Status Update
Comments
pe...@gmail.com <pe...@gmail.com> #2
sg...@google.com <sg...@google.com>
cl...@google.com <cl...@google.com> #4
pu...@gmail.com <pu...@gmail.com> #5
sg...@google.com <sg...@google.com> #6
Android Studio version: 0.8.12
buildToolsVersion 21.0.1
Gradle 1.11
pu...@gmail.com <pu...@gmail.com> #7
--set-max-idx-number=<value>
Unfortunately changing the default is not a solution since the linearAlloc limit can be reached at very different levels depending on the classes hierarchy and other criteria.
In addition for most applications, moving to multidex will only help to workaround the linearalloc limit for the installation. But the application will still crash against the same limit at execution. The only working use case where I know multidex can help with linearalloc is when the apk does not contains one application but distinct pieces running in separate process.
pu...@gmail.com <pu...@gmail.com> #8
It's nice to know about that command line option. I do not see it in the output of 'dx --help', might be good to add that.
I'm not very familiar with the 'linearAlloc limit' issue outside of the context of the dexopt step. My sample app is able to run once the lower idx value is set, although I do not actually call into any of the library code that is bundled with the app. I assume it's undefined when/if the 'linearAlloc limit' will be hit in a large application on gb.
I'm a bit confused as to the platform compatibility of multidex given the 'linearAlloc limit' bug. What specific versions of Android are supported? The multidex code implies back to v4 (
to...@yahoo.com <to...@yahoo.com> #9
The linearalloc limit is reached when loading classes. At install time dexopt is loading all classes contained in the dex so it's facing the limit immediately. At execution the limit may be reached after some delay dependending of the usage you have of the packaged classes. If you face it at install time but not at execution, this means you never trigger the loading of some classes. In a real application those never loaded classes should have been shrinked away manually or by Proguard. The exception is when there are different groups of classes in the dex files used in separate process.
About multidex library supported versions I've merged recently a change to try to be clearer
The summary is that the library should work down to API 4 (Donut), but below ICS applications will probably be hit by the linearalloc limit
bi...@google.com <bi...@google.com> #10
dexOptions {
additionalParameters = ['--multi-dex', '--set-max-idx-number=40000']
}
hu...@google.com <hu...@google.com> #11
hu...@google.com <hu...@google.com> #14
Re
to...@yahoo.com <to...@yahoo.com> #15
I tried last week... but I can't reproduce it in a brand new project (app + java-library). I can only reproduce it in my full project. My best guess is that the 'app' part uses more android classes and thereby triggers the faulty behaviour in the desugar code.
Not sure how to proceed - but I'm quite happy to try out any custom/beta build you might have?
Depending on time, I'll have another go later this week.
hu...@google.com <hu...@google.com> #16
It would be great to track down the root cause first before we come up with a fix.
Could you generate an
./gradlew :app:mergeLibDexDebug -Dcom.android.tools.r8.dumpinputtodirectory=.
make sure the build fails with the above error, and send us dumpNNNN.zip
(s) (possibly in a restricted comment)?
pu...@gmail.com <pu...@gmail.com> #17
I also have a project producing this error, but noticed it only happens in debug. In release, it builds fine. I can attach a r8dump.zip for it if necessary. Although in the zip produced, library.jar and program.jar are corrupted.
hu...@google.com <hu...@google.com> #18
Please run the build with -Dcom.android.tools.r8.dumpinputtodirectory=.
instead of -Dcom.android.tools.r8.dumpinputtofile
, hopefully it will not be corrupted.
Make sure the build fails with The given file ... is located outside the root directory
, and send us the dumpNNNN.zip
files. Thanks a lot!
to...@yahoo.com <to...@yahoo.com> #19 Restricted
to...@yahoo.com <to...@yahoo.com> #20
I can confirm #18 (thanks!, didn't think of trying that)
The failure only happens in debug builds (as uploaded in previous) while it builds fine and deploys without an issue in release builds.
hu...@google.com <hu...@google.com> #21
That's great, thanks! (Release builds use a different pipeline, so it doesn't hit this code path.)
@Clément: Could you take a look at the dump above and see why in the desugaring graph (see com.android.tools.r8.D8Command.Builder#setDesugarGraphConsumer)
D8 reports an edge to/from android.jar
and whether that is expected?
