We can't trust Apple's native linker to pick the right (non private)
framework when an older TargetVersion is used. It just prefer what's
available - even if specified with a WeakFramework :(
That was already dealt with for applications. However the native linking
of the Xamarin.Sdk.framework (code sharing with extensions) is done with
the `LinkTask` instead of the `NativeLinkTask` so it did not have the
"auto correct" code.
Unit test added.
reference:
https://bugzilla.xamarin.com/show_bug.cgi?id=61039
Apple rejects apps that links with the previously private IOSurface.framework.
We were not sure about this [1] and had it disabled (this is removed).
However another bug stopped our adjustment logic [2] from being executed. This
is needed since `clang` picks up the private framework if it does not find the
public one (even when asking for a weak framework),
[1] https://bugzilla.xamarin.com/show_bug.cgi?id=59201
[2] Added a while ago for JavascriptCore
* [mtouch] Improve how we make sure native symbols aren't stripped away. Fixes#51710 and #54417.
* Refactor required symbol collection to store more information about each
symbol (field, function, Objective-C class), and in general make the code
more straight forward.
* Implement support for generating source code that references these symbols,
and do this whenever we can't ask the native linker to keep these symbols
(when using bitcode). Additionally make it possible to do this manually, so
that the source code can be generated for non-bitcode platforms too (which
is useful if the number of symbols is enormous, in which case we might
surpass the maximum command-line length).
* Also make it possible to completely ignore native symbols, or ignore them on
a per-symbol basis. This provides a fallback for users if we get something
right and we try to preserve something that shouldn't be preserved (for
instance if it doesn't exist), and the user ends up with unfixable linker
errors.
* Don't collect Objective-C classes unless they're in an assembly with
LinkWith attributes. We don't need to preserve Objective-C classes in any
other circumstances.
* Implement everything for both Xamarin.iOS and Xamarin.Mac, and share the
code between them.
* Remove previous workaround for bug #51710, since it's no longer needed.
* Add tests.
https://bugzilla.xamarin.com/show_bug.cgi?id=54417https://bugzilla.xamarin.com/show_bug.cgi?id=51710
* [mtouch] Make sure to only keep symbols from the current app when code sharing.
This fixes a build problem with the interdependent-binding-projects test when
testing in Today Extension mode.
* [runtime] Fix Xamarin-debug.framework's install name.
This makes building to frameworks work in debug mode.
* [mtouch] Fix check to add frameworks to watchKit extensions.
* [mtouch] Never pass -read_only_relocs to the native linker when bitcode is enabled.
* [mtouch] Bitcode requires linking with c++.
This particular case applies to shared libraries/frameworks (we already link
with c++ when building statically).
Also stop using `mdb` as the name for debug symbols and remove
> static MdbReader mdb_reader;
since we're not mkbundl'ing mtouch anymore.
Related to https://github.com/xamarin/xamarin-macios/pull/2002 for mmp
The previous build system kept a forward-pointing single linked list of tasks
to execute: task X had a list of subsequent tasks to execute. If task X was
up-to-date, it was not created (and the next tasks were directly added to the
list of tasks to execute).
In this world it became complicated to merge output from tasks (for instance
if the output of task X and task Y should be a consumed by a single task
producing a single output, since the corresponding task would end up in both
X's and Y's list of subsequent tasks).
Example: creating a single framework from the aot-compiled output of multiple
assemblies.
So I've reversed the logic: now we keep track of the final output, and then
each task has a list of dependencies that must be built.
This makes it trivial to create merging tasks (for the previous example, there
could for instance be a CreateFrameworkTask, where its dependencies would be
all the corresponding AotTasks).
We also always create every task, and then each task decides when its executed
whether it should do anything or not. This makes it unnecessary to 'forward-
delete' files when creating tasks (say you have three tasks, A, B, C; B
depends on A, and C depends on B; if A's output isn't up-to-date, it has to
delete its own output if it exists, otherwise B would not detect that it would
have to re-execute, because at task *creation* time, B's input hadn't
changed).
Additionally make it based on async/await, since much of the work happens in
externel processes (and we don't need to spin up additional threads just to
run external processes). This makes us have less code run on background
threads, which makes any issues with thread-safety less likely.
The AOT-compilation occurs in the AOT-task now, and then we compile the result
using CompileTask.
This means that the error message in CompileTask was slightly incorrect, so
rectify it.
This makes dylibs automatically have the correct dylib id, which means no
fixups are required.
For instance: we'd build libpinvokes.armv7.dylib from libpinvokes.armv7.m,
which by default ends up with a dylib id of "libpinvokes.armv7.dylib". With
this fix no change is required, since we now build armv7/libpinvokes.dylib
from armv7/libpinvokes.m.
Use @rpath instead of @executable_path in dylibs, since it allows us to be
more flexible when placing dylibs in the app.
In particular with this change it's trivial to put libmonosgen-2.0.dylib in
the container app, and reference it from extensions.
Performance tests
-----------------
This is for a new watchOS extension project, built for release.
* The default (currently -O2) optimizations: 41s ( baseline ) 30.027.060 bytes ( baseline )
* All optimizations disabled (`--llvm-opt=all=`): 17s (-24s = -59%) 32.978.312 bytes (+2.951.252 = +10%)
* Optimized for size (`--llvm-opt=all=-Os`): 36s ( -5s = -12%) 28.617.408 bytes (-1.409.652 = -5%)
* Optimized for more size (`--llvm-opt=all=-Oz`): 35s ( -6s = -15%) 28.601.016 bytes (-1.426.044 = -5%)
* Optimized slightly (`--llvm-opt=all=-O1`): 35s ( -6s = -15%) 28.666.556 bytes (-1.360.504 = -5%)
* Optimized a lot (`--llvm-opt=all=-O3`): 41s ( 0s = 0%) 30.403.996 bytes (+ 376.936 = +1%)
Conclusions
-----------
* The fastest build by far (less than twice as fast) is if optimizations are
disabled, but this adds a 10% size penalty (~3 MB in this test case),
compared to the baseline, and 15% size penalty (4.3 MB) compared to -Oz.
* -Oz seems to have the best overall results: at least as fast as any other
optimized build, and the smallest app as well.
Caveats
-------
Some optimizations might not work the AOT compiled code. The resulting
binaries have not been tested.