1891726 - Thread names are overridden by JavaVM::AttachCurrentThread

Status: RESOLVED FIXED

(GeckoView :: General, defect)

Description

[Jamie Nicol [:jnicol] (on leave until 21st October)]

This has been a minor annoyance of mine for years, but never enough for me to track down the cause, until today. Some thread names appear in the Android Studio debugger as "Thread-<n>". eg the "Renderer" thread always appears as "Thread-3". We set the thread names when the threads are spawned, but they are getting overwritten, and today I captured the place this happens in a debugger:

    prctl 0x0000758b0f5e8400
    pthread_setname_np 0x0000758b0f5e71af
    art::SetThreadName(const char *) 0x0000758b0ce379a7
    art::Thread::CreatePeer(const char *, bool, _jobject *) 0x0000758b0ce08954
    art::Thread::Attach(const char *, bool, _jobject *, bool) 0x0000758b0ce073a2
    art::Runtime::AttachCurrentThread(const char *, bool, _jobject *, bool) 0x0000758b0cde8a17
    art::JII::AttachCurrentThreadInternal(_JavaVM *, _JNIEnv **, void *, bool) 0x0000758b0cc8fb59
    art::CheckJII::AttachCurrentThread(_JavaVM *, _JNIEnv **, void *) 0x0000758b0ca7e8e5
    [Inlined] _JavaVM::AttachCurrentThread(_JNIEnv **, void *) jni.h:1058
    mozilla::jni::GetEnvForThread() Utils.cpp:170
    mozilla::gl::CreateSurfaceFromNativeWindow(mozilla::gl::EglDisplay &, void *, void *) GLContextProviderEGL.cpp:184

It seems the JVM overwrites it when we first attach a thread. That was enough information to easily find the chromium bug when they encountered this issue.

Seems we can easily fix this by supplying the thread name as an arg to AttachCurrentThread()

Comment 1

[Jamie Nicol [:jnicol] (on leave until 21st October)]

Attachment: Bug 1891726 - Supply thread name to JavaVM::AttachCurrentThread(). r?#geckoview-reviewers

By default, AttachCurrentThread() overwrites the existing thread name
with "Thread-<n>", making it hard to identify these threads in a
debugger. By supplying the existing thread name as an argument to
AttachCurrentThread, we ensure that the thread keeps its name as
desired.

Comment 2

[Markus Stange [:mstange]]

Oh, nice, Denis and I were confused by this when we looked at a perfetto trace yesterday.

Comment 3

[Jeff Muizelaar [:jrmuizel]]

I was curious where the actual Thread-<n> was coming from and it seems to be here: https://cs.android.com/android/platform/superproject/+/master:libcore/ojluni/src/main/java/java/lang/Thread.java;l=705;drc=9a73823472955a225341bbd7905200b46b280a79

Comment 4

[Jamie Nicol [:jnicol] (on leave until 21st October)]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #3)

I was curious where the actual Thread-<n> was coming from and it seems to be here: https://cs.android.com/android/platform/superproject/+/master:libcore/ojluni/src/main/java/java/lang/Thread.java;l=705;drc=9a73823472955a225341bbd7905200b46b280a79

The comment directly underneath that is amusing

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/dbc4dbc65bbc
Supply thread name to JavaVM::AttachCurrentThread(). r=geckoview-reviewers,jrmuizel,m_kato

Comment 6

[Cristian Tuns]

https://hg.mozilla.org/mozilla-central/rev/dbc4dbc65bbc