hedning[m] has joined #nixos-borg
jtojnar has joined #nixos-borg
MichaelRaskin has joined #nixos-borg
<ekleog> wow, I guess something's really wrong: running `nix-build default.nix with http://ekleog.xelpaste.net/P0LWSu as default.nix gives the same error: http://ekleog.xelpaste.net/vuL1EA . Might it be due to the recent switch to nix 2? I can't see why, but…
<ekleog> that said when replacing the `(import <nix/config.nix>).shell` with `"/bin/sh"`, things work correctly
<ekleog> hmm, what's this <nix/config.nix> BTW? I don't appear to have a nix= in my NIX_PATH, can't seem to find any docs about it, and never saw it before
<jtojnar> how can I run grahamcofborg-eval locally?
<gchristensen> I wish you could :(
<gchristensen> you can emulate it
<jtojnar> gchristensen: I tried that but it works, unlike master
<gchristensen> what's going on?
<gchristensen> ohh master is broken
<gchristensen> jtojnar: which PR?
<gchristensen> master is broken
<gchristensen> and master must be fixed firsst
jtojnar has quit [Ping timeout: 246 seconds]
jtojnar has joined #nixos-borg
<gchristensen> MichaelRaskin: I can't wait to have auto-merge so we can disable pushing :(
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-borg
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-borg
jtojnar has quit [Quit: jtojnar]
MichaelRaskin has quit [Quit: MichaelRaskin]
<makefu> i think the move towards push by dr.borg is a fantastic development. if this feature will become the new standard i definitly will consider becoming an `official` maintainer. it would be even better if i could earn maintainership only for the small fraction of packages i could really help with.
orivej has quit [Ping timeout: 260 seconds]
<gchristensen> makefu: =)
<globin> gchristensen: btw re nixborg account shall I send you the creds?
<gchristensen> sure!
<gchristensen> I won't be able to switch over for a few days, but sure let's get that startted
mrphssh has joined #nixos-borg
mrphssh has quit [Client Quit]
<ekleog> gchristensen: Here is the rewrite of the comment parser to use nom and add the nixos-test alias :) Also, I don't really know what you meant by parsing the comment to an AST and then reducing from there, but it should be much easier to do now that a proper parser library is in use :)
<ekleog> (it doesn't break any test on my environment that weren't broken on next before the PR)
<gchristensen> ekleog: it was a bit of a joke, Mic92 and I have been talking about converting the trivial commands in to a proper AST and parser and having a full blown language
<globin> gchristensen: will send a mail
<gchristensen> ok, great!
<Mic92> gchristensen: this escalated quickly
<Mic92> whois borg-lang.org looks still free
<Mic92> When will borg become an Alexa skill?
<Mic92> ekleog: is it feasible to parse something as complex as nix with nom?
<ekleog> Mic92: well, rustc's parser is based on syn, which is a stripped-down version of nom, so… :°
<ekleog> s/syn/synom/ (syn is actually the name of the synom parser, and synom the stripped-down version of nom, iirc)
<Mic92> ekleog: I like the error messages that hand-written parser can be achieved over bison parser. Does nom gives me enough flexibilty to create reasonable error messages?
<gchristensen> Mic92: should probably be ofborg-lang.org, no?
<gchristensen> or locutus-lang
<Mic92> is the fun worth? £9.30/year ?
<gchristensen> I've spent more on less ...
<ekleog> at least theoretically it can, https://github.com/Geal/nom/blob/master/doc/error_management.md , though I haven't started to experiment with it
<globin> gchristensen: https://nlu.rasa.ai/ %)
<gchristensen> my god
<Mic92> ekleog: the first sentence already looks discouraging.
<ekleog> indeed, but given how rustc does give really nice errors, I guess it's not that bad :°
<globin> gchristensen: you shoild have mail
<gchristensen> globin: got it!
<gchristensen> thank you :D
<Mic92> depends how much time they spent on it.
<Mic92> Maybe nix is grammar is not too complex to parse it by hand?
<gchristensen> Mic92: or hughscript (https://en.wikipedia.org/wiki/I,_Borg)
<ekleog> true indeed, that said if it were me I'd rather start with nom, get something that works, then worry about adding nice error reporting :°
<Mic92> gchristensen: I was not socialised with Star Trek.
<gchristensen> it is okay
<gchristensen> what sounds like the best name?
<Mic92> ekleog: I thought of having something else to parse nix code then the C++ code base for formatting - then error messages becomes important when it is integrated into editor.
<Mic92> But maybe runtime errors are the ugly part of the nix language
<Mic92> and not the syntax errors.
<ekleog> I've never felt like syntax errors were a big issue, usually nix points pretty well where they are
<Mic92> probably
<Mic92> Stack traces are bigger issue
<Mic92> and anonymous closures
<Mic92> and infinite recursion
<ekleog> indeed, but nix being lazy could they be made much better? I feel like a good typing support would solve a big part of these issues before they even arise, but…
<Mic92> they would. But I don't see this to happen any time soon
jtojnar has joined #nixos-borg
cransom has quit [*.net *.split]
cransom has joined #nixos-borg
<zimbatm> nix already attaches a lot of information with each closure and strings that are being generated
<zimbatm> I'm pretty sure it should be possible to expose more that
orivej has joined #nixos-borg
MichaelRaskin has joined #nixos-borg
<MichaelRaskin> (sigh) do you consider the natural Plan B for tests a good idea?
<gchristensen> what is that?
<MichaelRaskin> Separate repo — maybe just in my account with access to everyone reasonable who asks, tests approximately track master and may also have .release-1803 versions, ofBorg support for running these
<gchristensen> ouch, why?
<MichaelRaskin> Ah, you don't read notifications even for the threads where you previously participated.
<gchristensen> I just haven't been reading much mail today
<MichaelRaskin> Note that all his point (well, except the idea that Perl is a good language for readable code) are addressed both in my email to ML and in the issue description at the moment he commented…
<MichaelRaskin> By the way, does GitHub API for checks require that every check is either a success of complete failure with no third state?
<gchristensen> there is an Error state
<gchristensen> ooohh
<gchristensen> that'd be good for the master branch being broken!
<MichaelRaskin> I am just defending ofBorg comment spam as not-obviously-insane in an issue, so I am wondering if builds can be shoehorned into the checks if they just succeed-or-undecisive
orivej has quit [Ping timeout: 246 seconds]
<samueldr> hmmmm, have I missed a notification about the github-to-irc bridge? looks like it hasn't notified in a while
<samueldr> well, since wednesday 15:38 EST
infinisil has quit [*.net *.split]
infinisil has joined #nixos-borg
<MichaelRaskin> Hm. @matthewbauer mentions that Error state doesn't do anything useful, as the overall summary state is still failure. Argh, does GitHub do anything right?
<samueldr> marketing
<samueldr> (as in, they took the "git" market early on)