If the behavior is correct in D8, I will fix this issue in AGP.
hu...@google.com <hu...@google.com> #22
Update: By patching AGP with the input dump above, I can see that D8 reports the following dependency to AGP (via com.android.tools.r8.DesugarGraphConsumer#accept
):
dependent = {ArchiveEntryOrigin@25758} "dump12086563528200/program.jar:com/.../.../channels/io/MyPipedInputStream.class"
dependency = {ArchiveEntryOrigin@25759} "Sdk/platforms/android-35/android.jar:java/io/InputStream.class"
Do we know why D8 reports that dependency given minSdk >= 24
(26 in the input dump)?
(It would be great to track this down before beta branching that's happening today.)
hu...@google.com <hu...@google.com> #23
Steps to reproduce
Based on the input dump above, I've managed to reproduce this issue in a project (see attached):
- Create a project with
app
andlib
(Android library), using AGP 8.5.2 to 8.7.1, minSdk = 24+ - In
app
, setisCoreLibraryDesugaringEnabled = true
andcoreLibraryDesugaring("com.android.tools:desugar_jdk_libs_nio:2.1.0")
- In
lib
, create a class that implementsjava.io.InputStream
orjava.io.PipedInputStream
:
class ExampleClass(pipeSize: Int) : java.io.PipedInputStream(pipeSize)
- Run
./gradlew :app:mergeLibDexDebug
, the build will fail with:
> Transform bundleLibRuntimeToDirDebug (project :lib) with DexingNoClasspathTransform
ERROR: D8: java.lang.IllegalStateException: The given file 'Sdk/platforms/android-34/android.jar' is located outside the root directory 'MyApplication/lib/build/intermediates/runtime_library_classes_dir/debug/bundleLibRuntimeToDirDebug'
Root cause
The issue is a combination of 2 factors:
- AGP doesn't expect desugaring dependencies between project's classes and
android.jar
- D8 in some cases reports dependencies between the
ExampleClass
above andjava.io.InputStream(RetargetInterface)
inandroid.jar
Workarounds
- Use
coreLibraryDesugaring("com.android.tools:desugar_jdk_libs_nio:2.0.4")
instead of2.1.0+
Fix
- AGP: Ignore dependencies between project's classes and
android.jar
- D8: Investigate root cause #2 and see why the behavior is inconsistent between
desugar_jdk_libs_nio:2.0.4
and2.1.0+
hu...@google.com <hu...@google.com> #24
- AGP: Ignore dependencies between project's classes and android.jar
We've done this in
Thanks a lot for this report and for helping us debug this issue!
Follow up
I'm closing this issue now, but it would be great if R8 team could verify the following, as it may indicate some bug in D8.
- D8: Investigate root cause #2 and see why the behavior is inconsistent between desugar_jdk_libs_nio:2.0.4 and 2.1.0+
ch...@google.com <ch...@google.com> #25
Bugjuggler: wait until 20241120 -> clementbera
an...@google.com <an...@google.com> #26
Thank you for your patience while our engineering team worked to resolve this issue. A fix for this issue is now available in:
- Android Studio Ladybug Feature Drop | 2024.2.2 Canary 9
- Android Gradle Plugin 8.8.0-alpha09
We encourage you to try the latest update.
If you notice further issues or have questions, please file a new bug report.
Thank you for taking the time to submit feedback — we really appreciate it!
ch...@google.com <ch...@google.com> #27
I opened
pu...@gmail.com <pu...@gmail.com> #28
Confirming it is fixed in Canary 9.
However, Canary 9 introduced a new unrelated issue for me:
to...@yahoo.com <to...@yahoo.com> #29
I don't get the error described, but any source code changes are not reflected when I hit "Debug". I have to manually build first.
I added myself to the newest issue...
cl...@google.com <cl...@google.com> #30
Could you confirm whether it is possible to have a (desugaring) dependency between the project's classes and android.jar when minSdk >= 24?
That is possible when using desugared library. Most desugaring happen below 24, but some methods are desugared up to latest android.
I see java.io.InputStream#transferTo(java.io.OutputStream)
is desugared with desugared library up to android 32 so that makes sense.
to...@yahoo.com <to...@yahoo.com> #31
meerkat / agp 8.9.x and suga 2.1.3 .... seems to build again ok.
thanks!
Description
Build failure after upgrading the desugar library to 2.1.0 It works in AGP 8.4.x.