391912 - class reflections table never freed if java VM not loaded

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Java: OJI, defect)

Description

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

When starting up Firefox (with about:blank) and shutting it down, I see two allocations leaked (the table and its storage) from the stack:

JSJ_NewHashTable[jsj3250 +0xF76C]
jsj_InitJavaClassReflectionsTable[jsj3250 +0xC693]
jsj_init_JavaClass[jsj3250 +0x7C9E]
JSJ_InitJSContext[jsj3250 +0x1B04]
nsJVMManager::InitLiveConnectClasses[oji +0x6452]
nsJSContext::InitializeLiveConnectClasses[gklayout +0x406C45]
nsJSContext::InitClasses[gklayout +0x408800]
nsJSContext::InitContext[gklayout +0x406AB3]
nsGlobalWindow::SetScriptContext[gklayout +0x3E0F7B]

This seems to be because the table is allocated at the above stack, but JSJ_DisconnectFromJavaVM only frees it if java_vm is non-NULL.

Comment 1

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Actually, there's a second allocation not freed:

JSJ_ConnectToJavaVM[jsj3250 +0x1982]
nsJVMManager::MaybeStartupLiveConnect[oji +0x7151]
nsJVMManager::StartupLiveConnect[oji +0x5F9E]
nsJSRuntime::Init[gklayout +0x409763]

which makes me think the problem may just be that we never call nsJVMManager::MaybeShutdownLiveConnect.

Comment 2

[Eli Friedman]

(In reply to comment #1)
> which makes me think the problem may just be that we never call
> nsJVMManager::MaybeShutdownLiveConnect.

That seems likely: nsJSRuntime::Init() calls StartupLiveConnect, but nsJSRuntime::ShutDown doesn't call ShutdownLiveConnect.

Comment 3

[Benjamin Smedberg]

Mass-closing bugs in the "OJI" component: OJI plugin integration was replaced with npruntime long ago, and these bugs appear to be irrelevant now. If there is in fact a real bug that remains, please file it new in the "Core" product, component "Plug-ins".