45155 - assertion in nsPSMComponent (jar packaging)

Status: VERIFIED FIXED

(Core :: Security: PSM, defect, P3)

Description

[Warren Harris]

One of our readers, Dave Price, from Mt View California writes:

        This assertion is happening when they try to get 
chatzilla/content/chatzillaOverlay.xul out of the .jar file.

Here's the stack trace.

Dave

nsDebug::Assertion(const char * 0x100cb564, const char * 0x100cb54c,
const char * 0x100cb518, int 355) line 246 + 13 bytes
nsServiceManagerImpl::GetService(nsServiceManagerImpl * const
0x00b16f68, const nsID & {...}, const nsID & {...}, nsISupports * *
0x02a7fb7c, nsIShutdownListener * 0x00000000) line 355 + 44 bytes
nsServiceManagerImpl::GetService(nsServiceManagerImpl * const
0x00b16f68, const char * 0x033efbfc, const nsID & {...}, nsISupports * *
0x02a7fb7c, nsIShutdownListener * 0x00000000) line 494
nsServiceManager::GetService(const char * 0x033efbfc, const nsID &
{...}, nsISupports * * 0x02a7fb7c, nsIShutdownListener * 0x00000000)
line 607
nsGetServiceByProgID::operator()(const nsID & {...}, void * *
0x02a7fb7c) line 66 + 22 bytes
nsCOMPtr<nsISignatureVerifier>::assign_from_helper(const nsCOMPtr_helper
& {...}, const nsID & {...}) line 856 + 18 bytes
nsCOMPtr<nsISignatureVerifier>::nsCOMPtr<nsISignatureVerifier>(const
nsCOMPtr_helper & {...}) line 553
nsJAR::VerifyEntry(const char * 0x02cca1b0, const char * 0x026804c0,
unsigned int 591) line 708 + 33 bytes
nsJARInputStream::Init(nsJAR * 0x02cca228, const char * 0x02cc9540, int
1) line 89 + 38 bytes
nsJAR::CreateInputStream(const char * 0x02cc9540, int 1, nsIInputStream
* * 0x02cc9f3c) line 362 + 20 bytes
nsJAR::GetInputStream(nsJAR * const 0x02cca228, const char * 0x02cc9540,
nsIInputStream * * 0x02cc9f3c) line 293
nsJARChannel::GetInputStream(nsJARChannel * const 0x02cc9038,
nsIInputStream * * 0x02cc9f3c) line 1000 + 37 bytes
nsFileTransport::Process() line 397 + 71 bytes
nsFileTransport::Run(nsFileTransport * const 0x02cc9eec) line 352
nsThreadPoolRunnable::Run(nsThreadPoolRunnable * const 0x024963b8) line
689 + 12 bytes
nsThread::Main(void * 0x02496738) line 84 + 26 bytes
_PR_NativeRunThread(void * 0x02496780) line 399 + 13 bytes
MSVCRTD! 1020c323()
KERNEL32! 77e92ca8()

nsAppShellService::Run(nsAppShellService * const 0x00e4e7f8) line 387
main1(int 1, char * * 0x00b16b48, nsISupports * 0x00000000) line 913 +
32 bytes
main(int 1, char * * 0x00b16b48) line 1099 + 37 bytes
mainCRTStartup() line 338 + 17 bytes
KERNEL32! 77e87903()


My pithy comeback...

Warren Harris wrote: 

  It looks like there's a circular dependency between nsPref, and 
nsPSMComponent. I don't quite see it, but I'm
  guessing that nsPSMComponent initialization needs prefs, but 
pref_VerifyLockFileSpec needs the
  nsPSMComponent (the nsISignatureVerifier). Doug should probably look at this. 

  Doug: It's only manifest with Dave's jar packaging stuff that he's working on. 

  Dave: Can you send us a simple jar manifest that will allow us to reproduce 
this problem? Also, maybe you could set
  a breakpoint on nsPSMComponent::CreatePSMComponent and pref_VerifyLockFileSpec 
and show us their stack
  traces. 

  Warren


Doug Turner replies...

psm definitely depends on libpref.  In fact, when we are created, we get a whole 
mess of preferences: 
http://lxr.mozilla.org/seamonkey/source/extensions/psm-glue/src/nsPSMComponent.c
pp#539 

But it looks like this code: 
http://lxr.mozilla.org/seamonkey/source/modules/libpref/src/nsPref.cpp#1331 

is only called when we are messing around with a lock file: 
http://lxr.mozilla.org/seamonkey/source/modules/libpref/src/nsPref.cpp#428 

Also, it seams very wrong that this returns success when the 
nsISignatureVerifier does not exists. 
  

