Status Update
Comments
mk...@google.com <mk...@google.com>
je...@google.com <je...@google.com> #2
Lint options:
lintOptions {
lintConfig = file('lint.xml')
abortOnError true
xmlReport true
htmlReport true
checkDependencies true
}
Also, our project has custom lint rules. Failure happens with a dependency of 26.6.3 and 27.0.0.
Is any way to downgrade lint runner? This problem is keeping us from using AGP 4.0
al...@google.com <al...@google.com>
[Deleted User] <[Deleted User]> #3
Attached slightly different stacktrace for the same problem.
Wonder if disabling some lint rules could help to "fix" this problem, as it is also blocking for us 4.0 AGP upgrade.
al...@google.com <al...@google.com> #4
Notes:
I suspect the exceptions are happening now because JetBrains recently changed an assert
statement into a hard failure (see
I.e., it's possible that Lint has been hitting this code path all along, and only now it is causing a problem.
This has been difficult to investigate so far because the failure is nondeterministic and also deep inside the Kotlin compiler. Plus, JetBrains seems to be hitting this code path occasionally too (hence why they changed the assert into a hard failure), so I'm not even sure whether Lint is implicated or not.
al...@google.com <al...@google.com> #5
It could also explain why not all lint rules were working from time to time (
Is any public issue created I could follow?
mi...@gmail.com <mi...@gmail.com> #6
Has anyone tried this with 4.1.0-beta2? We're using a more recent version of UAST and the Kotlin compiler there. It's not clear that it's fixed but -- there's a chance :-)
cm...@google.com <cm...@google.com> #7
I tried. The same error. I am wondering, is any way to know which lint checks cause failure? I have a few custom lint checks. Or, is any way downgrade compiler for lint checks?
cm...@google.com <cm...@google.com> #8
I did more tests and probably narrow down a cause. I did all tests on 4.1.0-beta2
. Original configuration have enabled checkDependencies
and included custom lint checks. Custom lint checks are compiled using 26.6.3
version:
checkDependencies = true, customLintChecks = true
= 20% chance of successcheckDependencies = true, customLintChecks = false
= 57% chance of successcheckDependencies = false, customLintChecks = true
= 100% chance of success
It is a clear evidence that checkDependencies
feature is not stable. In addition, custom lint checks is not friendly with checkDependencies
and increases the risk of failure. Or maybe just an old lint SDK used is to blame.
It is very similar to
I have a theory, but do not know how to test. Maybe crashes happening because there are some lint checks from libs built upon old lint SDK version?
fr...@gmail.com <fr...@gmail.com> #9
To me this is happening on 4.2.0-alpha04 too.
Can someone confirm that this is happening if you have these three conditions met?:
- checkDependencies=true on your app module
- have a custom lint rule using lintChecks on your app module
- have a custom lint rule using lintChecks on your library module
For me the error went away after I removed the lintChecks on the library module.
Description
The new Variant API SourceDirectoriesImpl.addGeneratedSourceDirectory() allows to add a custom task generating source files that need to be compiled within the build process, but does not add the generated source directory to the IDE model in TaskProviderBasedDirectoryEntryImpl .
It is used in replacement of the old API BaseVariantImpl.registerJavaGeneratingTask() that automatically added the generated source directories to the IDE model.
It seems to me that the old behavior should be maintained, i.e., either:
shouldBeAddedToIdeModel
totrue
by default in the constructor ofTaskProviderBasedDirectoryEntryImpl
,shouldBeAddedToIdeModel = true
when creatingTaskProviderBasedDirectoryEntryImpl
inaddGeneratedSourceDirectory
,addGeneratedSourceDirectory
for letting the user set the desired value ofshouldBeAddedToIdeModel
.Note that it should go the same way with SourceDirectoriesImpl.addStaticSourceDirectory() .
Also, I would personally not set the path of the output directory here if it has already been set by the user before calling
addGeneratedSourceDirectory
, since this overwriting is not documented and not obvious to the user. Or similarly toshouldBeAddedToIdeModel
, this behavior (setting or not the output directory) could be controlled through an optional argument toaddGeneratedSourceDirectory
.Please consider this other issue too that might be relevant as well (I did not try to reproduce it), it apppears not only in TaskProviderBasedDirectoryEntryImpl but also in ProviderBasedDirectoryEntryImpl .
Studio Build: #AI-213.7172.25.2211.8571212 (Android Studio Electric Eel | 2022.1.1 Canary 2)
Version of Gradle Plugin: 7.4.0-alpha02
Version of Gradle: 7.5-rc-1
Version of Java: 1.8
OS: Ubuntu 18.04