97397 - Preferences from netscape.cfg and autoconfig.jsc are being written to prefs.js

Status: RESOLVED INCOMPLETE

(Core :: Preferences: Backend, enhancement)

Description

[lrg]

I see this as a problem primarily because of a case in which the netscape.cfg is 
removed or modified (if a new netscape.cfg file is distributed for example).  
While preferences set in the netscape.cfg take presidence over those in the 
prefs.js, any preferences that were set inside of a previous .cfg but are no 
longer present, would be executed. 

Additionally there is some concern over allowing the values present inside of 
the netscape.cfg file to the average users.  While the hashing mechanism isn't 
very secure, having all of the values written to the prefs.js file is much less 
so.

Comment 1

[Mitesh Shah]

I am not sure why does communicator writes the preferences set in netscape.cfg
back to prefs.js. Is there any way we can avoid writing the autoconfig
preferences  back to prefs.js ?

Comment 2

[tao]

mitesh is no longer with us. reassigning to default module owner. bug 113387
tracks the original list of mitesh bugs

Comment 3

[tao]

Is this a regression from Communicator 4.x? I suspect what might have happened
is that N6 writes in the in-memory prefs back to prefs.js. The in-memory prefs
values are filled with the overwritten settings from netscape.cfg. 

Comment 4

[Brian Nesse (gone)]

Yes, that is exactly what happens. When the preferences are written out any
preferences in memory, which are not default values, are written to prefs.js. I
do not believe this functionality is any different from that of 4.x.

Comment 5

[tao]

If I undersand this correctly: the pref values (from cfg) got written into 
user profile's prefs.js only when the values in the prefs.js are different from 
those in cfg. Am I right? Does it matter if the pref values are locked or not?

Could QA confirm how 4.x behaves? thx!

Comment 6

[lrg]

Only locked prefs are being written into the prefs.js file, others are not. 
This is the same behavior as we see in netscape 4.79, so it isn't a regression.

Comment 7

[tao]

Luke: thanks for the comment. looks like this is a RFE. Let's revisit it in the
future.

Comment 8

[Jeff Hooker]

removed jhooker from cc: list

Comment 9

[Brian Nesse (gone)]

*** Bug 171225 has been marked as a duplicate of this bug. ***

Comment 10

[djk]

Please update the summary to include 'user.js' also, as that's what my duped bug
was about.

Comment 11

[Brant Gurganus]

[RFE] is deprecated in favor of severity: enhancement.  They have the same meaning.

Comment 12

[Conrad Carlen (not reading bugmail)]

Taking some of Brain's bugs since he's no longer here.

Comment 13

[(not currently active) Ted Mielczarek]

*** Bug 252307 has been marked as a duplicate of this bug. ***

Comment 14

[Carlo 'Carcarlo' Carmagnini]

(In reply to comment #13)
> *** Bug 252307 has been marked as a duplicate of this bug. ***

Is there a solution for a user to fix this bug? maybe we've to delete manually 
the line with the general.useragent.override string in prefs.js?
There are a lot of people who change the UA ID for just a while, then delete it 
(some online banks refuse Firefox). I think it's a big bug, just for all the 
people who browse the www not knowing they're giving a wrong UA ID to the sites 
they visit. 

Comment 15

[timeless]

[e-mail]: in about:config you can right click and choose reset.

the right thing to do for this bug is sorta complicated, sorta simple, and 
almost entirely unrelated to your problem.

preferences need to remember who set them, if the user set them, then they need 
to be serialized <period>. otherwise you suffer from the problem where 
switching between apps will eventually set your prefs to match those of the app.
if the user did not set the pref, then it should not be serialized <period>.

Comment 16

[OstGote!]

IMO the user.js should be removed from the summary (the dupes are wrong). The
user.js issue was invalidated with Bug 50038.

Comment 17

[Mike Kaply [:mkaply]]

This is so old, I'm going to mark it as incomplete.