@joelmoss@aarondfrancis I think most of these tools support connecting to an external IDE? Shouldn't need to be "wrapped" by the IDE, as long as both processes are running they can communicate
@timacdonald87 I think the JIT sort of does this, but probably not with helpers like env() because that's a userland helper. Maybe $_ENV could be different. Either way, dead code, including branch, elimination doesn't do much in PHP since the perf bottleneck is typically IO
@brendt_gd Specifically disable it because to my knowledge there's pretty much no performance benefit in regular apps that don't do any heavy math or similar, and the implementation has some issues (if you look at the php-src repo you can see a bunch of JIT segfault issue threads)
If you have serious projects on GitHub, it's a good idea to set up mirrors of your repos on other forges. Very easy to do with Gitea/Forgejo.
Self-hosting Forgejo is also super easy, if you have a NixOS server you can just add it as a service and put it behind nginx like this
@brendt_gd I don't like locking things like PHPStan since it consistently gets better - usually manifests as failing CI due to a new check or smarter logic that makes some ignores redundant. Typically a quick fix and I know I'm getting the most out of the tool
@gonedark Speaking of wifi, one underappreciated aspect of latency is how bad (and variable) wifi can be. If you ping your own router, depending on how congested the channel is (+ other interference), you could be seeing some unexpected latency spikes
@ctjlewis@tszzl@mpopv with what the gold card site says, it seems they're going to try to make it part of EB1/EB2 and use the large fee as evidence of extraordinary ability, basically commercial success. I think at least some of those subcategories have clearly defined criteria directly in INA though
@mholt6@jaydrogers Kind of, but benchmarking without any I/O is somewhat pointless if you're trying to compare how two production-like deployments perform. I think Jay's idea here is to test just that, rather than only benchmarking the web *servers* themselves.
@jaydrogers Should definitely involve DB queries (possibly complex ones) to replicate a real-world app
With both FrankenPHP and PHP-FPM configured for maximum possible throughput given the hardware. Noticed a lot of these benchmarks don't configure PHP-FPM very well so def a focus point
@JackEllis I think book pics tend to get bookmarked and the algo seems to really love bookmarks, a couple months ago a good % of my feed was just pics of totally random books no one will ever read that are just good at attracting bookmarks https://t.co/YMTPDEP1RG