Closed Bug 36781 Opened 24 years ago Closed 24 years ago

Seg fault on exit running the Linux installer

Categories

(SeaMonkey :: Installer, defect, P2)

Product:
SeaMonkey
Component:
Installer
Platform:
x86
Linux
Type:
defect

Tracking

(Not tracked)

Status:
VERIFIED FIXED

People

(Reporter: depman1, Assigned: samir_bugzilla)

References

Details

(Keywords: crash, platform-parity, Whiteboard: [nsbeta2+] sgehani's fix ETA: 07/22) 

Linux only. build 2000-04-21-13-M16.
1. Download stub installer for Linux. gzip and tar extract.
2. Run the installer.
3. Continue pressing Next and then OK.
4. Continues through downloading and installing phases.
5. Watch console window.
Result: Get these "cannot unlink" and "cannot remove directory" msgs (for .tmp 
directories). Then, a segmentation core dump. No TalkBack available.
Have to redo unlinking.  This is mine.
Assignee: cathleen → sgehani
Target Milestone: --- → M16
Status: NEW → ASSIGNED
QA Contact: jimmylee → depstein
This doesn't have to do with unlinking.  It is from within the Linux installer 
code when exiting.  Investigating.
Component: Installer: XPInstall Engine → Installer
Priority: P3 → P1
Adding crash keyword.
Keywords: crash
exiting linux installer will cause crash
need to fix in beta2
Keywords: nsbeta2
Still crashes. I did a diff between unzip/extraction of M16 (in folder called
may08-unzip) and installer package folder (containing all files installed before
the core crash). Here is the diff:

/u/depstein/builds/M16/may08-unzip >diff package/ ../may08-inst2/
Only in package/: TestGtkEmbed
Only in package/: bloaturls.txt
Common subdirectories: package/chrome and ../may08-inst2/chrome
Only in ../may08-inst2/: component-1.reg
Binary files package/component.reg and ../may08-inst2/component.reg differ
Common subdirectories: package/components and ../may08-inst2/components
Common subdirectories: package/defaults and ../may08-inst2/defaults
Common subdirectories: package/icons and ../may08-inst2/icons
Only in package/: libcmt.so
Only in package/: libgtkembedmoz.so
Only in package/: libgtkxtbin.so
Only in package/: libjpeg.so
Only in package/: libprotocol.so
Only in package/: libzlib.so
Only in package/: mozilla-config
Only in package/: mozilla-installer-bin
Common subdirectories: package/plugins and ../may08-inst2/plugins
Only in ../may08-inst2/: registry
Common subdirectories: package/res and ../may08-inst2/res
Common subdirectories: package/searchplugins and ../may08-inst2/searchplugins
Only in package/: timebombgen
Only in package/: xpidl
Only in package/: xpt_dump
Only in package/: xpt_link
/u/depstein/builds/M16/may08-unzip >ls
mozilla-i686-pc-linux-gnu.tar  package/
re is the diff:


After a discussion with David, we decided to open a separate bug regarding his 
last comment which has to do with packages-unix manifest differences rather than 
install failure.  He plans on following up with a bug number for reference.
submitted bug 38603. pasted diff file listing (below) into that bug. 38603 
addresses a crash after entering a URL in the Linux installer build.
This isn't directly related to the bug, but there was an error that went by
stating that components.reg was not found (I was running the install perl script
in a tree within which I had not uet started mozilla). I don't think we should
be packaging components.reg (if that is what is being attempted).
Syd,
We won't be paackaging the components.reg eventually.  Thanks for pointing this 
out.  Folks have vacillated over this issue much.
Putting on [nsbeta2+] radar for beta2 fix.
Whiteboard: [nsbeta2+]
Blocks: 37900
No longer blocks: 37900
just a quick 'me too'...

crashed for me twice with may 11 build. even during download part of the
install, it's sucking 100% CPU which is a good tip off something is already
wrong at that point.

first crash wasen't really... it just hung installing a file named
'Shareware.<something>', second was while on Mail & News.
Retargeting to M17.  This is a bug, not a feature.
Target Milestone: M16 → M17
After consultation with dp, bug 40074 filed against XPCOM.  We have a workaround 
for this release: don't unload xpistub (which leaves xpcom around in turn).  
Fix checked in.
Status: ASSIGNED → RESOLVED
Closed: 24 years ago
Resolution: --- → FIXED
message on console today:

nNCL: registering deferred (0)
Segmentation fault (core dumped)

build 2000052308
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
This is not the same bug.  Here is it is failing to autoreg.  Please open a 
a separate bug for this problem.  My bug is an installer app seg fault on exit 
problem.  Resolving again.

