MSP-Greg

Confused. `@options[:workers] > 1` is false for `-w1`. This was the change I suggested, along with removing the `else` clause. The change does not execute `sleep` for `-w1`. `@clustered` is...

@dentarg I'm confused. I added the patch as above. ``` echo 'app { [200, {}, ["OK"]] }' | RUBY_MN_THREADS=1 bundle exec puma --config /dev/stdin -w1 -t1:1 --port 7777 hey -n...

> So: sleep seems related. Speaking of sleep, I just shut down all Puma work. > showed significant time spent in Mutex#synchronize Another rabbit hole I've peered into is replacing...

I just ran this with Ruby before and after @amuta's commit. That's a serious increase in performance, not expected. Also, see https://bugs.ruby-lang.org/issues/21560#note-7, which also showed an increase. Note the `+MN`...

> That is very unexpected. Agreed. I just posted in ruby about whether the fix will be backported to Ruby 3.3 & 3.4. 3.4 is listed as 'required', 3.3 isn't,...

> My workflow has setup-python step followed by setup-ruby. If you reverse the steps, does it work? The MSYS2 path entries are needed for compiling extension gems. I don't believe...

@eregon @hsbt IIRC, the ruby-loco code for ucrt & mingw Ruby is identical except for the compiler setup. Or, they should behave identically. mswin is a bit different, but should...

@eregon > I plan to fix this soon in TruffleRuby Thanks for the info. I don't recall if I pinged you, but previously Puma had a several of the non...

See https://github.com/puma/puma/pull/3488#issuecomment-2357055924 Some quick testing shows that it behaves essentially the same with keep-alive connections and non. Re closing connections, with Puma's current code structure, it isn't necessary to close...

> If the algorithm can be changed to serve all requests in order and reusing any keep-alive connections I think several people are looking into this. I also am but...