1249962 - Can't inspect variables in debugger after stepping out of function

Status: RESOLVED WORKSFORME

(DevTools :: Debugger, defect, P2)

Description

[hbenl]

Attachment: bug.html

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:44.0) Gecko/20100101 Firefox/44.0
Build ID: 20160211213400

Steps to reproduce:

- open the attached file "bug.html" in Firefox
- open the debugger
- set a breakpoint at line 6 (where it says "return o.x;")
- click the button labeled "Click me" -> the debugger stops at the breakpoint
- click "Step Out" twice
- try to inspect the variables in the "Variables" pane (by clicking on the triangles next to the variable names) 


Actual results:

The properties of the "this" variable are shown, but not the properties of any other variable.
Also, only the contents of the top scope ("Function scope [f]") are shown but the contents of the other scopes ("Block scope" and "Global scope [Window]") can't be opened.


Expected results:

I should be able to inspect all variables and all scopes.

Comment 1

[hbenl]

I have reproduced this with Firefox 44.0.2 on Windows 7 and on Arch Linux.

Another interesting observation: The script in the attached test case contains a function h() which is never called but would call the function g() (which is the function that I step out of to trigger the bug). If I remove h() or rewrite it so that it wouldn't call g(), the bug is not triggered anymore!

Comment 2

[Eddy Bruel [:ejpbruel]]

Haven't actually confirmed this yet, but if this is an actual bug, it breaks basic functionality that should just work. Marking this as P1 because we should look into it asap.

Comment 3

[hbenl]

Attachment: firefoxbug.png

Reproduced with Firefox 45.0 on Arch Linux

Comment 4

[hbenl]

Attachment: firefoxbug-firebug.png

Reproduced with Firefox 45.0 using Firebug on Arch Linux

Comment 5

[hbenl]

I have just re-tested this with Firefox 45.0 on Linux and Windows 7. The bug is still there.
I have attached 2 screenshots showing the bug, the first is done with the Firefox debugger, 
the second with Firebug.
In the Firebug screenshot you can see the error message "No such actor for ID: server1.conn0.pausedobj139".

Could someone please have a look at this? Reproducing this bug with the attached example takes
just a few seconds...

Comment 6

[Eddy Bruel [:ejpbruel]]

(In reply to hbenl from comment #5)
> I have just re-tested this with Firefox 45.0 on Linux and Windows 7. The bug
> is still there.
> I have attached 2 screenshots showing the bug, the first is done with the
> Firefox debugger, 
> the second with Firebug.
> In the Firebug screenshot you can see the error message "No such actor for
> ID: server1.conn0.pausedobj139".
> 
> Could someone please have a look at this? Reproducing this bug with the
> attached example takes
> just a few seconds...

Like I said in comment 2, this is high on our priority list, but unfortunately, that list is quite long. We only have two people (myself including) actively working on the debugger backend, and they are both occupied with other tasks at the moment.

There's only so much we can do. Sorry for not getting to this sooner!

Comment 7

[hbenl]

OK, sorry for being a bit impatient.

In case this helps, I have tried the attached test case with a few different versions of Firefox
and the bug first appears in version 41.0b1, whereas the test case works fine in 40.0.3.
When I start Firefox 41.0b1 from the console and run the test case, the following messages appear
in the console:

console.warn: noSuchActor: No such actor for ID: server1.conn0.pausedobj71
console.warn: noSuchActor: No such actor for ID: server1.conn0.pausedobj37

Comment 8

[James Long (:jlongster)]

We will look at this but we have other P1 bugs that we are fixing right now.

Comment 9

[hbenl]

This may be related to https://bugzilla.mozilla.org/show_bug.cgi?id=1262654, 
which is another scenario where the debugger responds with noSuchActor messages

Comment 10

[hbenl]

I just re-tested this with Firefox 58.0.1 (64-bit, on Arch Linux) and found that the bug is gone!