Integrated terminal resets scroll position to top when emacs is closed

[mjbvz]

Issue Type: Bug

Reopening of #62329

steps

1. In the intergrated terminal.
2. Run a command that generates more than one screen of content: ls -alF /dev/
3. Scroll to bottom
4. Now run emacs and then close emacs using ctrl+c ctrl+x

Bug
Terminal scroll position is sometimes reset to top of buffer

VS Code version: Code - Insiders 1.40.0-insider (86405ea, 2019-11-06T16:56:25.487Z)
OS version: Darwin x64 18.7.0

System Info

Item	Value
CPUs	Intel(R) Core(TM) i7-4770HQ CPU @ 2.20GHz (8 x 2200)
GPU Status	2d_canvas: enabled flash_3d: enabled flash_stage3d: enabled flash_stage3d_baseline: enabled gpu_compositing: enabled metal: disabled_off multiple_raster_threads: enabled_on oop_rasterization: disabled_off protected_video_decode: unavailable_off rasterization: enabled skia_renderer: disabled_off surface_control: disabled_off surface_synchronization: enabled_on video_decode: enabled viz_display_compositor: enabled_on viz_hit_test_surface_layer: disabled_off webgl: enabled webgl2: enabled
Load (avg)	2, 2, 2
Memory (System)	16.00GB (0.02GB free)
Process Argv	-psn_0_1614218
Screen Reader	no
VM	22%

[Tyriar]

Can't repro 🙁 some things that might help diagnose:

• Debug VS Code to see what's getting hitting hit. Calls to this when exiting would be particularly interesting https://github.com/xtermjs/xterm.js/blob/dd4eb7bdecf1c13748c27de65c7bc32d95e3d30e/src/Terminal.ts#L1027
• Try repro in the xterm.js demo
• Enable escape sequence logging and capture the entering emacs logs and exit emacs logs for both the case where it's fine and the case where the scroll is off.

[mjbvz]

Seems to happen most reliably when I run emacs as the git editor, but as the screenshot shows it can also happen for normal runs.

From debugging:

Normal case

• scroll called with undefined
• _onScroll called twice with _ignoreNextScrollEvent == true
• scrollLines not called by _onScroll

Bugged case

• scroll called with undefined
• _onScroll called with _ignoreNextScrollEvent == true
• _onScroll called again _ignoreNextScrollEvent == false
• scrollLines called by _onScroll with -3145

[Tyriar]

_onScroll called again _ignoreNextScrollEvent == false
scrollLines called by _onScroll with -3145

That's definitely the problem, could you get the stacktrace for both of those?

[mjbvz]

Stacks look the same. Here _onScroll called with _ignoreNextScrollEvent === true:

`:

And with _ignoreNextScrollEvent === false:

---

It seems like a timing issue. Between the call to _innerRefresh which sets _ignoreNextScrollEvent = true and _scroll, someone else is setting _ignoreNextScrollEvent = false. Here's one possible call sequences I see opening emacs in the bad case (true/false is the value of _ignoreNextScrollEvent ):

And from the good case:

For what it's worth, I also currently see this more often in insiders (vs OSS)

[Yanpas]

I don't use emacs or something, just run some tests which output a lot of stuff, then hide and unhide terminal - after a couple of moments I am on top of terminal. Scrolling it with a scrollbar to the bottom doesn't bring me to the last line, so I press enter

[rianquinn]

Yeah, I have to press enter to get to the bottom of the terminal. I can confirm that this is an issue with Linux and Windows (I see this on both), so it doesn't appear to be specific to any OS. This also happens with both native and remote development, so it is not specific to remote development (which I use a lot). For me, it starts happening after about an hour of use, and then becomes really frequent (almost every time I open the terminal). Its causing me to form a habitual response to just always press enter when I open the terminal ☹️