Lint intermediate files is breaking gradle remote cache - lint-definite.xml [349534809]

Fixed

Bug

Status Update

No update yet.

Description

eb...@netflix.com

created issue #1

Jun 26, 2024 01:52PM

DESCRIBE THE ISSUE IN DETAIL:

Lint intermediate files is breaking gradle remote cache because the file ordering in build/intermediates/lint_partial_results/debug/lintAnalyzeDebug/out/lint-definite.xml isn't deterministic (Linux vs MacOS). We noticed it on the IconLocation lint rule.

STEPS TO REPRODUCE:

Build a project with 2 lint warnings on the IconLocation rule
Execute lintDebug on a Mac
Execute lintDebug on a Linux
Compare the files lint-definite.xml from build/intermediates/lint_partial_results/debug/lintAnalyzeDebug/out. In our experience the 2 warnings are swapped.

FILE CONTENT ON MAC

<incident
    id="IconLocation"
    severity="warning"
    message="Found bitmap drawable `res/drawable/internal_hk_modal_preview_asset_1.webp` in densityless folder">
    <location
        file="${:hk-components*debug*MAIN*sourceProvider*0*resDir*0}/drawable/internal_hk_modal_preview_asset_1.webp"/>
</incident>

<incident
    id="IconLocation"
    severity="warning"
    message="Found bitmap drawable `res/drawable/internal_hk_modal_asset_wide_1.webp` in densityless folder">
    <location
        file="${:hk-components*debug*MAIN*sourceProvider*0*resDir*0}/drawable/internal_hk_modal_asset_wide_1.webp"/>
</incident>

FILE CONTENT ON LINUS

<incident
    id="IconLocation"
    severity="warning"
    message="Found bitmap drawable `res/drawable/internal_hk_modal_asset_wide_1.webp` in densityless folder">
    <location
        file="${:hk-components*debug*MAIN*sourceProvider*0*resDir*0}/drawable/internal_hk_modal_asset_wide_1.webp"/>
</incident>

<incident
    id="IconLocation"
    severity="warning"
    message="Found bitmap drawable `res/drawable/internal_hk_modal_preview_asset_1.webp` in densityless folder">
    <location
        file="${:hk-components*debug*MAIN*sourceProvider*0*resDir*0}/drawable/internal_hk_modal_preview_asset_1.webp"/>
</incident>

RESULT

This break gradle remote build cache for CI vs Mac scenario.

IMPORTANT: Please read https://developer.android.com/studio/report-bugs.html carefully and supply all required information.

Studio Build: Android Studio Koala Feature Drop | 2024.1.2 Canary 7 (irrelevant since happen in CI too)
Version of Gradle Plugin: AGP 8.5.0
Version of Gradle: Gradle 8.8
Version of Java: 17.0.11-zulu
OS: Mac Sonoma 14.4.1 (23E224) vs Linux 6.5.0-1020-aws (amd64)

Screenshot 2024-06-26 at 6.21.38 AM.png

17 KB

View

Download

Comments

rp...@google.com <rp...@google.com> Jun 26, 2024 01:59PM

Assigned to an...@google.com.

pa...@google.com <pa...@google.com> Jun 26, 2024 04:01PM

Status: New

tn...@google.com <tn...@google.com> Jul 2, 2024 11:43PM

Assigned to tn...@google.com.

pa...@google.com <pa...@google.com> #2Jul 3, 2024 10:45AM

It looks like LintCliClient provides a function LintCliClient.sortResults() for sorting its definiteIncidents (not provisionalIncidents, and not partialResults), and this ends up being called after partial results are serialized to disk, anyway (in the analysis phase). Perhaps the call to this function should be removed, and instead sorting could occur in LintCliClient.storeState(project) (definite, provisional, partial) and LintCliClient.performReporting() (just definite).

I guess XmlWriter.writePartialResults needs to iterate over the keys (Issues) in order, and XmlWriter.writeLintMap needs to iterate over the keys (Strings) in order.

Message last modified on Jul 3, 2024 10:47AM

tn...@google.com <tn...@google.com> #3Jul 8, 2024 02:45PM

Paul, do you want to grab this? I saw you made it a P1 without assigning it to yourself so I grabbed it on the shuttle right before the 4th of July break but it looks like you've done some investigation already. I reached the same conclusions - the sort call in LintCliClient#run happens to late and it should probably be moved up (and yes, done in store results, as long as it's also done for non-partial clients, such as g3).

The other thing I was thinking is that in unit tests, we do a lot of "normalization" (via TestLintClient overrides), to ensure that tests are stable across platforms. For example, we replace File.separator with /, and \r\n with \n in strings, such that Windows and Linux+OSX have the same expected output. I wonder if we should consider taking some of this out now (in particular, the things that show up in intermediate results) to help have more test coverage for the stability of these artifacts.

pa...@google.com <pa...@google.com> Jul 8, 2024 05:49PM

Reassigned to pa...@google.com.

pa...@google.com <pa...@google.com> #4Jul 18, 2024 01:12PM

Marked as fixed.

Fixed by https://cs.android.com/android-studio/platform/tools/base/+/8ee3d7db053cb6a892c7ea509ccb224e41a1c036

an...@google.com <an...@google.com> #5Jul 22, 2024 04:32PM

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 | 2024.1.3 Canary 2
Android Gradle Plugin 8.7.0-alpha02

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!