If I use sudo to run regxpcom, it creates the .mozilla dir with permissions of 2700 owner root, and normal group, in the normal users home dir This prevents me from using sudo, or another 'fake' root shell to install and register the components BuildID: (compiled from source) 2000091317
frb: what sudo are you using (RH, Debian, etc?) In most cases (I believe that this is the RH default, not sure for Debian), sudo does not change the value of $HOME, which is probably why you are seeing the behavior you are seeing. If $HOME is set to /root/, and mozilla still installs in /home/$USER, then there is a problem with mozilla. Otherwise, it is probably a problem with your setup- you need to obtain root in some way that changes $HOME, or change $HOME manually and then install mozilla.
frb: are you still seeing this error? I'm going to close the bug out as INVALID if not. Please post back to the bug and let us know.
I changed my script to create directory and set HOME to it, so I don't know if the offending code is still in regxpcom
This is because of the Special Magic at http://lxr.mozilla.org/seamonkey/source/xpcom/components/nsRegistry.cpp#456 which we should really tear out. XPCOM might well be used by people that don't want to muck with .mozilla, so I think we should only be creating that directory if the application is specifically asking for the per-user registry. scc, what say you? I can whip up a patch.
Created attachment 17482 [details] [diff] [review] Only ensure/create ~/.mozilla or equiv when looking for user registry
adding keywords, relnote: if you run an app that gives you admin priveleges please see this bug to understand what to do. [I'm not quite clear on whether doing is setting $HOME or preventing a setting, or resetting or doing something else]. pdt: this bug would be low risk, but probably should not be plussed until after it is checked into trunk. [And you probably won't plus it, that's fine]
adding the people involved who might replace running "netscape -installer" with "regxpcom && regchrome"
*** Bug 57644 has been marked as a duplicate of this bug. ***
adding ray, who owned the dup.
r=dveditz on Shaver's patch. It's quite safe, just changes the timing on when some existing code is executed so that it's run by the application when needed, but not by applications that simply embed XPCOM. scc: is it possible to get this in? Having .mozilla created by root can't be good :-)
The PDT doesn't even see your comments until this gets reviews and moves to rtm+ state. Since the patch exists, you could get reviews and move to rtm+ rather easily. Is the attached patch the smallest, lowest risk fix possible? Please include a description of the testing done before moving to rtm+. Is there any actual usage of this tool in the packages real users will download and install? If not, this is probably rtm-.
Uh, sure, I can check in shaver's patch. Or shaver can take this bug and check it in himself, whichever is best. Where do we want this? trunk only? or trunk and branch?
> Is there any actual usage of this tool in the packages real users will > download and install? If not, this is probably rtm-. This tool is used by me in the debian packages, and by blizzard in the RPMs. It will probably be used by any other person on a multi-user OS. So no, it isn't used by the official mozilla tarballs on the official ftp site but that should not be a reason to just dismiss it
Either way (trunk or branch) this needs a sr=. Since shaver wrote the patch scc should be able to provide the sr=. Then if you don't want to fight PDT red tape that's all you need to check into the branch, and then mark the bug [rtm-]. If you do want to get this into the branch as well then mark the bug [rtm+] after getting the sr= and wait for their judgement. Include explanation of why we can't ship Netscape 6 without this fix. To get into the branch this needs a sr= (it has an r= from me) and pdt approval. To land on the trunk you still need an sr=. Since Shaver wrote the patch scc should be able to supply the sr=.
sr=scc all reviews are in place, do we want this in? Changing whiteboard status accordingly
rtm-, not ship stopper.
fix checked in
*** Bug 145060 has been marked as a duplicate of this bug. ***
no one verified, but someone filed this bug again. back to the drawing board.
this is now relnoted. i'm gone until sunday and don't intend to look at this before monday, so if this is important to anyone they should look into it ASAP. otoh in practice today's probably the last day for 1.0 so by the time you read this comment you've probably run out of time for that.
why was the bug re-opened? did something regress the previously checked in patch? timeless? what's up with this? re-assigning to you since all the activity seems to be you.