Grace, the new bug should be nominated for nsbeta2 please.  Thanks.
Status: REOPENED → RESOLVED
Closed: 24 years ago → 24 years ago
Resolution: --- → FIXED
opened bug 40310
*** Bug 40310 has been marked as a duplicate of this bug. ***
Misinterpreted depstein's and gbush's comments.  Sorry.  Reopening.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Status: REOPENED → ASSIGNED
This was another problem (I introduced while fixing the xpcom related crash on 
exit).  Fix checked in.
Status: ASSIGNED → RESOLVED
Closed: 24 years ago → 24 years ago
Resolution: --- → FIXED
this is happening- same message as above- on build 2000052508
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Here's the deal: if you run as an X client on the same machine that the X server 
is (i.e. without exceed or rlogin and setting a remote display) then the 
installer works fine.  Else, it dies.  So, that's a different bug.  But, since 
the empirical result is the same we'll track it here.
Status: REOPENED → ASSIGNED
results 5/26 build
1. using exceed to rlogin to remote machine (usual test)
   installer gives the following message on console (does not core dump)
    nNCL: registering deferred (0)
    Xlib: unexpected async reply (sequence 0x14437)!
it hangs - ctrl-c out
installed files appear to be there but when I try to launch- this error appears:
error in loading shared libraries: libraptogfx.so: cannot open a shared object 
file etc....

2. logging in directly to a machine
intaller  core dumps as before
Logging in directly I didn't core dump.  Will investigate Grace's env further.
Oops, examining gbush's build revealed that the fix for the seg fault while 
displaying locally is in my tree but not in.  Will check that in.  Will continue 
investigating the cause of the remote display seg fault.  Sorry for the 
confusion.  
Should no longer crash when displaying on same machine as X server is running.  
Remote displaying still seg faults on exit (i.e., partial fix checked in).  
Continuing investigation.
Priority: P1 → P2
Summary: Crash (core dump)after running Linux installer. → Seg fault on exit running the Linux installer on a remote display
*** Bug 43788 has been marked as a duplicate of this bug. ***
*** Bug 44399 has been marked as a duplicate of this bug. ***
Summary: Seg fault on exit running the Linux installer on a remote display → Seg fault on exit running the Linux installer
This happened to me repeatedly on a non-remote display as well. See bug 44399.
Changing summary to reflect this.
Whiteboard: [nsbeta2+] → [nsbeta2+] sgehani's fix ETA: 07/14
downloaded installer on linux, crashed.  7/7 bits off of sweetlou.


I still get the installer bug:

./mozilla-installer
Passive mode on.
Passive mode on.
Passive mode on.
nNCL: registering deferred (0)
Segmentation fault

I'm using Mandrake 7.0 w/Gnome 1.2 (HelixGnome) and the Sawfish window manager.
I've had this problem for awhile now.
Ooops: mozilla build 2000071020
*** Bug 41784 has been marked as a duplicate of this bug. ***
still seen on Linux commercial build 2000-07-13-08-M17.
adding me to CC
Samir, I see you checked in something for this.  Is it completely fixed now?
Fixed "local display" bug.  "Remote display" bug is probably thread related.
Summary: Seg fault on exit running the Linux installer → Seg fault on exit running the Linux installer remote displaying
Whiteboard: [nsbeta2+] sgehani's fix ETA: 07/14 → [nsbeta2+] sgehani's fix ETA: 07/19
still on linux commercial build 2000-07-18-08-M17 and instead of going into the 
installation and taking a little time before the core dumps it had the 
segmentation right after i clicked on the install button.
I got this morning's build (I guess 2000-07-18 M17) and I didn't get the
segfault. But after going thru the install process, I get:

> sudo ./mozilla-installer
Passive mode on.
Passive mode on.
Passive mode on.
Passive mode on.
Passive mode on.
Passive mode on.
nNCL: registering deferred (0)


> ./mozilla
./run-mozilla.sh ./mozilla-bin
MOZILLA_FIVE_HOME=.
  LD_LIBRARY_PATH=.
     LIBRARY_PATH=.
       SHLIB_PATH=.
          LIBPATH=.
       ADDON_PATH=.
      MOZ_PROGRAM=./mozilla-bin
      MOZ_TOOLKIT=
        moz_debug=0
     moz_debugger=
Could not obtain CmdLine processing service

Whenever I got the segfault, I would get the same CmdLine error reported here.
So it's not crashing but it's not installing completely either.

