Benoit Daloze
Benoit Daloze
Psych 5 no longer bundles libyaml, so installing psych 5 will use the system libyaml. Other than that, we can either wait that TruffleRuby updates to Ruby 3.2 or make...
We discussed using the system libyaml on Slack. The ABI of libyaml seems quite stable since `soname` `0.2`, even more than zlib actually: https://abi-laboratory.pro/index.php?view=timeline&l=zlib https://abi-laboratory.pro/index.php?view=timeline&l=openssl https://abi-laboratory.pro/index.php?view=timeline&l=yaml So we might not...
With https://github.com/oracle/truffleruby/pull/2899 TruffleRuby now uses the system libyaml, like CRuby 3.2+ and Psych 5+. That improves warmup of `YAML.parse` quite a bit as showed above. We need to look at...
> If I [modify](https://github.com/Shopify/bootsnap/pull/447) bootsnap to enable the YAML compile cache it speeds `Faker::Config.locale` for me by 50%. So that caches the YAML with MessagePack? That's a C extension that...
@djoooooe Any idea what we could improve regarding https://github.com/oracle/truffleruby/issues/2089#issuecomment-1613936967 ?
I have taken another look at this. To repro, clone prawn, `bundle` and then `bundle exec rspec spec/prawn/images/png_spec.rb`. To profile: `TRUFFLERUBYOPT=--cpusampler=flamegraph bundle exec rspec spec/prawn/images/png_spec.rb` The whole test suite takes...
Here is a list of default gems with extensions in 3.1: ``` $ grep -a -R extensions lib/gems/specifications/default | sort lib/gems/specifications/default/bigdecimal-3.1.1.gemspec: s.extensions = ["ext/bigdecimal/extconf.rb".freeze] lib/gems/specifications/default/cgi-0.3.5.gemspec: s.extensions = ["ext/cgi/escape/extconf.rb".freeze] lib/gems/specifications/default/date-3.2.2.gemspec: s.extensions...
We tried this but using TruffleFile causes a significant performance overhead and it's only partial anyway, it seems too difficult to use TruffleFile for regular Ruby IO/File objects. So closing...
When we implement this we should reenable rdoc CI: https://github.com/ruby/rdoc/pull/1029
Another idea, because it's quite complicated to store ReturnID for methods & lambda in FrameDescriptor without a lot of changes or performance impact: * We could handle `return` in `BEGIN...