Skip to content

[Database][Android] SIGSEGV in Database::DeleteInternal when embedded bridge-class init fails; FindClassInFiles passes JNI-form names to loadClass on Android #1936

Open

Description

@Tb1281

[Database][Android] SIGSEGV in Database::DeleteInternal when embedded bridge-class init fails; FindClassInFiles passes JNI-form names to loadClass on Android

SDK: Firebase C++ SDK 13.12.0, Realtime Database (libFirebaseCppApp-13_12_0.so), arm64 Android.

Two related defects, one crash. Reporting both together because the second turns the first from "error return" into "process death".

1. Init failure crashes instead of returning an error (the SIGSEGV)

When Database::GetInstance cannot finish DatabaseInternal init (in our case: embedded Java bridge classes fail to resolve, see §2), the code path ends in delete database → Database::DeleteInternal, which dereferences a null App* (MakeKey(app=null) → app->name()).

Tombstone (deterministic, every run once the init fails):

signal 11 (SIGSEGV), SEGV_MAPERR, fault addr 0x8 (null+8 read)
#00 libFirebaseCppApp-13_12_0.so  firebase::database::Database::DeleteInternal()+192
#01 libFirebaseCppApp-13_12_0.so  firebase::database::Database::GetInstance(...)+792
#02 libFirebaseCppApp-13_12_0.so  Firebase_Database_CSharp_InternalFirebaseDatabase_GetInstanceInternal

Expected: init failure returns an error / null instance. Actual: SIGSEGV. This half is environment-independent — any init failure ends in a crash rather than a diagnosable error, which is also what made §2 so hard to root-cause (no error, just death).

2. Android branch of FindClassInFiles skips the slash→dot conversion the desktop branch does

app/src/util_android.cc, FindClassInFiles (verified on current main):

  • Desktop branch converts before calling loadClass:
    std::string class_name_str(class_name);
    std::replace(class_name_str.begin(), class_name_str.end(), kPathSeparator, '.');
    with LogDebug("Load class %s (a.k.a. %s)", ...).
  • Android branch passes the raw JNI-form name straight into DexClassLoader.loadClass:
    LogDebug("Load class %s", class_name);
    jstring class_name_object = env->NewStringUTF(class_name);
    ... kLoadClass ...

ClassLoader.loadClass takes binary names (dotted). On typical standalone apps ART tolerates the slashed form so nobody notices; under a custom class-loader setup the lookup fails. Preceding log line in our runs:

E firebase: Java class com/google/firebase/database/internal/cpp/CppTransactionHandler not found.

Suggested fix: the same 3-line std::replace in the Android branch — makes both branches consistent, zero behavior change where lookup already succeeds.

Environment (disclosed so this isn't mis-triaged)

The game runs hosted in another app's process (launcher creates the game context via createPackageContext and hosts its activity). That is why the class-loader topology differs from standalone and why §2 bites here first. But §1 needs no hosted setup to matter: any GetInstance init failure on any app crashes today. And §2 is a contract violation that only happens to work standalone.

Happy to attach the full tombstone + logcat and to test a patched build. Thanks for the SDK.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions