410706 - setTimeout and setInterval stops repeating after system time date or clock changed

Status: NEW

(Core :: DOM: Core & HTML, defect, P5)

Description

[Matias Jose]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.11) Gecko/20071127 Firefox/2.0.0.11
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.11) Gecko/20071127 Firefox/2.0.0.11

We have to timers there reflecting time in two different ways. If I change the system time, they will stop working. The function called by a setInterval statement just stops being called afterwards

Reproducible: Always

Steps to Reproduce:
1.Open the link
2.Change system time

Actual Results:  
The timers stopped (setInterval sopped working)

Expected Results:  
I was expecting the timers to keep running on every setInterval function call.

Comment 1

[Brendan Eich [:brendan]]

No crash involved though, right? "crash" means abnormal/abrupt termination of the Firefox process.

/be

Comment 2

[Bob Clary [:bc] (inactive)]

Matias, can you remove the console.log call? It is undefined in Minefield.

Comment 3

[Jeff Walden [:Waldo]]

Note for the fix: make sure the invisible "lateness" argument provided to the called function is the correct value, too.

Comment 4

[WADA:World Anti-bad-Duping Agency]

(In reply to comment #0)
> If I change the system time, they will stop working.
> The function called by a setInterval statement just stops being called afterwards

Phenomenon depends on "change of system time" (far future or far ancient) and software's (Firefox in your case) implementation which relates to timer (setInterval/setTimeout when user coded JavaScript).
"From what date/time to what date/time" did you change system-time?
What is your "setInterval" value?

Comment 5

[Matias Jose]

(In reply to comment #1)
> No crash involved though, right? "crash" means abnormal/abrupt termination of
> the Firefox process.
> 
> /be
> 

Correct, no crash involoved. just setInterval stops working

Comment 6

[Matias Jose]

(In reply to comment #4)
> (In reply to comment #0)
> > If I change the system time, they will stop working.
> > The function called by a setInterval statement just stops being called afterwards
> 
> Phenomenon depends on "change of system time" (far future or far ancient) and
> software's (Firefox in your case) implementation which relates to timer
> (setInterval/setTimeout when user coded JavaScript).
> "From what date/time to what date/time" did you change system-time?
> What is your "setInterval" value?
> 

Any time change on windows XP will cause the setInterval to stop, when clicking on "Apply" or "Ok" button. I tested it by changing seconds (+- 30 sec) and hours (+- 1 hour)

Comment 7

[Bob Clary [:bc] (inactive)]

Attachment: testcase

on linux, if i set the time to the past the timers will appear to stop for the period of the adjustment. setting the time into the future does not cause any stoppage.

Comment 8

[Bob Clary [:bc] (inactive)]

ditto on windows. invalid?

Comment 9

[Matias Jose]

(In reply to comment #2)
> Matias, can you remove the console.log call? It is undefined in Minefield.
> 

See if it works now, console is an object which belongs to firebug extension. I put it on a conditional, it should be called if console doesn't exist. But it's usefull to notice when the function is called (or not)

Comment 10

[Bob Clary [:bc] (inactive)]

(In reply to comment #9)

yes, that is essentially the same modification I made. The timer works now in Firefox without the extension.

Comment 11

[Pavlos Touboulidis]

(In reply to comment #8)
> ditto on windows. invalid?
> 

The timer stops when I set the time to one hour in the past. The same thing happens with setTimeout().

Aren't setInterval/setTimeout supposed to fire after a delay (in milliseconds) and not at a fixed currentDate+delay? Even if that was the case, there's a bug here because of the inconsistent behavior when changing the system time to the past or to the future.

IE7 & Opera 9.5 do not seem to have this problem.

Comment 12

[Dan]

I've got the exact same problem. I'm also pretty certain that this wasn't happening in the past (but I can't say I know exactly which version).

"Aren't setInterval/setTimeout supposed to fire after a delay (in milliseconds)
and not at a fixed currentDate+delay? Even if that was the case, there's a bug
here because of the inconsistent behavior when changing the system time to the
past or to the future." - YES!

I think that what's happened here is that someone has changed either (1) the way that these two functions measure time to be based on the new Date() function or (2) they're always used the new Date() function, but the code that handles possible client changes to the past has been removed.

I will try to fix this and report back.

Comment 15

[dziastinux]

Has been fixed in 4.0b11?

Comment 16

[nemo]

Has this been fixed in 53b?

(I'm sorry, I couldn't resist)

Comment 17

[Marion Daly [:mdaly]]

overholt: Is this still an issue?