createdDate and startOfWeek behavior changed

[Bruno Brzozowski]

Hi all,

It seems like the way createdDate and startOfWeek or startOfDay interact with each other, has changed in the last couple of weeks.
I have 3 filters which show me when a ticket is unassigned for more than 10 minutes (during the week and during business hours) or unassigned for more than 30 minutes (during evening hours or the weekend). These filters were working fine until about 2 weeks ago or so. Something must have changed because I get results now when I shouldn't.

The filters are something like this:

project = Fixlist AND resolution = Unresolved AND assignee in (EMPTY, Operations) AND ((created >= startOfWeek(-1d) AND created < startOfWeek("+1d") AND (created < -30m or updated < -30m)) OR ( created < startOfWeek("+1d")))

project = Fixlist AND resolution = Unresolved AND assignee in (EMPTY, Operations) AND ( (created >= startOfDay("+17h") AND created <= startofDaY("+24h") AND created < -30m) OR (created >= startOfDay() AND created <= startofDaY("+1h") and created < -30m) OR ( createdDate < startOfWeek("+1d")))

project = Fixlist AND resolution = Unresolved AND assignee in (EMPTY, Operations) AND ((created >= startOfWeek("+1d") AND created <= startOfWeek("+5d") AND created < -10m AND created >= startOfDay("+1h") AND created <= startofDaY("+17h")) or ( created < startOfWeek("+1d")))

Has any of this changed lately? Am I using these incorrectly?
Thank you in advance.

[Bill Sheboy]

Hi @Bruno Brzozowski 

If you just noticed this behavior start less than a month ago, and have not changed your locale settings or the query, I recommend working with your Site Admin to contact the Atlassian Support team: https://support.atlassian.com/contact/#/

I see no open defects for this symptom in the public backlog, and I recall they are continuing to "improve" JQL handling with the changes to the different search views.  Prod Support will be the best source to answer you questions about any side-effects from recent releases.

Kind regards,
Bill

[Jim Knepley - ReleaseTEAM]

Hi @Bruno Brzozowski 

Can you give us an example of the expected behavior and the actual behavior? I don't see anything wrong in your queries, some examples might help reveal the issue.

[Bruno Brzozowski]

Absolutely. Thank you @Jim Knepley - ReleaseTEAM for replying.

So just to explain the expected behavior, I have 3 filters:
1. Tickets unassigned for more than 10 minutes, between the hours of 1am to 5pm, Monday to Friday.
2. Tickets unassigned for more than 30 minutes, between the hours of 5pm to 1am, Monday to Friday.
3. Tickets unassigned for more than 30 minutes, on Saturday and Sunday.
I have a subscription which runs every 5 minutes and emails me if there are results.

One example I have is a ticket was created at 3:42am Sept 30th (Monday).
At 3:45am, I received an alert from each of the 3 filters, which is different behavior that I was experiencing before.
At that time, the ticket was 3 minutes old, but each of the filters emailed me. If the ticket stayed unassigned, I would have expected the first email to come out at 3:55am from filter #1 because at that point the ticket would be more than 10 minutes old and unassigned. The other two filters would not have emailed at all. Which is how things had been working.
It looks like Sep 23rd is when it first started happening.

[Jim Knepley - ReleaseTEAM]

I can't speak to any changes in Jira cloud that could explain the different behavior,

From the "eh, it's worth a try" file, you might try changing how you are offsetting the startOfWeek() function, so try startOfWeek() - 1d instead of startOfWeek("-1d").

[Bruno Brzozowski]

Thank you. I'll try that.