State of last subscription modification it not restored after restart

[Ross]

Environment

ABP Edge 0.9.9
Edge 38.14393.0.0 / Windows 10

How to reproduce

1. Install ABP for Edge.
2. Add all of the custom subscriptions from the first run page.
3. Open the ABP Options page.
4. Enable the Adblock Warning Removal List subscription.
5. Restart edge.
6. Observe the Adblock Warning Removal List is disabled again.

Observed behaviour

The state of the Adblock Warning Removal List is not restored (see attached video).

Expected behaviour

The state of the Adblock Warning Removal List (and all others) to be restored to the values set by the user before restart.

Notes

Reproduced 4/6 times using the steps above. Could reproduce using when disabling a different list at Step 4 (such as one just added from the first run page like malware). If two subscriptions are disabled, only one is restored (second video).

Could not reproduce when just adding one list from the first run page (tried a couple of times).

[oleksandr]

I cannot reproduce it 6/6. Could it be that you were closing Edge before the changes have been saved? (It does take more time to save the lists with more filters being present especially on slower hardware). Can you reproduce it with a more substantial delay?

[Ross]

It does not seem reproducible with a longer delay (8-10s). Is reproducible with a relatively short delay (1-3s). Not a major issue as most users won't close the browser after changing settings so quickly.

[sebastian]

Also I'm not sure whether we can do something about it on our end anyway if Edge terminates before the data get a chance to be written. Closing the issue, for now, and moving forward with the release.

[mapx]

ABP could keep the settings (what lists are enabled, other parameters) separate from the filters.
Even if someone closes quickly edge at least the list of lists will be saved (in the settings file / sql container).

On start, it should check if the 2 entities are coherent (in case there arent: the complete filters list will be regenerated).

[sebastian]

If it is IndexedDB, or our code serializing the filter data, being slow here, that could help. But if other means of storage will have a similar delay (not unlikely), this wouldn't change anything.

Either way, this doesn't seem to be a new regression, neither is it established how likely this issue will be noticed by users. Note that Ross presumably used a fairly slow virtual machine, and closing the browser just after changing your filter settings isn't a particular common use case either.

[Ross]

Yes it was on a fairly slow virtual machine. I do access to a proper non-VM Win 10 machine sometimes so I'll be sure to try this out again there.