1465579 - Crash when breakpointing a page with devtools.debugger.new-debugger-frontend;false

Status: RESOLVED INVALID

(Core :: DOM: HTML Parser, defect, P3)

Description

[nemo]

STR:
①  Disable devtools.debugger.new-debugger-frontend  (I had done this long ago in the original profile when the new frontend was utterly unusable)
② Go to http://bl.ocks.org/mbertrand/raw/5218300/
③ Breakpoint line 74
④ Hit Refresh
⑤ Hit Refresh again

Crashed reliably every time with that flag enabled.
https://crash-stats.mozilla.com/report/index/1b1a506f-03b7-4f72-a4fc-e4cc70180530

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

What has this to do with GC?

Comment 2

[nemo]

Er... Sorry removed "in GC" - I think I'm confusing it with another crash I was encountering:
https://crash-stats.mozilla.com/report/index/7b751351-aa0e-4acf-b00d-c57090180530  - that one which is actually in stable in and a totally different profile - apologies in particular to sfink for the confusion


And to Ollie who replied before I could correct this, causing mid-air collision 

Comment 3

[Henri Sivonen (:hsivonen)]

To understand the severity here: Is setting devtools.debugger.new-debugger-frontend to false considered a supported configuration?

Comment 5

[Henri Sivonen (:hsivonen)]

It seems that the old debugger front end makes JS run at a time when the HTML parser believes JS execution to be impossible.

Is the old debug considered supported? If not, is it going to be removed?

Does the new debugger front-end guarantee that it won't run JS at times when normal Web semantics would prevent JS execution?

Comment 6

[Brian Grinstead [:bgrins]]

(In reply to Henri Sivonen (:hsivonen) from comment #5)
> It seems that the old debugger front end makes JS run at a time when the
> HTML parser believes JS execution to be impossible.
> 
> Is the old debug considered supported? If not, is it going to be removed?

The old debugger UI is not on by default in any environment. Last I heard we were waiting on a feature or two to remove it (see Bug 1314057), but at this point I believe it's considered unsupported as far as bugfixes. A crash may be extra motivation to go ahead and remove the code. Harald, is that a fair summary?

> Does the new debugger front-end guarantee that it won't run JS at times when
> normal Web semantics would prevent JS execution?

That I can't say, but forwarding to Jason who may have an idea.

Do you know the particular line in the debugger JS (or stack trace) that's causing the crash? That would help us check to see if the new UI could also get into this scenario.

Comment 7

[:Harald Kirschner :digitarald]

> A crash may be extra motivation to go ahead and remove the code. Harald, is that a fair summary?

Plan is to remove the debugger in 64 (while speccing out event handler debugging for implementation in 65, which was the only feature that we considered a blocker).

Comment 8

[Henri Sivonen (:hsivonen)]

(In reply to Brian Grinstead [:bgrins] from comment #6)
> Do you know the particular line in the debugger JS (or stack trace) that's
> causing the crash?

I don't anything beyond the repro steps here.

(In reply to :Harald Kirschner :digitarald from comment #7)
> Plan is to remove the debugger in 64

In that case, it's probably not worthwhile to change the HTML parser here.

Comment 9

[Jason Laster [:jlast]]

We are in the process of removing the old debugger.

Comment 10

[nemo]

Triaging my bugs, noticed this one still open. I very much doubt anyone cares about the old debugger anymore. Closing.