Benoit Daloze

The TruffleRuby team will monitor the CI status via https://github.com/eregon/truffleruby-gem-tracker and help if it fails (like I mentioned on https://github.com/prawnpdf/prawn/pull/1189#issuecomment-761567300). I don't know why the tests are slower. It'd be...

If the CI time is a concern (it's 6min30s in [this run](https://github.com/prawnpdf/prawn/runs/5137930247?check_suite_focus=true), still sounds fairly fast): * We could try to make it faster by investigating and e.g. using `--engine.Mode=latency`...

It's expected more optimizing Ruby JITs run test suites slower (e.g., similar on JRuby), because tests constantly run different code and so rarely run any existing & JIT-compiled code. And...

From the [flamegraph](http://eregon.me/flamegraphs/prawn-flamegraph_20220214-102939.svg) it seems most of time (31.2% using `Search`) is spent in `Prawn::Images::PNG#split_image_data` and specifically in `StringIO#read` and `String#byteslice` called from there. The main cost is probably scanning...

There is notably `rb_binding_new()`, that I hear captures the binding of the Ruby frame right above it, and from that Binding object it's of course possible to write, etc, at...

> Recently there's been a change so Kernel#binding can't be called from a non-Ruby frame: https://github.com/ruby/ruby/pull/5767 Indeed, although that doesn't affect rb_binding_new(), https://gist.github.com/eregon/fc779aca9c37595ffde7473b424769c7 prints: ``` [:main] [:my_method_var] ``` (I actually...

> We might be able to get away with not invalidating until someone actually uses binding to write to a frame (most bindings only read). Is there a reliable way...

JRuby recently used libfixposix to get `pty` working in https://github.com/jruby/jruby/issues/6552, possibly relevant for us, as from reading that issue posix_spawn doesn't have that ability. But we could probably also do...

`selenium-webdriver` no longer uses `childprocess` (in upcoming releases) and the test suite passes on truffleruby :tada: https://github.com/SeleniumHQ/selenium/issues/11251#issuecomment-1314051027

The specs executed in 1min23s on both truffleruby and truffleruby-head in https://github.com/rubocop/rubocop-ast/actions/runs/3049983186/jobs/4916606764. While that's still slower than on CRuby it's pretty fast for CI in general, and around the same...