I believe I've seen these exceptions for the past year or so on our setup as well, though I hadn't actually confirmed the diffs do exist. I did confirm that nobody ever manually deleted diffs though.

Thanks!

Yeah, ON DUPLICATE KEY UPDATE is sensitive to order. Setup:

I think this is:

Presuming setting Public view policy on repo resolves the issue. Let us know if not!

Just throwing in my support for this. I'd love to move away from paid services (aka Jira + Tempo) and use Phabricator.

In T8238#196204, @epriestley wrote:

I'm a little confused about how these tags are appearing so widely in developer working copies in the first place. Is git clone fetching them by default? Do you have unusual default options configured? Do users regularly explicitly fetch all tags present in the remote?

That is, my expectation is that git pull does not "bring everything back".

In T8238#196204, @epriestley wrote:

I'm a little confused about how these tags are appearing so widely in developer working copies in the first place. Is git clone fetching them by default? Do you have unusual default options configured? Do users regularly explicitly fetch all tags present in the remote?

That is, my expectation is that git pull does not "bring everything back".

(The major goal of using refs instead of tags wouldn't really be to hide them per se, just to make them extremely difficult to fetch by accident.)

Maybe my assumptions about default behaviors are off, or the more-recent base tags (but not necessarily the diff tags) are getting fetched because they're often ancestors of master?

I'm a little confused about how these tags are appearing so widely in developer working copies in the first place. Is git clone fetching them by default? Do you have unusual default options configured? Do users regularly explicitly fetch all tags present in the remote?

In T8238#196187, @epriestley wrote:

I think cleaning up closed tags probably isn't a particularly great long-term solution since some fraction of the revisions likely just get left open forever -- so you'd have 500 tags instead of 50000, but I assume that still causes about as many problems, and you'll be back to 50000 in just 100 more years anyway.

I think cleaning up closed tags probably isn't a particularly great long-term solution since some fraction of the revisions likely just get left open forever -- so you'd have 500 tags instead of 50000, but I assume that still causes about as many problems, and you'll be back to 50000 in just 100 more years anyway.

See some discussion in T7944, particularly T7944#194982.

We're planning to put application-level stemming on the MySQL index (see T6740). Are there problems with the InnoDB search that you're aware of that aren't fixable with application adjustments?

With InnoDB now mostly finished indexing, search seems to be working and the service isn't getting killed by locked tables. More info at the link I pasted above ^

We see how bad this is (does pushing tags to the same repository create a big user-facing mess?) and then go from there.

a quick and dirty fix for this could be

Wikimedia is considering another stab at elasticsearch because we just hit a fairly serious scalability issue with myisam (https://phabricator.wikimedia.org/T146673) and it looks like innodb isn't a lot better. In addition to that, the mysql fulltext engines don't seem to handle stemming or even simple prefix matching.

I clicked "Update Now" and the update frequency for our most active repository has dropped back down to 15 seconds.

+1 for mentions, that would serve the same purpose as see also: or related tasks: would. Primarily it would just be nice to be able to see them in one place instead of hiding in comments...

@beber does setting EFL to "Public" resolve your issue?s

Yes, thanks!

https://phab.enlightenment.org/diffusion/EFL/ view policy is set to "All Users".

This has been solved AFAIK since I haven't seen this issue in a while.

(I'll follow up on this once I have a chance to bang on it a bit, but I want to try to finish up the RRULE stuff while it's still somewhat fresh in my mind.)

This went out a while ago and I confirmed the fix in production.

I'm going to presume we got this, but reopen or yell if you're still seeing issues on the new code.

Thanks for the report! The detailed reproduction instructions were very helpful in resolving this.

I did order an iPhone 7 plus, so I'll have a better understanding if 2x is actually noticeably bad.

This isn't like something I'm looking for immediately, btw. Just like, next time I have to make 50 images.

I'm also not sure if we can use SVG as a data: URL in CSS. It probably should work, but if it doesn't we'll have to rasterize (and, I guess, serve 5+ different sizes to every user).

Alright -- we can have bin/celerity map rasterize SVG into arbitrary sizes if I'm not comfortable serving SVG directly from Celerity.

I'd like to generate 1 image and have them work 1x, 2x, 3x, etc or any size I need. I can't think of any other long term solution.

• Create a badge.
• Award it to a user.
• Visit their profile page.
• On the profile page, the badge title should link to the badge detail page (currently: no way to get to the badge detail page).

I'm not really sure this is actually useful given that every method is now named <application>.whatever and you can search by substring.

(That will give us slightly bad scale-downs on 1X screens, but I bet they're not worse than scaling down an SVG would be.)

I think we can just do 2x in all cases for these, if that's reasonable?

fwiw, I want to convert all bitmaps to SVG if possible, so we can skip building 1x and 2x images.

Don't know if D16178 is worth comandeering? But feel free.

This one is a little scattered, so let me know where I can fill in any gaps. Most of the details are in D16178.

Thanks!