zimbatm has quit [Read error: Connection reset by peer]
manveru[m] has quit [Write error: Connection reset by peer]
manveru[m] has joined #nixos-ruby
zimbatm has joined #nixos-ruby
jasmin has joined #nixos-ruby
<jasmin>
Hello, I can not send mail via Rails. I get the error: "535 5.7.1 Authentication failed", yet via telnet, authentication works fine. Here is my conf: https://paste.ofcode.org/XGQPuWLZaB6MmQSBkxbZPm.
<jasmin>
I do not stop searching on google for a solution, try options ...
* manveru
has a serious case of deja-vu
<manveru>
jasmin: i'm curious, since you already solved that issue like a month ago, and i told you that this isn't the right channel... did you forget everything since? :)
<manveru>
anyway, i can't help with that issue, sorry
<jasmin>
I didn't see this
<zimbatm>
manveru: can you touch a bit on the motivations for bundix v3? is that you wanted the new CLI or where there other ideas behind it?
<manveru>
zimbatm: well, it's first small step in a long process...
<zimbatm>
so it's creating the foundations for more change?
<manveru>
eventually i want to replace bundler :)
<zimbatm>
aha :)
<manveru>
well, the Gemfile/Gemfile.lock parts of it anyway
<manveru>
so generating those was the first step
<manveru>
now we need a new gemset format that can capture the needed information
<manveru>
and making it JSON seems a good thing to do as well
<zimbatm>
maybe it's an opportunity to rename it to Gemfile.nix or Gemfile.lock.nix
<manveru>
so it is a bit more inter-operable with other tools
<zimbatm>
or .json
<manveru>
well, the original plan was from qyliss
<manveru>
i just was too impatient to wait for it to happen
<zimbatm>
sounds good either case :)
<manveru>
so this is mostly a preview, not the final version
<manveru>
i also wanted the new CLI to put in documentation and subcommands
<manveru>
and get rid of magic
<jasmin>
manveru: my conf was wrong
<zimbatm>
I didn't know abou the "main" gem
<jasmin>
issue solved !!
<manveru>
cmdargs is a tiny wrapper around OptionParser
<jasmin>
yeppeee
<manveru>
i first wanted to use "main", but it's been unmaintained for years and throws warnings on recent ruby versions
<manveru>
and thor... is a monster
<manveru>
jasmin: who would've thought :)
<manveru>
jasmin: contrats
<manveru>
zimbatm: anyway, there's a lot of things to discuss and improve... while bundix2 "usually" does the "right" thing, it's always a pain to figure out what's going wrong
<manveru>
i want to get a clean separation of concerns from the ground up
<manveru>
and then we can stuff it full of features :)
<zimbatm>
manveru: main is still in the gemset config
<manveru>
hmm
<zimbatm>
+1 for the approach
<manveru>
also this can finally handle `gemspec` entries
<manveru>
for nixpkgs
<manveru>
need to add a ton more testcases, so if you know some nice weird projects, bring em on :D
<manveru>
either way, i think a lot of projects could be packaged by simply ignoring bundler... only use it to resolve dependencies once, then write a gemset.json and setup the GEM_PATH appropriately
<manveru>
that's also why i did the ruby.withPackages first
<manveru>
while it generates the gemset using bundler, the implementation doesn't use bundler at all
<zimbatm>
yeah makes sense
jasmin has quit [Quit: Leaving]
<manveru>
though now that bundler is in stdlib... gonna be hard to avoid it
<zimbatm>
really? ugh
<manveru>
since 2.6, yeah :(
<zimbatm>
great separation of concerns...
<manveru>
i said the same when they added rubygems to stdlib :P
<manveru>
but well
<zimbatm>
next, create a meta-bundler
<zimbatm>
at the moment I'm fighting with libv8 and mini_racer
<zimbatm>
it's a bit of a mess
<zimbatm>
we're injecting v8 into the libv8 gem but it's the wrong version
<zimbatm>
and if we don't do that, libv8 tries to fetch the source code at build time
<manveru>
yeah
<manveru>
but using another v8 shouldn't be hard, right?
<zimbatm>
it's not very generic, mini_racer only compiles with specific versions of v8
<manveru>
damn
<manveru>
guess i've been lucky so far
<qyliss>
zimbatm: I can send you a working mini_racer build
<qyliss>
It was so horrible I decided there was no way it was going in nixpkgs
<manveru>
that sounds more intriguing that i want to admit :D
<zimbatm>
qyliss: fire away :)
<zimbatm>
it might end-up in nixpkgs though as I'm trying to package discourse for it