Breakpoint 2, 0x4037a540 in mkdir () from /lib/libc.so.6 (gdb) where #0 0x4037a540 in mkdir () from /lib/libc.so.6 #1 0x4010a568 in do_mkdir (path=0x8083a88 "/tmp/test/.mozilla", flags=170, mode=509, _retval=0xbffff028) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsLocalFileUnix.cpp:383 #2 0x4010a5e4 in nsLocalFile::CreateAndKeepOpen (this=0x80839f8, type=1, flags=170, permissions=509, _retval=0xbffff028) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsLocalFileUnix.cpp:397 #3 0x4010a6f8 in nsLocalFile::Create (this=0x80839f8, type=1, permissions=509) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsLocalFileUnix.cpp:437 #4 0x400d3b01 in nsAppFileLocationProvider::GetProductDirectory (this=0x80541b0, aLocalFile=0xbffff1e8) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsAppFileLocationProvider.cpp:392 #5 0x400d31e7 in nsAppFileLocationProvider::GetFile (this=0x80541b0, prop=0x401977e5 "UserPlugins", persistant=0xbffff238, _retval=0xbffff234) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsAppFileLocationProvider.cpp:248 #6 0x400d47bf in nsAppDirectoryEnumerator::HasMoreElements (this=0x8050028, result=0xbffff300) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsAppFileLocationProvider.cpp:463 #7 0x400d4f1f in nsPathsDirectoryEnumerator::HasMoreElements (this=0x8050028, result=0xbffff300) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsAppFileLocationProvider.cpp:551 #8 0x4014afb8 in AppendFromDirServiceList (codename=0x401b9621 "APluginsDL", aPath=0x804ffe8) at /mnt/ibm/mozhack/mozilla/xpcom/reflect/xptinfo/src/xptiInterfaceInfoManager.cpp:215 #9 0x4014b447 in xptiInterfaceInfoManager::BuildFileSearchPath (aPath=0xbffff3e8) at /mnt/ibm/mozhack/mozilla/xpcom/reflect/xptinfo/src/xptiInterfaceInfoManager.cpp:273 #10 0x4014a531 in xptiInterfaceInfoManager::GetInterfaceInfoManagerNoAddRef () at /mnt/ibm/mozhack/mozilla/xpcom/reflect/xptinfo/src/xptiInterfaceInfoManager.cpp:64 #11 0x40152088 in XPTI_GetInterfaceInfoManager () at /mnt/ibm/mozhack/mozilla/xpcom/reflect/xptinfo/src/xptiInterfaceInfoManager.cpp:2128 #12 0x4009fa0a in NS_InitXPCOM2 (result=0xbffff5f0, binDirectory=0x0, appFileLocationProvider=0x0) at /mnt/ibm/mozhack/mozilla/xpcom/build/nsXPComInit.cpp:598 #13 0x0804afad in main (argc=1, argv=0xbffff664) at /mnt/ibm/mozhack/mozilla/xpcom/tools/registry/regxpcom.cpp:182 #14 0x402cd17d in __libc_start_main () from /lib/libc.so.6 (gdb) up #1 0x4010a568 in do_mkdir (path=0x8083a88 "/tmp/test/.mozilla", flags=170, mode=509, _retval=0xbffff028) at /mnt/ibm/mozhack/mozilla/xpcom/io/nsLocalFileUnix.cpp:383 383 return mkdir(path, mode); (gdb) So the problem is that nsAppFileLocationProvider::GetProductDirectory absolutely creates the directory, it doesn't appear that it needs to. unfortunately my first xpcshell test run resulted in bug 196241
Created attachment 116482 [details] [diff] [review] change nsAppFileLocationProvider::GetProductDirectory to return what the directory would be without forcing it to exist
Hi, Did the install on Mandrake 9 as root. -Choose to delete old mozilla stuff -Received error [ERROR]Fatalerror (-616): Extraction of XPCOM failed[/ERROR] -Read about this bug -Tried the script... no luck there is no mozzila anymore on my machine -Only the users-stuff is hanging around at user level. What to do? Is this the save bug? thanx, Bastiaan
*** Bug 198863 has been marked as a duplicate of this bug. ***
regxpcom doesn't exist any more