402477 - https error page triggers "ASSERTION: Using observer service off the main thread!"

Status: RESOLVED DUPLICATE of bug 413627

(Core :: Security: PSM, defect, P2)

Description

[Jesse Ruderman]

Steps to reproduce:
1. Load https://dave.sni.velox.ch/.

Result: The https error page appears correctly, but this assertion appears on the console:

###!!! ASSERTION: Using observer service off the main thread!: 'Error', file /Users/jruderman/trunk/mozilla/xpcom/ds/nsObserverService.cpp, line 130


I don't know how to get a stack out of gdb with XPCOM_DEBUG_BREAK=trap when the assertion happens off of the main thread, so here's a stack trace from using XPCOM_DEBUG_BREAK=stack and fix-macosx-stack.pl:

nsObserverService::AddObserver(nsIObserver*, char const*, int) (/Users/jruderman/trunk/mozilla/xpcom/ds/nsObserverService.cpp:130)
nsCertOverrideService::Init() (/Users/jruderman/trunk/mozilla/security/manager/ssl/src/nsCertOverrideService.cpp:158)
nsCertOverrideServiceConstructor(nsISupports*, nsID const&, void**) (/Users/jruderman/trunk/mozilla/security/manager/ssl/src/nsNSSModule.cpp:199)
nsGenericFactory::CreateInstance(nsISupports*, nsID const&, void**) (nsGenericFactory.cpp:80)
nsComponentManagerImpl::CreateInstanceByContractID(char const*, nsISupports*, nsID const&, void**) (/Users/jruderman/trunk/mozilla/xpcom/components/nsComponentManager.cpp:1800)
nsComponentManagerImpl::GetServiceByContractID(char const*, nsID const&, void**) (/Users/jruderman/trunk/mozilla/xpcom/components/nsComponentManager.cpp:2233)
CallGetService(char const*, nsID const&, void**) (nsComponentManagerUtils.cpp:94)
nsGetServiceByContractID::operator()(nsID const&, void**) const (nsComponentManagerUtils.cpp:278)
nsPrintfCString::~nsPrintfCString()+0x00000960  (/Users/jruderman/trunk/mozilla/security/nss/lib/crmf/../../../../dist/include/xpcom/nsCOMPtr.h:1294)
nsPrintfCString::~nsPrintfCString()+0x000009B9  (/Users/jruderman/trunk/mozilla/security/nss/lib/crmf/../../../../dist/include/xpcom/nsCOMPtr.h:677)
nsNSSBadCertHandler(void*, PRFileDesc*) (/Users/jruderman/trunk/mozilla/security/manager/ssl/src/nsNSSIOLayer.cpp:2637)
_ssl3_HandleCertificate (ssl3con.c:7126)
_ssl3_HandleHandshakeMessage (ssl3con.c:7782)
_ssl3_HandleHandshake (ssl3con.c:7898)
_ssl3_HandleRecord (ssl3con.c:8161)
_ssl3_GatherCompleteHandshake (ssl3gthr.c:206)
_ssl_GatherRecord1stHandshake (sslcon.c:1258)
_ssl_Do1stHandshake (sslsecur.c:151)
_ssl_SecureSend (sslsecur.c:1152)
_ssl_SecureWrite (sslsecur.c:1197)
_ssl_Write (sslsock.c:1472)
nsSSLThread::Run() (/Users/jruderman/trunk/mozilla/security/manager/ssl/src/nsSSLThread.cpp:1029)
nsPSMBackgroundThread::nsThreadRunner(void*) (/Users/jruderman/trunk/mozilla/security/manager/ssl/src/nsPSMBackgroundThread.cpp:44)
__pt_root (/Users/jruderman/trunk/mozilla/nsprpub/pr/src/pthreads/ptthread.c:221)
__pthread_body (/usr/lib/libSystem.B.dylib)

Comment 1

[Boris Zbarsky [:bzbarsky]]

Yeah, that's bad.  Don't do that.

Comment 2

[Jesse Ruderman]

https://freaks-unidos.net/x triggers four additional assertions:

###!!! ASSERTION: Using observer service off the main thread!
###!!! ASSERTION: nsMacLocale not thread-safe
###!!! ASSERTION: nsPlatformCharset not thread-safe
###!!! ASSERTION: nsMacLocale not thread-safe
###!!! ASSERTION: nsCharsetAlias2 not thread-safe

Comment 3

[Kai Engert (:KaiE:)]

Does this bug has the chance to make things crash or get corrupted?
Then it should get a higher priority than P5.

Comment 4

[Benjamin Smedberg]

Yeah, this kind of thing can cause instability pretty quickly.

Comment 5

[Carsten Book [:Tomcat]]

reproduced on windows

Comment 6

[Kai Engert (:KaiE:)]

Are you still able to reproduce?

I don't see assertions with the latest build on Linux, neither console, neither error console window.

Comment 7

[Honza Bambas (:mayhemer)]

It was fixed at bug 413627. See http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&root=/cvsroot&subdir=mozilla/security/manager/ssl/src&command=DIFF_FRAMESET&file=nsCertOverrideService.cpp&rev2=1.4&rev1=1.3

Hope you are aware of possibility the code might be reentered what would be disaster.