Need the ability to install an overlays.rdf file which points to non existant xul overlays. This is a work around until the installer poke the chrome registry. see bug http://bugzilla.mozilla.org/show_bug.cgi?id=27178 waterson, I am assigning this to you since I understand that you own this sort of thing. ping me back if I am wrong.

hey, can you guys try out the attached patches and see if they work for you? rginda, I'd especially be interested to see if this solves your chatzilla problems.

PDT: I'm removing the PDT- on this, because of the amount of hassle that this "bug" has caused already. (I think that four verification builds have had to be respun because of this!) Fixing this will avoid future meltdowns: "I forgot to add blah.xul to the packager; now the browser won't start!" Instead, the problem will be localized to "hey, there's no 'blah' menu item in the build that I downloaded from sweetlou."

Without this fix, tweaking the UI for any other reason can result in a mysterious hang which then has to be tracked down by the regression team. waterson: are you at least reporting a warning for missing files?

I meant to say "We should take this fix NOW".

I thought about making a warning dialog, but then realized (e.g., ChatZilla) that it may be fine to have a missing overlay. So for now, I have a printf() wrapped in #ifdef DEBUG. I could take out the #ifdef DEBUG and *always* printf()...

We often run into missing XUL files at the release build level because someone fumbled the installer handoff. I guess this is my vote for always printing out a warning.

Putting on PDT+ for beta1. Please fix by 03/08

fix checked in, r=harishd, a=jevering

John, please verify

I really hate to do this, but Reopening ... win98 is OK (modulo some small errors), but linux and mac are both DOA (fail to run). (Let me know if this is an inappropriate test, but) The specific test I used to check: 1) with a build that is known to launch and work correctly 2) shut it down 3) move ./content/default/NavSecurityOverlay.xul out of the way 4) start mozilla (with -console on win98) What happens: ---------------------------------------------------------------------- win98 2000030813 opt comm bits - error message "*** Failed to load overlay ..." prints to console - browser continues to load - after fully loaded there are a few problems - the home page is not automatically loaded - the sidebar is empty (need to do 'View->Sidebar' twice to populate it) - the throbber is initally painted in the wrong place - however the browser functions acceptably well, as well as mailnews, prefs panels and the sidebar. ---------------------------------------------------------------------- linux rh6.0 2000030809 opt comm bits - error message "*** Failed to load overlay ..." prints to console - shortly thereafter the browser stops moving forward (just sits and spins) - final bit of console messages: I am inside the initialize Hey : You are in QFA Startup (QFA)Talkback loaded Ok. WEBSHELL+ = 2 assuming d&d is off for Navigator nsCollationUnix::Initialize mLocale = C nsCollationUnix::Initialize mCharset = ISO-8859-1 *** Failed to load overlay chrome://navigator/content/NavSecurityOverlay.xul /u/jrgm/mozilla-netscape/package/Cool/Xpcs Registry.dat: invalid ELF headernsCollationUnix::Initialize mLocale = C nsCollationUnix::Initialize mCharset = ISO-8859-1 WEBSHELL+ = 3 The next message should be "Setting content window", if all was well. ---------------------------------------------------------------------- mac 8.6 2000030813 -- mac also just sits and spins -- macsbugs shows various stack traces like the following: Calling chain using A6/R1 links Back chain ISA Caller 00000000 PPC 0B8F1EA4 03B46240 PPC 0B8E11C4 main+0016C 03B461B0 PPC 0B8E0B0C main1(int, char**, nsISplashScreen*)+00624 03B460C0 PPC 0B73E188 nsAppShellService::Run()+3E188 03B46080 PPC 0B71280C nsAppShell::Run()+00038 03B46030 PPC 0B712F18 nsMacMessagePump::DoMessagePump()+00028 03B45FE0 PPC 0B712FC8 nsMacMessagePump::GetEvent(EventRecord&)+0003C 03B45F90 PPC FFD43348 WaitNextEvent+00028 03B45EA8 68K 001954D6 'scod BFAF 0002'+01EB6 03B45E7C 68K 0047B576 03B45E38 68K 0019555E 'scod BFAF 0002'+01F3E 03B45E26 68K 0019DCB8 'scod BFAF 0002'+0A698 03B45DC5 PPC 0022A35C EmToNatEndMoveParams+00014 03B45D40 PPC 0BEEC794 03B45D00 PPC 0BEEC73C Closing log Calling chain using A6/R1 links Back chain ISA Caller 00000000 PPC 0B8F1EA4 03B46240 PPC 0B8E11C4 main+0016C 03B461B0 PPC 0B8E0B0C main1(int, char**, nsISplashScreen*)+00624 03B460C0 PPC 0B73E188 nsAppShellService::Run()+3E188 03B46080 PPC 0B71280C nsAppShell::Run()+00038 03B46030 PPC 0B712F18 nsMacMessagePump::DoMessagePump()+00028 03B45FE0 PPC 0B712FC8 nsMacMessagePump::GetEvent(EventRecord&)+0003C 03B45F90 PPC FFD43348 WaitNextEvent+00028 03B45EA8 68K 001954D6 'scod BFAF 0002'+01EB6 Closing log

Ok, let's change the definition of the bug :-). Yes, if you remove components from the browser, Bad Things will happen. The point of fixing this was that, now, at least you will be warned before the Bad Things start to happen.

per the amended summary, marking verified (with the caveat that Mac has no console/warning).