Release: 2.0.9.107 (Canary) 2024-09-11

Repros for flakies in general haven’t been readily findable, otherwise they would be given.
It’s also possible that the 3 A.M wake that Windows does to check for updates is part of it.

In general it seems like more tabs open in Duplicati is more prone to having odd behavior.
When I was testing with just one, GUI was less flaky. Recently I’ve had more than normal.

You can see above where I was leaving a trail of broken ones. The next bunch was due to upgrading to 2.0.9.108. TrayIcon doesn’t seem to emit to stdout any URL to initial connect, therefore I open yet more tabs through TrayIcon request. Anyway, here’s one specific test:

Make a backup of a short file. Duplicate the tab to make 8 total. Reload the tabs if you like.
Run Compact now maybe 10 times (no need to be super fast). Maybe a status bar will stop responding to Compact now, or will go into a progressstate loop even though all is done.

For a more vivid visual display, do the above with a Destination folder without write access.
Survey tabs later to see which ones gave a red popup. No access at all gives similar result.

In either case, wander around the tabs to look for status bars that are stuck or don’t update.

EDIT 1:

Browser is Edge on Windows 10, but I don’t know if it matters.

EDIT 2:

Refreshed all eight tabs, had PC sleep (except for Window wake at around 3 local, 7 UTC).
Woke this morning with Connection lost.popups on all eight, though dev tools log varied.
FWIW there’s a 15 minute sleep on wake set to delay a job that would otherwise raise load.

Six wound up like this, with a very dark Duplicati screen underneath (look hard):

image

with roughly this stop:

image

Two wound up with home page visible below, less dimmed from the usual look:

image

with roughly this stop:

image

The 11:07 UTC entries would be around the PC wake time by me.

EDIT 3:

Since one can see the jobs in the background of the second image,
the first one began as a no write test, allowing read and list access.
This took more advanced ACL work, but no access got same result.
For faster fail, set number-of-retries=0. Job test 1 was normal.
Main oddity with it was it didn’t show up on Save – until I refreshed.

EDIT 4.

The night after my eight tab Edge test, I did a four tab Chrome test.
All failed, half with the super dark background and longer stop path.
Half with the slightly greyed background and the shorter stop path.

One or two tabs seem to survive. All this may be system dependent.
Coming out of sleep is a super busy time for a system, and it’s slow.

Last night I tried four Edge tabs without a past-due job at wake time.
Typical wake is around 7, and I moved job from 5:20 to 7:20. Result:
All failed, just like four tab Chrome test, with 50/50 split failure mode.

I’ll say again that this is now 2.0.9.108, but the history is here in 107.