8237 - [PP] Prefs need to save when apprunner quits

Status: VERIFIED FIXED

(SeaMonkey :: UI Design, defect, P1)

Description

[Steve Lamm]

I use a preference to save the open/close state of the Sidebar. When I
implemented this last week, it only worked if I explicitly saved the prefs file
after changing the preference. I should not have to save the prefs at all.

My JavaScript is in contentframe.js,

http://lxr.mozilla.org/mozilla/source/xpfe/browser/src/contentframe.js

I may have only tested this on Linux, but think I had the same behavior on
Windows too.

Comment 1

[don]

Hmmmm, I thought the prefs were being saved.

Comment 2

[Steve Lamm]

I see this bug in a build from today on Linux. My Windows build is messed up, so
I have not retested it there. This does not need to get fixed for M7, since I
can save the file explicitly as a work around.

Comment 3

[Steve Lamm]

This works if I save the prefs explicitly (except on some Windows builds where
prefs are broken, bug #8384). So, I have a work around for M7. Is this a dup? I
seem to remember writing this somewhere else.

Comment 4

[mcmullen]

This is working today. Marking worksforme.

Yesterday, with a breakpoint on nsPref::SavePrefsFileWith(), the variable
gErrorOpeningUserPrefs was set to true, preventing saving the file. Today, this
is not occurring (why? I don't have time to investigate).

I ran the browser, opened the sidebar, closed the window. The pref was correctly
saved at application shutdown.

Steve, can you verify this on unix?

Comment 5

[Steve Lamm]

The current checked in contentframe.js calls SavePrefsFile(). If I take that
out, the "sidebar.open" pref never gets saved.

I left in the SavePrefsFile() calls as a work around.

Comment 6

[Steve Lamm]

The actual fix could wait until M8.

Comment 7

[don]

Move to M8 ...

Comment 8

[mcmullen]

Hey Don, not so fast and frantic, please!  Maybe Steve's case is fixed, but if

SavePrefFile is broken, then maybe stuff from the prefs window is not being

saved, either. Steve, is this the case on unix? If this is broken, it's a major

regression.

Comment 9

[leger]

Clearing the WorksForMe since this is now reopened.

Comment 10

[mcmullen]

There are two bugs.

1. If the prefs filed doesn't exist, then there is a bug that sets an error flag,
preventing writes to the prefs file thereafter. This occurs on Mac and Windows,
and it would occur on Linux, except that there's a worse bug there (#8106).

2. On Linux, bug 8106 means that instead of executing the cleanup code in
nsAppRunner/ main, the program just calls exit(). The code that saves the prefs
file is never called.

Comment 11

[mcmullen]

I checked in the fix for two XP problems here. The Linux problem (calling exit())
prevents fixing this for Linux.

Comment 12

[mcmullen]

Since the only case left here is Linux, and since it depends on 8106 (currently
assigned to ramiro), I'm assigning this to ramiro, too. Setting the platform to
Linux now, since that is the only remaining case.

Comment 13

[mcmullen]

Since this never worked on Linux (how could it?) I'm changing the [REGRESSION] to
[PP].

Comment 14

[ramiro]

I just checked in a fix for this.

Marking fixed.

Comment 15

[rubydoo123]

verified using 0714 build