feat(cli) - Runtime binary for Deno

[yos1p]

deno_runtime crate will produce lite binary: deno-rt in target folder (#8757)

CC @bartlomieju @lucacasonato

[lucacasonato]

This looks good, but could you split the PR so this PR only adds the deno-rt binary? Then we can change how deno compile works in a follow up PR.

[yos1p]

Okay. I reverted some of the changes in Deno compile. I also move some of the code from cli/standalone.rs to runtime/standalone.rs. Running from CLI should will call run function from runtime crate now.

[bartlomieju]

@yos1p looks good, nice work!

That said I'm not sure if we should add a binary target to deno_runtime, maybe instead we should add a second binary target to cli? (I know I suggested to put it in the runtime dir in the original issue, but I'm having doubts now). With this change concepts of standalone binaries are split into cli and runtime and I believe it's preferable to have all of that code in cli.

@ry WDYT?

[bartlomieju]

Also we need to track the size of the "lite" binary in the benchmarks: https://deno.land/benchmarks#executable-size I'll be fine to add that in a separate PR though.

[ry]

Looks good - but can we bike shed on the name of executable a bit?

Generally we avoid dashes in filenames. "rt" is not very intelligible.

What about denolite ?

[bartlomieju]

Looks good - but can we bike shed on the name of executable a bit?

Generally we avoid dashes in filenames. "rt" is not very intelligible.

What about denolite ?

Good point, +1 for denolite

[lucacasonato]

denolite sounds like a lite version of Deno for general use. But it is not. It is purely the runtime for standalone executable. I would prefer denort or denoruntime.

[yos1p]

Thank you for review! Okay. I'll try to move the code and binary to be generated by cli crate.

Also we need to track the size of the "lite" binary in the benchmarks: https://deno.land/benchmarks#executable-size I'll be fine to add that in a separate PR though.

I can do it in this PR.

For the naming, I kind of prefer denort instead of denolite, but name can be easily changed later. I'll just follow what you guys decide.

[ry]

Generally looks good @yos1p.

When I run denort in without an arguments I get a zero exit code (success) and no output. I think this is going to be confusing to people who don't know about this. It would be better to print an error message and have a non-zero exit code.

[ry]

I think we need to have a test of some sort... How will we know if this thing breaks?

[yos1p]

I think we need to have a test of some sort... How will we know if this thing breaks?

I think we can use similar tests like standalone, but this can only be applied when we already implement compile with denort instead of deno.

[yos1p]

@ry Any suggestions for the error message?

[bartlomieju]

@yos1p this is fine for now, I will review this PR in more detail today/tomorrow and merge it. @lucacasonato and I will look into using this new binary target for standalone executables.

[ry]

"This executable is used internally by 'deno compile', it is not meant to be invoked directly."

[bartlomieju]

@yos1p thank you for the PR, I really appreciate it, but @lucacasonato and I took a deeper look at it and we found out we needed to do a lot more changes. We've opened a new PR at #9041. Closing without merge.

[yos1p]

No worries. Looking forward for the new PR!