Dave, when you have reduce this to a test case, let me know.

Comment 1

[Doug Turner (:dougt)]

Assert is here:

http://lxr.mozilla.org/seamonkey/source/xpcom/components/nsServiceManager.cpp#35
5

Comment 2

[Doug Turner (:dougt)]

There must be another thread doing something.  I do not see any reentrant calls 
to the service manager in these stacks.  

Comment 3

[dprice (gone)]

I've emailed the test case to Doug.  If it works without any problems, I'll post 
it in the bug as well.

Dave

Comment 4

[dprice (gone)]

To reproduce this defect:

- make certain you have the latest verions of the following:
        mozilla/config/make-jars.pl
        mozilla/config/rules.mak
        mozilla/netwerk/base/src/nsIOService.cpp
- rebuild mozilla/netwerk
- check out mozilla/extensions/irc/jar.mn
- go to mozilla/extensions/irc and run nmake -f makefile.win xinstall
- delete *.rdf in mozilla/dist/WIN32_D.OBJ/bin/chrome/
- make certain that installed_chrome.txt is in 
mozilla/dist/WIN32_D.OBJ/bin/chrome/ and the contents match the sample at the 
end of this comment
- remove the directory
mozilla/dist/WIN32_D.OBJ/bin/chrome/packages/chatzilla/ 
- run mozilla

If you have problems, just let me know.
Dave

<<begin installed_chrome.txt>>
content,install,url,resource:/chrome/packages/widget-toolkit/ 
content,install,url,resource:/chrome/packages/core/ 
locale,install,url,resource:/chrome/locales/en-US/ 
content,install,url,jar:resource:/chrome/packages/chatzilla.jar!/ 
locale,install,url,jar:resource:/chrome/packages/chatzilla.jar!/ 
skin,install,url,jar:resource:/chrome/packages/chatzilla.jar!/ 
skin,install,url,resource:/chrome/skins/modern/ 
skin,install,url,resource:/chrome/skins/classic/ 
locale,install,url,resource:/chrome/locales/en-GB/ 
content,install,url,resource:/chrome/packages/messenger/ 
<<end installed_chrome.txt>>

Comment 5

[Doug Turner (:dougt)]

running nmake -f makefile.win xinstall does not work.

There are files that are not checked in that are listing in the jar.mn file.  

Comment 6

[Doug Turner (:dougt)]

also I had to manually enter the perl command because xinstall (name sucks) does 
not work:

+++ make: packaging jars from jar.mn
+++ jaring ..\..\dist\WIN32_D.OBJ\bin\chrome/packages\chatzilla.jar
The name specified is not recognized as an
internal or external command, operable program or batch file.

Comment 7

[Doug Turner (:dougt)]

--> mstoltz.

mstoltz I think that this is yours

Comment 8

[Jay Patel [:jay]]

Putting on [nsbeta2-] radar. Not critical to beta2. 

Comment 9

[Warren Harris]

PDT: Removing [nsbeta2-] because this bug is blocking the jar file packaging bug 
18433.

Comment 10

[Mitchell Stoltz (not reading bugmail)]

This looks like it has something to do with Neeti's code at 
http://lxr.mozilla.org/seamonkey/source/modules/libpref/src/nsPref.cpp#1331 .
She calls nsISignatureVerifier(which is another interface to nsPSMComponent) in 
order to calculate an MD5 hash. It would be much more convenient if we could put 
an MD5 function into Mozilla rather than have to call over to PSM to do it.

Comment 11

[dprice (gone)]

I'll pull the changes and test in my tree.

Comment 12

[dprice (gone)]

I pulled a new tree this evening and I am still seeing the assert.  

to reproduce:
pull and build a new tree
go to mozilla/extenions/irc and run nmake -f makefile.win xinstall
edit mozilla/dist/WIN32_D.OBJ/bin/chrome/installed_chrome.txt as mentioned above
run mozilla

Comment 13

[Jay Patel [:jay]]

Putting on [nsbeta2+] radar for beta2 fix, but only to prevent the blocking of 
bug 18433.

Comment 14

[Mitchell Stoltz (not reading bugmail)]

Could someone (warren?) give me instructions for reproducing this bug? I'll take 
a look at it.

Comment 15

[Warren Harris]

I fixed this.

Comment 16

[leger]

Updating QA Contact.

Comment 17

[dprice (gone)]

at gbush's request, I've verified that this is fixed.

Comment 18

[timeless]

mozilla/extensions/psm-glue/src/nsCrypto.cpp 	1.10
mozilla/extensions/psm-glue/src/Attic/nsCrypto.cpp 	1.10