75445 - Core dump after exiting browser after triggering specific xpi

Status: RESOLVED INVALID

(Core Graveyard :: Installer: XPInstall Engine, defect)

Description

[Jimmy Lee]

Build: 2001-04-09-09-trunk(LINUX)

1. From http://jimbob/trigger3.html, click on Acceptance drop-down and choose
   a_execute_unix
2. Click Trigger case button
3. Click OK button from confirm dialog
4. Exit browser

RESULT:
Core dump occurs.  The Install.log looks like the installation was successful.  
The executable called from a_execute_unix is suppose to create a text file 
a_execute_unix.txt in my home directory, but this did not occur.  Running this 
executable from the command line does successfully create the expected text 
file.

EXPECTED RESULT:
No core dump.  a_execute_unix.txt is found from the home directory.  The file 
contains the current date and time stamp.

Comment 1

[Jimmy Lee]

Adding keywords accordingly.  Nominating for Beta.  I don't know what kind of 
executables may be called, but I know this one appears to cause a core dump.  
Calling the other test a_execute_unix_parameter does work correctly.

Comment 2

[Syd Logan]

Gee that dialog is huge (the one to select stuff to install) much wider than my
screen.

Comment 3

[Syd Logan]

A couple of questions:

The dialog that I think you refer to as a confirm dialog -- is this the one that
is shown in the attached gif? 

How come there is nothing in the list? And "OK" is enabled? Please give me some
information on how this dialog should work, and what is supposed to be in the
list. The fact that the list is empty for me makes me wonder if that is the real
bug; hitting OK here makes zero sense to me.

Comment 4

[Syd Logan]

Attachment: full screen shot (empty list, wacko dialog size)

Comment 5

[Syd Logan]

Oh, and I didn't core dump. So, Jimmy, if you can duplicate this for me I'd like
to run down to your cube and see this.

Comment 6

[Daniel Veditz [:dveditz]]

How old is your tree? These are symptoms of the xul and skin whackage and were 
recently fixed.

Comment 7

[Jimmy Lee]

I can still duplicate this bug with today's build.  Feel free to drop in.

Comment 8

[Syd Logan]

I'm not crashing here. I've seen the full sized window before (that was my gif
that was attached, after all) but it works for me now. Puzzling. What is
*really* annoying is that after I hit the OK button, the install progress dialog
(???) comes up, flashes, goes away, and the user has zero clue whether the
install actually happened or not. Terrible UI. So, what is the reason for this?
Is it the install script, or the engine? Let me know so I can take a look at
fixing it, or point me at a bug that may already exist on this issue.

Comment 9

[Jimmy Lee]

I think it is the responsibility of the install script writer to control some of 
the users' experience.  In this particular case, the executable should provide 
some UI for the end user, so he/she knows that something is happening.  My 
test is just a test, so the UI is poor.  Also, an install script writer can 
display a simple alert indicating the installation is complete.

Personally, I like that we do not try to present a "canned" installation.  We 
allow the install script writer to manage the "look" that the end user receives 
through javascript and third-party executables as examples.

Comment 10

[Syd Logan]

Ok, fine (I'm just learning about this stuff, so thanks for the patience).

Next question, is this still happening for you? Or is the bug worksforme? I'd 
recommend keeping this on the daily test list until we can be certain that the 
UI changes that are apparently behind this crash are actually resolved and 
solid.

Comment 11

[Jimmy Lee]

I am still able to reproduce the bug with today's build, 2001-04-24-08-trunk.

Comment 12

[Jimmy Lee]

I am still able to reproduce the bug with today's build,  2001-06-05-04-0.9.1.

Comment 13

[Blake Ross]

Missed 0.9.3.

Comment 14

[Daniel Veditz [:dveditz]]

Jussi, any rationale for marking this nsenterprise+? Does that indicate eClient
is volunteering resources?

Comment 15

[Syd Logan]

Can I have the source code for "test_unix"? When I run this manually I do *not*
get any files created in my home directory.

Thanks,

syd

Comment 16

[Jimmy Lee]

Attachment: Source for executable in xpi package

Comment 17

[Hong Kwon]

not a stopper for emojo, removing nsenterprise+

Comment 18

[Jimmy Lee]

Nominating.

Comment 19

[Syd Logan]

over to dprice

Comment 20

[chris hofmann]

dprice is on sabitcal. moving to next milestone. 

Comment 21

[Syd Logan]

Resetting milestone of all nsbeta1-bugs, only nsbeta1+ bugs should have a target
milestone.

Comment 22

[Syd Logan]

Resetting milestone, only nsbeta1+ bugs can have a milestone on them, these are
niminated, but not yet plussed.

Comment 23

[Daniel Veditz [:dveditz]]

Jimmy: does this still happen? If so, is it possible it's a duplicate of bug 123649?

Comment 24

[Jimmy Lee]

Today the browser is freezing during Installing...and no core dump occurs.  I do
not think this is relate to bug 123649 because the executable are different. 
But I cannot be 100% sure.

Comment 25

[David Epstein]

Test case can be triggered from
http://www.mozilla.org/quality/smartupdate/xpinstall-trigger.html
Select a_execute_unix from the Acceptance Test Case menu.
Bug needs to be reassigned.

Comment 26

[Jesse Ruderman]

I don't think this bug is still valid.  Support for install.js is gone.  Also, most of the files in this bug are gone, making it very hard to retest.