You would have to explicitly set complete_doas arrays. It might be doable based off current complete_* variables listed via set, but It would take some doing.
Yeah that doesn’t seem accurate because .po is for localization I think, and it has a lot of tests. You don’t want to count the localization tests against it. But yes bash is too big and slow – it even says so in the man page.
And nice article! I’ve worked on completion a bit for the OIl shell, and it’s definitely a big pain point with bash. In fact I have some broken custom completions right now that I can’t figure out, and I’m pretty good at debugging bash. It’s actually a non-deterministic bug where they stop working after awhile in the shell and I have to restart a new shell to make them work.
I had no idea! That’s awesome! I debated not including all the things.. ultimately decided give complete totals because picking individual languages seemed too arbitrary.
And nice article!
Thanks! That sounds like a bug one definitely wouldn’t want in their shell!
I also did not know about this!
Is there a way to make this work for doas, so it picks up the autocompletes for the command?
So
$ doas rcctl s<tab>would use rcctl’s autocompletes?You would have to explicitly set
complete_doasarrays. It might be doable based off currentcomplete_*variables listed viaset, but It would take some doing.Is there much benefit from pledging the shell? I thought pledge wasn’t inherited across fork/exec.
This question has of course been asked before :)
http://marc.info/?l=openbsd-tech&m=144451217831726&w=2
[Comment removed by author]
I included everything in the source directory, here is a breakdown:
Yeah that doesn’t seem accurate because .po is for localization I think, and it has a lot of tests. You don’t want to count the localization tests against it. But yes bash is too big and slow – it even says so in the man page.
And nice article! I’ve worked on completion a bit for the OIl shell, and it’s definitely a big pain point with bash. In fact I have some broken custom completions right now that I can’t figure out, and I’m pretty good at debugging bash. It’s actually a non-deterministic bug where they stop working after awhile in the shell and I have to restart a new shell to make them work.
I had no idea! That’s awesome! I debated not including all the things.. ultimately decided give complete totals because picking individual languages seemed too arbitrary.
Thanks! That sounds like a bug one definitely wouldn’t want in their shell!
There’s still no fish port, is there? I tried pkg add a little while ago and got nothing, but maybe I used the wrong name?
It’s in wip - I am going to stick with ksh, but I still plan on sending the patches upstream.
fish-shell