Heikki Linnakangas

On 15/01/2019 18:07, Jacob Champion wrote: > Looks like Concourse caught something: > > ``` > subprocess.CalledProcessError: Command 'psql 'funny"db'"'"'with\\quotes' -t -c " CREATE EXTENSION IF NOT EXISTS gp_replica_check "'...

The culprit is in the `simple_equality_predicate_refuted()` function: ``` /** * Check to see if the predicate is expr=constant or constant=expr. In that case, try to evaluate the clause * by...

On 15/06/2020 12:45, Richard Guo wrote: > Partition constraint: (n = '1'::numeric) > Qual in query: (length(n::text) = 3) > > It substitutes 'n' in the Qual with '1'::numeric from...

> The above case is based on postgres 12. You can see 1 and -1 are both in the same leaf partition which is just the same as the case...

This is another iteration of https://github.com/neondatabase/neon/pull/2096. I'll respond to @SomeoneToIgnore 's comments here

> Would header building be able to check things like https://github.com/neondatabase/postgres/pull/138? > > If it's not, let's do the whole build since there's no other linting for such erros. >...

> One reason to put style checks into a separate workflow was the desire to run the e2e tests despite style errors. Hmm, I see. I'm only running the formatting...

Stripping the useless .debug_pubnames and .debug_pubtypes sections from the rust binaries would help a little, see https://github.com/neondatabase/neon/pull/2514. That PR doesn't change the docker build, but we could add a similar...

The "wait for all timeline tasks to finish" logic is a bit broken with GC and compaction. We have this: ``` pub async fn delete_timeline(tenant_id: TenantId, timeline_id: TimelineId) -> anyhow::Result...

This forces image creation for a key-range, even if there have been no updates after the last image. That's not good.