Thread: Commitfest app reports wrong patch as latest
I made a CF entry [1] for a patch I'm planning to commit fairly soon, just in hopes of getting the cfbot to verify it. I was surprised to see that the CF app reported an old patch in the same thread as being latest: Key management with tests First at 2021-01-01 04:50:47 by Bruce Momjian <bruce at momjian.us> Latest at 2022-04-21 17:53:20 by Tom Lane <tgl at sss.pgh.pa.us> Latest attachment (v13-0001-pg_upgrade-perserve-database-OID-patch.patch) at 2022-01-22 07:20:30 from Shruthi Gowda <gowdashruat gmail.com> The cfbot seems to have tested the right patch, ie the one at [2], which surprised me even more because I thought it relied on the CF app's opinion of what's latest. Not sure what's going on there, but the thread in question is very long --- maybe the CF app just gets confused at some point? regards, tom lane [1] https://commitfest.postgresql.org/38/3624/ [2] https://www.postgresql.org/message-id/3353761.1650563600@sss.pgh.pa.us
On Thu, Apr 21, 2022 at 12:12 PM Tom Lane <tgl@sss.pgh.pa.us> wrote: > Not sure what's going on there, but the thread in question is very > long --- maybe the CF app just gets confused at some point? Could it just be a cache invalidation issue? -- Peter Geoghegan
Peter Geoghegan <pg@bowt.ie> writes: > On Thu, Apr 21, 2022 at 12:12 PM Tom Lane <tgl@sss.pgh.pa.us> wrote: >> Not sure what's going on there, but the thread in question is very >> long --- maybe the CF app just gets confused at some point? > Could it just be a cache invalidation issue? Maybe so, because I just went back to the CF entry's page to close it, and now it shows the right thing. regards, tom lane
On Thu, Apr 21, 2022 at 1:34 PM Tom Lane <tgl@sss.pgh.pa.us> wrote: > Maybe so, because I just went back to the CF entry's page to > close it, and now it shows the right thing. I've noticed that http caching (I believe we use Varnish for this) is often behind problems that I run into with our web based infrastructure. Not that I have problems all that often, mind you. -- Peter Geoghegan
Peter Geoghegan <pg@bowt.ie> writes: > On Thu, Apr 21, 2022 at 1:34 PM Tom Lane <tgl@sss.pgh.pa.us> wrote: >> Maybe so, because I just went back to the CF entry's page to >> close it, and now it shows the right thing. > I've noticed that http caching (I believe we use Varnish for this) is > often behind problems that I run into with our web based > infrastructure. Not that I have problems all that often, mind you. I don't think it was an http-level issue, because the page showed the wrong thing from inception, and reloading it didn't help. I am guessing it was built from a stale view of what the end-of-thread was. Anyway, the problem did go away eventually, so it was a cache problem somewhere. regards, tom lane
On Thu, Apr 21, 2022, 23:17 Tom Lane <tgl@sss.pgh.pa.us> wrote:
Peter Geoghegan <pg@bowt.ie> writes:
> On Thu, Apr 21, 2022 at 1:34 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Maybe so, because I just went back to the CF entry's page to
>> close it, and now it shows the right thing.
> I've noticed that http caching (I believe we use Varnish for this) is
> often behind problems that I run into with our web based
> infrastructure. Not that I have problems all that often, mind you.
I don't think it was an http-level issue, because the page showed the
wrong thing from inception, and reloading it didn't help. I am guessing
it was built from a stale view of what the end-of-thread was. Anyway, the
problem did go away eventually, so it was a cache problem somewhere.
There is no cache for the commitfest app.
There is however a cronjob that syncs info with the archives. And it's not very forgiving to temporary network problems - if those happen it can take some time before it manages to process the full queue of "pending attachments".
I think the cf bot uses the cf web to find the threads, but then independently looks them up on the archives to get the attachments.
/Magnus