Fix auto-install of pnpm and node on mode init - #16

I suspect this doesn't work because my shell env first prints my lovely terminal art before running anything else

That IS lovely! Interesting, I will check how I can distinguist command output and general shell output.

(Ah yes, I remember now, I am using a full-on interactive shell so we can have all the env variables, but that of course also triggers any user-defined shell startup code in .bashrc or equivalent.)

@stephanmax this is working for me on gentoo/zsh/wayland /cc @lee2sman

how can i test? do i need to build from source? i don't have a kotlin or java environment set up for example.

let us know what you're thinking for handling running shell commands @stephanmax

@lee2sman ooops nevermind. ill let you know when this is merged. i forgot you're not building

I think the easiest workaround is only considering the last line of the command output. This is not as rigorous a solution as I’d like, but my other ideas fall short as well.

• We could redirect the command output to std error and read that stream. User-defined output would still go to std out and we could dinstinguish them.
• We could just consider the error code. But I like the debugging information that pnpm and Node version give us.
• Come up with a more clever test for a certain binary (without relying on an interactive shell)?!

@stephanmax i like the idea of redirecting the the output to another stream. i suppose you could use stderr - though i do wonder how easy it would be for to output it to a new stream so that it doesn't pollute stderr?

As for your second idea - i love error codes. is there another way for getting the debug information without having to spit it all out to the console? Or a flag?

I think your questions are valid and all culminate in my third bullet point :)

Happy to invest more time on this, but I’d say for now just considering the last line of output works well enough.

is there a way to test now or should i wait til there' a new release?

@lee2sman it's merged in, so you can go ahead and test the way you planned to in the beginning

It's included in the new 0.6.0 release ☺️