroidNdkVersion)` property.
      * `platform-tools`: `%(Version)` is controlled by the
        `$(AndroidSdkPlatformToolsVersion)` property.
      * `platforms;android-{API}`, where `{API}` and `%(Version)` is the
        API level for the `$(TargetFrameworkVersion)` in the `.csproj`.
        For example, `platforms;android-27` for
        `$(TargetFrameworkVersion)`=`v7.1`.
      * `tools`: `%(Version)` is controlled by the
        `$(AndroidSdkToolsVersion)` property.

commit 0d822e2
Author: Jonathan Pryor <jonpryor@vt.edu>
Date:   Mon Feb 12 10:39:48 2018 -0500

    [Xamarin.Android.Build.Tasks] Improve `javac -version` parsing (#1292)

    The [Ubuntu build][0] is [failing][1]:

    	warning : Failed to get the Java SDK version as it does not appear to contain a valid version number.
    	`javac -version` returned:
    	```openjdk version "1.8.0_03-Ubuntu"
    	OpenJDK Runtime Environment (build 1.8.0_03-Ubuntu-8u77-b03-3ubuntu3-b03)
    	OpenJDK 64-Bit Server VM (build 25.03-b03, mixed mode)
    	```
    	...
    	Task "AdjustJavacVersionArguments"
    	  …/bin/Debug/lib/xamarin.android/xbuild/Xamarin/Android/Xamarin.Android.Common.targets:
    	  error : Error executing task AdjustJavacVersionArguments: Required property 'JdkVersion' not set.

    [0]: https://jenkins.mono-project.com/view/Xamarin.Android/job/xamarin-android-linux/779/
    [1]: https://jenkins.mono-project.com/view/Xamarin.Android/job/xamarin-android-linux/779/consoleText

    This was introduced/broken by commit 0e1d1c8, which turned
    `AdjustJavacVersionArguments.JdkVersion` into a `[Required]`
    parameter. The value for `AdjustJavacVersionArguments.JdkVersion`
    comes from the `<ResolveSdks/>` task `JdkVersion` output parameter,
    which itself comes from parsing `javac -version` output.

    The problem is that the regex we used to parse the output of
    `javac -version` wasn't properly parsing the output on Ubuntu:

    	openjdk version "1.8.0_03-Ubuntu"

    Specifically, the `-Ubuntu` text was causing the regex to not match.

    Update the regex to be a bit more lenient, enabling it to work with
    the Ubuntu `javac -version` output, which *should* allow the build to
    progress further.

commit 0114586
Author: Jonathan Peppers <jonathan.peppers@gmail.com>
Date:   Sun Feb 11 12:53:52 2018 -0600

    [build] fix api-xml-adjuster.targets on Windows (#1291)

    Since 7d705bf, the Windows builds on VSTS have seemed to be failing. I
    tested this locally, and noticed the use of command line arguments such
    as:

        -parameter-names='%(SomeVariable)'

    Unfortunately, this isn't working on Windows due to the single quote. It
    is more appropriately expressed as:

        -parameter-names=&quot;%(SomeVariable)&quot;

    This isn't very pretty, but it should work on all platforms.

    I also fixed all the tabs I saw in this file--in favor of spaces, and
    fixed other XML code conventions.

    I will now return to my regularly scheduled baby duty.
@pjcollins
Copy link
Member Author

Verified fixed in d15-7.

@ghost ghost locked as resolved and limited conversation to collaborators Jun 9, 2022
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
Area: App Runtime Issues in `libmonodroid.so`.
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants