Seemingly inconsistent state on dashboard for some of my CLs

[WANG Xuerui]

Hi,

Recently I've noticed two of my CLs (https://go.dev/cl/455017 and
https://go.dev/cl/455075) have displayed inconsistent statuses in their
respective detail page and dashboard. Namely they seem to not satisfy
Code-Review when viewed in the dashboard, but the Code-Review
requirement appears green in the CL detail page.

While I'm posting here mainly to get some attention for the two issues
(they would show up in the needs-review dashboard otherwise, it seems),
I'm still a bit curious about the reason behind this. I just got trybot
access a while ago, so I'm afraid I've done something wrong in kicking
the bots. (I'll probably poke them instead next time, and not forget to
TRY=loong64 all the time. /facepalms)

[David Chase]

I'll give them a look. To be clear, is the problem is that those two
CLs don't show up in https://go.dev/s/needs-review , but they lack two
Google touches and thus they should. (One of them won't show up now
because I +1'd it and it moved to ready-to-submit, which means that it
*should* have been on that list, but was not.)

Improperly kicking trybots is a thing, it does not help that last time
I checked, there is no check that the requested trybot even exists and
it would silently do nothing. (Guess how I know.)

> --
> You received this message because you are subscribed to the Google Groups "golang-dev" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to golang-dev+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/golang-dev/56ad2967-5afd-e131-7fa0-93ce8c3d7e9a%40xen0n.name.

[David Chase]

There's a claim that there's some internal bug that makes the
dashboard use stale data sometimes, not sure it is worth further
investigation with your CLs as test rats but it's odd that you had two
of these.

On Mon, Apr 10, 2023 at 1:12 PM WANG Xuerui <i.go...@xen0n.name> wrote:

>
> On 4/10/23 23:22, David Chase wrote:
> > I'll give them a look. To be clear, is the problem is that those two
> > CLs don't show up in https://go.dev/s/needs-review , but they lack two
> > Google touches and thus they should. (One of them won't show up now
> > because I +1'd it and it moved to ready-to-submit, which means that it
> > *should* have been on that list, but was not.)
>

> Thank you very much. Yes the problem was that they didn't appear in that
> dashboard while they should. (When I tried to remove filtering
> conditions from the dashboard's template to "debug", it appeared
> removing the CR+2 condition made the CLs show up, yet they definitely
> hadn't lost their CR+2 votes at the time.)

[WANG Xuerui]

On 4/11/23 01:57, David Chase wrote:
> There's a claim that there's some internal bug that makes the
> dashboard use stale data sometimes, not sure it is worth further
> investigation with your CLs as test rats but it's odd that you had two
> of these.

Ah, that kinds of explains the thing. I remember after ki^H^Hprodding
the bots in the (UTC+8) evening and getting TryBot-Result +1, I briefly
saw the 2 CLs appear in the dashboard, but by midnight they disappeared
again.

Since the bug is certainly non-deterministic, it may well happen again
for someone else, and perhaps not limited to the Go instance of Gerrit.
I think it's probably fine to not look into my CLs, they'll perhaps
encounter the bug later nevertheless; thanks again for your help looking
into this and sharing the info!

[WANG Xuerui]

On 4/10/23 23:22, David Chase wrote:

> I'll give them a look. To be clear, is the problem is that those two
> CLs don't show up in https://go.dev/s/needs-review , but they lack two
> Google touches and thus they should. (One of them won't show up now
> because I +1'd it and it moved to ready-to-submit, which means that it
> *should* have been on that list, but was not.)

Thank you very much. Yes the problem was that they didn't appear in that
dashboard while they should. (When I tried to remove filtering
conditions from the dashboard's template to "debug", it appeared
removing the CR+2 condition made the CLs show up, yet they definitely
hadn't lost their CR+2 votes at the time.)