Matthias van de Meent

I updated the vendor/postgres to the latest version that addresses all comments from https://github.com/neondatabase/postgres/pull/188. Apart from that, I've registered a PR in the cloud repo that should address the change...

Note/observation: of the time spent in handle_get_rel_page_at_lsn (36.05%), * 43.5% is spent in ImageLayer::get_value_reconstruct_data * 19.1% is spent in tracing * 13.6% is spent in LayerMap::search * 10.3% is spent...

> How and where this results were produced? These are pulled from your SVG file, by selecting handle_get_rel_page_at_lsn and then searching for `^.*ImageLayer.*get_value_reconstruct_data`, `tracing`, `^.*LayerMap.*search` and `^.*DeltaLayer.*get_value_reconstruct_data` respectively; then noting...

I'm getting 8.4% tracing from the flamegraph that [you submitted here](https://github.com/neondatabase/neon/issues/2221#issuecomment-1221240909) 

> Sorry, but how 19% of of 36% can be the same as 1.4%? That's the difference between PERF flamegraphs which usually counts instructions, vs the benchmark flamegraphs which counts...

My approach so far has been a unit-testing approach, in which I assume that if we can extract all records from a file in a correct manner, then we can...

The current release branch of PG15 may still see feature reverts and/or catalog updates, so we should release PG15 to customers only after RC0 is out (otherwise we might need...

I don't think we can release PG15 beta 4 to customers, as there have been catalog updates since that was stamped. RC1 is probably going to be the first upstream...

Why close this one in favor of a closed (not: merged) PR?

This may be somewhat related to #164. The DOM update scripts, however, currently dirty all the dom elements in the library except for the container element, and also re-builds the...