Oh, the 
ftp://ftp.mozilla.org/pub/mozilla/nightly/2000-07-17-20-M17/linux-xpi

I ran it again and got:

Passive mode on.
Passive mode on.
Passive mode on.
Passive mode on.
Passive mode on.
Passive mode on.
nNCL: registering deferred (0)

Gtk-CRITICAL **: file gtkprogress.c: line 518 (gtk_progress_set_percentage):
assertion `percentage >= 0 && percentage <= 1.0' failed.

Gtk-CRITICAL **: file gtkprogress.c: line 518 (gtk_progress_set_percentage):
assertion `percentage >= 0 && percentage <= 1.0' failed.

Gtk-CRITICAL **: file gtkprogress.c: line 518 (gtk_progress_set_percentage):
assertion `percentage >= 0 && percentage <= 1.0' failed.

Gtk-CRITICAL **: file gtkprogress.c: line 518 (gtk_progress_set_percentage):
assertion `percentage >= 0 && percentage <= 1.0' failed.

Gtk-CRITICAL **: file gtkprogress.c: line 518 (gtk_progress_set_percentage):
assertion `percentage >= 0 && percentage <= 1.0' failed.

I tried again and had the same error. Weird.

This bug is to track the seg fault issue.  I'd like to keep it at that so 
pointy-heads parsing for info can get to the crux quickly.  In the future, I'd 
appreicate it if new bugs are created for issues that are other than the 
installer actually seg faulting.  Thanks.
Keywords: pp
I'm getting this crash too.
Changed title - has nothing to do with being remote.

Here is a stack:


Program received signal SIGILL, Illegal instruction.
0x40400521 in ?? ()
(gdb) where
#0  0x40400521 in ?? ()
#1  0x8054c82 in nsXIEngine::Install ()
#2  0x805342d in nsInstallDlg::WorkDammitWork ()
#3  0x40230ce9 in pthread_start_thread (arg=0xbf7ffe7c) at manager.c:204

Samir: do you have any clues on this, or need help?
Summary: Seg fault on exit running the Linux installer remote displaying → Seg fault on exit running the Linux installer
7/19 is past and Samir won't be back until Friday.
Whiteboard: [nsbeta2+] sgehani's fix ETA: 07/19 → [nsbeta2+] sgehani's fix ETA: 07/22
45716 is the fact that the linux installer doesn't work.  No working linux
installer and Netscape is shipping a tarball full of misc cruft for PR2.

FYI - for anyone working with the installer, if the installer fails it will
leave files around with "-1" in them, i.e. components/libnecko-1.so.  I've found
that often times deleting the component.reg file and lib*-1.so can sometimes
make the installer build work.
Let's try this once again: the linux installer crashes when remote displaying.  
I have isolated this bug (but due to the threaded nature am still working on 
it).  I *know* this bug exists.  If you are seeing a crash local displaying 
(another manifestation of which I fixed a couple of days ago) then *please* log 
*another* bug.  Let's leave this bug for the remote display problem.  Thank you.
Summary: Seg fault on exit running the Linux installer → Seg fault on exit running the Linux installer remote displaying
*** Bug 46110 has been marked as a duplicate of this bug. ***
*** Bug 46082 has been marked as a duplicate of this bug. ***
Summary: Seg fault on exit running the Linux installer remote displaying → Seg fault on exit running the Linux installer
Since bug 45716 is NSCP-only permission'd, and Samir wants this bug for tracking
remote display issues, is there any public bug open on the fact that linux
installer hasent worked for about 4 month, on a local display? Segfault every
time, on every build.
Fixed.  

No, really, it's fixed this time: remote and local display.  No longer crashes 
on exit.  What's that?  Don't believe me?  Try it out for yourself with the next 
veriication builds in the morning.  Really.  :o)
Status: ASSIGNED → RESOLVED
Closed: 24 years ago → 24 years ago
Resolution: --- → FIXED
I still get a segmentation fault with linux commercial build 2000-07-24-08-M17 
but later in the installation tha before; reopening
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Please verify this with the mozilla installer.  The commercial installer is 
hosed due to bug 46263 and is separate from this bug.
Status: REOPENED → RESOLVED
Closed: 24 years ago → 24 years ago
Resolution: --- → FIXED
works for me on 2000072408 build
no seg fault on the Mozilla build 2000-07-24-08-M17. Verifying this bug.
Status: RESOLVED → VERIFIED
*** Bug 47151 has been marked as a duplicate of this bug. ***
Product: Browser → Seamonkey
You need to log in before you can comment on or make changes to this bug.