Design Meeting 2021-09-02

[kitsonk]

• Review of open items on the Q3 Roadmap Q3 2021 roadmap #11168
  • Subprocess API is making good progress, but blocked on Deno.kill() stabilisation (discussed later)
  • Testing overhaul to be discussed later
  • Good progress on docs, with a lot of the infrastructure delivered, but it will be Q4 before a refreshed doc.deno.land is done.
  • Config file making progress.
  • No progress on code signing.
  • No progress on prompt by default.
  • No type check by default has a PR, but needs to be discussed (discussed later)
  • No progress on cross-process broadcast channel
  • Crypto is making good progress. Likely more changes/features for 1.14
  • LSP v2 registry, import map improvements not likely for 1.14
• Testing proposal update
  • Going to implement the imperative API (with step metadata) first
  • An API for a declarative approach will be built on top of the imperative API. It will be then determined if it gets baked into CLI, lives in std or lives as a community API.
  • A POC polyfill for mocha has been done by @lucacasonato though it needs the implementation of the imperative API to really prove it, but the concept is sound.
  • Action open an issue for next steps (Implement test steps API #11901)
• Microtask queue issues (Microtask queue is not handled correctly in Deno #11731)
  • A breakout meeting indicated that it isn't easily addressable.
  • Impact is several WPT tests are failing, which is a strong indication that there are edge cases in some of the APIs that people are encountering, though it is really hard to pin them down to repeatable re-producable issues because of the nature of the problem.
  • Action further investigations by the team to see if there is something that can be done.
• Fetch file:// URLs via fetch() fetch() a local file #2150
  • There is an internal proposal to resolve the issue, being an "MVP" which doesn't support things like HEAD requests, range requests, relative URLs or content-type.
  • Not incompatible with the spec, but defines undefined behaviour, which is a risk.
  • Broad team support.
  • Action @kitsonk to research behaviour in Firefox of local file handling and raise an issue for implementation
• Let move deno_webgpu out of CLI's tree (Run WebGPU CTS on CI gfx-rs/wgpu#1421)
  • The Rust wpgu implementation which we use in Deno would like our JavaScript bindings to be moved over to their org, so they can run the spec conformance tests and iterate quicker.
  • This is a win/win for everyone, as we haven't had the time to maintain the bindings and there are long cycles in updating, as well as it is good collaboration between projects.
  • There are some implementation issues to work out so we can re-consume back into Deno, but it feels worth the effort.
  • Action let's do it!
• Stabilize Deno.kill() API Stabilize Deno.kill API #11757
  • Agreed it should be stabilised, but we need to get rid of the clunky LinuxSignal and MacOSSignal enums first.
  • Action open an issue (Move away from using enums for Deno.Signal API #11900) and implement
• Config file discovery
  • @ry presented the outcome of a breakout meeting regarding config file auto-discovery.
  • Of the 4 logical potential options, what is termed as target' was recommend, that the target of the subcommand will be used as a base for searching for a Deno.json, ascending parent directories to the root.
  • There will be some challenges to work out with implementing for the LSP, but should be achievable.
  • Action @ry will open an issue for implementation, target 1.15
• Markdown formatting - wrapping links like prettier (proseWrap: always should not wrap link title dprint/dprint-plugin-markdown#53)
  • Everyone agrees that no-wrap is good and aligned with prettier
  • Action @dsherret to implement
• No type checking by default ([DRAFT] feat: skip type checking by default #11810)
  • Open questions with the PR about deno test not type checking being surprising, deno cache behaviour changes significantly and the potential introduction of deno check.
  • There is a warning, that can't be disabled, that is really annoying.
  • @lucacasonato wonders if we should just wait for 2.0
  • Action punt to 2.0
• What about 2.0?
  • The no type checking prompted a discussion about 2.0.
  • Consensus that we should explore what 2.0 would look like, but that we would want to do it "quickly", like after 1.15.
  • Action explore what 2.0 scope would look like at next design meeting.