Theia: [git] "Stopped/Started watching" git repo, on every change

[marcdumais-work]

Description

Every time a file is changed, I see a pair of traces in the backend log, saying that the git repo watching has stopped and right after, started again.

I do not know what this exactly means, but it sounds like it could be that the OS file watchers, watching a git repo, are being freed and re-allocated, again and again (hopefully not the case).

Reproduction Steps

• build and start the browser example
• watch the terminal where the backend was started
• open a file in theia's editor.
• make changes in the file, saving once in a while or letting auto-save do it for you
  -> each time a change is saved, a pair of traces (see below) appears in the B-E log.

OS and Theia version:
ubuntu 16.04, theia browser example app, built from latest master

Diagnostics:

root INFO Stopped watching the git repository: file:///opt/ericsson-dev/theia
root INFO Started watching the git repository: file:///opt/ericsson-dev/theia

[akosyakov]

If it is reproducible we have to fix it asap. Probably another scm PR bug (

[akosyakov]

yes, i was able to reproduce @vinokurig Can you take care about it today? cc @evidolob

[vinokurig]

@akosyakov Yes, working on it

[vinokurig]

@akosyakov Are you sure that this bug is relates to the SCM PR? I've reverted it localy and I could reproduce the bug.

[akosyakov]

@vinokurig checking now against https://github.com/theia-ide/theia/commit/2a1c24973ed9833c54b5b2838f52b184f3343be2

[vinokurig]

I've prepared a PR about refactor Git module to SCM https://github.com/theia-ide/theia/pull/5339 that fixes this bug

[akosyakov]

@vinokurig i cannot reproduce it with revision before SCM PR

I've prepared a PR about refactor Git module to SCM #5339 that fixes this bug

checking

[akosyakov]

5339 is too big and breaking, is there a way for small fix of this issue?

[vinokurig]

5339 is too big and breaking, is there a way for small fix of this issue?

I'll try to extract a particular fix for it

[vinokurig]

@akosyakov Looks like it is not possible to extract a particular fix. All the changes are needed. I am working on the comments from the PR.

[akosyakov]

@marcdumais-work Can we do 7.0.1 without SCM fixes? I won't have time to review today and it will be breaking. And we do later 7.1.0 this week?

[marcdumais-work]

@anton yes, we can go ahead today with a 0.7.1 bugfix release, and discuss later this week what we'll do with the SCM fixes (0.7.2 bugfix release?)

[lmcbout]

Thanks @kittaakos great improvement with issue #5284 , I notice that the watchers with electron for the git repos start about 6 times, then if you wait about ~20-30 secondes, a timer error occurs and then start/stop the git repo occurs in a loop. Was reproduce-able on a window environment this morning (~5-6 times in a row)

[vince-fugnitto]

Thanks @kittaakos great improvement with issue #5284 , I notice that the watchers with electron for the git repos start about 6 times, then if you wait about ~20-30 secondes, a timer error occurs and then start/stop the git repo occurs in a loop. Was reproduce-able on a window environment this morning (~5-6 times in a row)

I don't believe it has to do with electron, the following can also be seen in the browser app.