18638 - [DOGFOOD] Need to be able to access XPConnect

Status: VERIFIED FIXED

(Core :: Preferences: Backend, defect, P3)

Description

[Chris Waterson]

I *need* to be able to access XPConnect in viewer. This is the only way that I
can isolate some of my bugs. Please make this work!!!!

thanks,
chris

Comment 1

[Chris Waterson]

...and no, viewer doesn't read prefs, so setting the magic XPConnect pref
doesn't work.

Comment 2

[Norris Boyd]

Is there a macro defined for viewer that's off for apprunner? Then I could turn
off security.

Comment 3

[leger]

Putting on PDT+ radar.

Comment 4

[Chris Waterson]

okay, i figured out how to "#if 0" out the thing that was tripping me up. I
don't think you want a macro: the same caps.dll code will be used in both
apprunner and viewer. Maybe some global variable? Or maybe we just need to make
viewer read in prefs. In which case this really isn't your problem.

Comment 5

[leger]

Dogfood, putting on M12 radar.

Comment 6

[leger]

Romoving PDT+ for anoth review per norris' comments via mail:
"This isn't even apprunner, but viewer. Is this really dogfood? Can't people use
apprunner if they need to try out XPConnect components?

I'm trying to figure out how to get my PDT+ bugs fixed in the next 12 days and
not having to worry about this would help."

Comment 7

[rickg]

Clayton -- can you get Norris some help on this one?

Comment 8

[clayton]

Mike, would you look into this one?  Norris has some information:  running
tests requires altered security access which is done via prefs.  Testing is
often done in Viewer which does not use prefs and there isn't a good alternate
mechanism.  Norris suggests a couple of possible approaches:  move the Apprunner
prefs mechanism to viewer so it can handle prefs, or figure out a way to
determine whether viewer or apprunner is in control.  If viewer, turn off
security.  Please let me know if this seems approachable.

Comment 9

[Mike McCabe]

Waterson -

"i figured out how to '#if 0' out the thing that was tripping me up"

... which is that?

Accepting bug as assigned.

Comment 10

[leger]

Putting on the PDT+ radar.

Comment 11

[Mike McCabe]

Here's what I've found out so far.  Documenting, changing component to libPref,
and CC'ing in case somebody can suggest a good solution.

Viewer *has* prefs, but fails silently when initializing the prefs object at
nsPref.cpp:pref_InitInitialObjects().  The relevant code is

    NS_WITH_SERVICE(nsIFileLocator, locator, kFileLocatorCID, &rv);
    if (NS_FAILED(rv))
    	return JS_TRUE;

(*why* return JS_TRUE on failure?)

The NS_WITH_SERVICE call fails to find an nsIFileLocator with an
NS_ERROR_FACTORY_NOT_REGISTERED result.  nsIFileLocator is implemented in
xpfe/appshell; is the viewer seeing it?

All hints welcome.

Comment 12

[neeti]

nsIFileLocator is not being registered with viewer. I tried adding the
nsComponentManager::RegisterComponentLib(kFileLocatorCID,  ...) call to
mozilla/viewer/nsSetupRegistry.cpp file. After this is done, all the prefs get
read. This  is a hack since we are making viewer depend on appshell.dll. I have
attached the diffs below.

Neeti
cvs server: Diffing .
Index: nsSetupRegistry.cpp
===================================================================
RCS file: /cvsroot/mozilla/webshell/tests/viewer/nsSetupRegistry.cpp,v
retrieving revision 1.152
diff -c -r1.152 nsSetupRegistry.cpp
*** nsSetupRegistry.cpp 1999/11/24 03:41:45     1.152
--- nsSetupRegistry.cpp 1999/11/24 17:21:31
***************
*** 50,55 ****
--- 50,58 ----
  #include "nsISound.h"
  #include "nsIFileSpecWithUI.h"

+ #include "nsFileLocations.h"
+ #include "nsIFileLocator.h"
+
  #if defined(XP_UNIX) && !defined(MOZ_MONOLITHIC_TOOLKIT)
  #include "nsIUnixToolkitService.h"
  #endif
***************
*** 67,72 ****
--- 70,76 ----
      #define PLUGIN_DLL "gkplugin.dll"
      #define CAPS_DLL   "caps.dll"
      #define LIVECONNECT_DLL    "jsj3250.dll"
+     #define APPSHELL_DLL "appshell.dll"
      #define OJI_DLL    "oji.dll"
  #elif defined(XP_MAC)
      #define WIDGET_DLL    "WIDGET_DLL"
***************
*** 165,170 ****
--- 169,177 ----
  static NS_DEFINE_IID(kCPluginHostCID, NS_PLUGIN_HOST_CID);
  static NS_DEFINE_CID(kCPluginManagerCID,          NS_PLUGINMANAGER_CID);

+ static NS_DEFINE_CID(kFileLocatorCID,     NS_FILELOCATOR_CID);
+
+
  // OJI
  #ifdef OJI
  static NS_DEFINE_CID(kCLiveconnectCID, NS_CLIVECONNECT_CID);
***************
*** 354,359 ****
--- 361,368 ----
    // PLUGIN
    nsComponentManager::RegisterComponentLib(kCPluginHostCID, NULL, NULL, PLUGIN
_DLL, PR_FALSE, PR_FALSE);
    nsComponentManager::RegisterComponentLib(kCPluginManagerCID, NULL, NULL, PLU
GIN_DLL, PR_FALSE, PR_FALSE);
+
+   nsComponentManager::RegisterComponentLib(kFileLocatorCID,      NULL, NS_FILE
LOCATOR_PROGID, APPSHELL_DLL, PR_FALSE, PR_FALSE);

  #ifdef OJI
    nsComponentManager::RegisterComponentLib(kCLiveconnectCID, "LiveConnect", "c
omponent://netscape/javascript/liveconnect", LIVECONNECT_DLL, PR_FALSE, PR_FALSE
);
cvs server: Diffing mac
cvs server: Diffing public
cvs server: Diffing resources
cvs server: Diffing samples
cvs server: Diffing throbber
cvs server: Diffing unix
cvs server: Diffing unix/gtk
cvs server: Diffing unix/motif
cvs server: Diffing unix/qt
cvs server: Diffing unix/xlib

Comment 13

[Mike McCabe]

Fixed with recent checkin.  Thanks, Neeti.

(Whoops, forgot to put r=neeti on the checkin message :( )

Comment 14

[dshea]

Looks to be an internal development issue.
Waterson has this issue been solved?

Comment 15

[Mike McCabe]

This is just about fixed; I actually had to back out my fix because of bustage.
Waterson, could you confirm after I land?

Comment 16

[Chris Waterson]

yes, just poke the bug and i'll verify

Comment 17

[Mike McCabe]

Fix re-landed.  Waterson?

Comment 18

[Chris Waterson]

marking verified

Comment 19

[leger]

Moving all libPref component bugs to new Preferences: Backend component.  
libPref component will be deleted.