Checking multiple items in quick succession causes "Corrupted" diffs

[BryanWall]

I wanted to mention something that I've seen happen, and have been able to reproduce. Let me know if you would like an issue open for this. It can be avoided once you know about it, but it's possible this could have side effects that break something.

If you either edit/save, recheck, or add new items quickly one after the other, the snapshots from one item end up in another item, and consequently the diff is calculated by comparing the previous snapshot from one item with the new snapshot of a different item. Even doing a re-check doesn't get the correct snapshot, at least not the first few rechecks. I was able to eventually get it to grab the correct snapshot, but I'm not sure why, maybe something being cached.

This happens using the webdriver. I tried testing with "Basic fast" instead of the webdriver, but later noticed the change to the fetch setting didn't save.

I was able to consistently reproduce it by doing this:

1. Copy a URL to the clipboard
2. Click Recheck on an item ("Item-A")
3. Quickly paste the URL ("Item-B") into the box at the top and click Watch

What happens:

• Both items show "Checking now" at the same time
• The content from Item-B ends up in the snapshot for Item-A

I think any sequence that triggers a check of multiple items in quick succession can cause this behaviour. It's probably dependent on the response time of the webdriver. Is it possible to wait until one check finished before starting the next?

In case this is helpful - the most recent time I tried this, the following was at the top of the diff for both items:
Unable to read /datastore/30ac4862-9c20-4a1b-b508-0c244feed198/f361bd2c-b3a3-456c-8fc7-aaa2b601b25c.stripped.txt

Edit: Ok, now I'm getting Unable to read /datastore/... on other items, also, so maybe this issue does cause some general data corruption. Even deleting history snapshot data did not fix this. I had to delete the item and re-add it to get it to work again.

[BryanWall]

Here are a couple of screen shots. I tested adding checks for Hacker News and Slashdot. I didn't actually do this too quickly, because I was trying to do something else, but this is what happened. The UI showed the check for the first was finished before I added the second, but something hadn't "settled" somewhere in the code or database. Notice that for both checks the "before" is from Hacker News and the "after" is from Slashdot.

[dgtlmoon]

how many watches do you have?

[dgtlmoon]

Yeah weird, thanks for the report, I can reproduce this in a slightly different way - have an existing item in 'check' status, then add a new item, for some reason the preview link appears for the new item, when I click on that item - i get the preview for the other item

[dgtlmoon]

@BryanWall alright, i think i found something, building master now, could you try once the new docker container is out (about 35 minutes) or pull latest?

[BryanWall]

@BryanWall alright, i think i found something, building master now, could you try once the new docker container is out (about 35 minutes) or pull latest?

I did a quick test and wasn't able to break it like before. The new build didn't fix the previously broken items, but you may have known that.

I noticed that you bumped Apprise to 0.9.8.3 so I tested that also. Still getting   in the title.

[dgtlmoon]

The new build didn't fix the previously broken items, but you may have known that.

yeah sorry :( if more people report an issue, i could add some method to clean it up, but as this is unpaid work and i'm low on time its kind of tough :(

apprise - please comment over on the relative apprise issue thread