#context directory blockers
west stone
urgh, this really is just the tip of the iceberg - more of a writeup here: https://github.com/dagger/dagger/pull/7744#discussion_r1670959469
context directory blockers
there's some entirely non-trivial issues in here, tbf, most of which are a result of some very tricky module loading logic.
@livid bobcat do you have any major objections if i look at reworking some of this pretty (maybe drastically)? i'm really struggling to understand how absolute/relative paths are supposed to interact, and it seems like there's some fairly big issues where we get confused between local modules loaded from the host and from the local
fyi @modern pebble
glass cave
Just to reiterate what I told @west stone verbally: there is no pressure to merge this to fit a particular release timeline, or version number. It's ready when it's ready.
west stone
oh also - i'm also happy to take over some of the more subtle module loading issues here, i think some of them are of my own doing with the views changes 😄
livid bobcat
there's some entirely non-trivial issues in here, tbf, most of which are a resul...
Commented on the issue: https://github.com/dagger/dagger/pull/7744#discussion_r1671111212
tl;dr There might be "simple" fixes to these issues in terms of "diff size", but we should nonetheless finish centralizing all the loading logic server-side to reduce the remaining complication (believe it or not this used to be much worse).
And I specifically think we're at the point where we should straight-up stub out at minimum dagger init/install/develop/call so that the CLI just makes a single core API for each one and the engine does all the work.
west stone
I'm on board with that 🙂
modern pebble
Thanks for the notice, that's funny because all the tests were good 😦
west stone
Commented on the issue: https://github.com/dagger/dagger/pull/7744#discussion_r1...
Thanks thanks, I'll see if I can find time to pick that up at some point!
I'd forgotten what state we left all of it in, I know you've done so much work in the area ❤️
livid bobcat
I'm uncertain how much of this needs to be a blocker on the context directory. It might make sense to patch bugs and then do the larger effort of finishing centralization separately. Mainly because IME this whole area takes forever to get every detail aligned.
paths are the equivalent of that for this whole area ^
west stone
Thanks for the notice, that's funny because all the tests were good 😦
Honestly, the only reason I caught this was because I was trying to add a test for remote git modules that use context dirs, and ended up doing something a bit weird to hit this
west stone
I'm uncertain how much of this needs to be a blocker on the context directory. I...
Yeah fair enough - I think the everything is absolute approach might fix us for now
I also really like the idea of having a ScratchSource or something, since currently, we're very confused on where module sources come from
Though again, maybe not worth blocking this on
I think me and @modern pebble are a little stretched for time before the release, so I still think it's worth pushing it into v0.12.1 for now 😭
livid bobcat
I think me and <@281874480651829250> are a little stretched for time before the ...
Makes sense to me, not worth rushing to that extent. Obviously don't need utter perfection to release something but this seems like there's enough left to do to just wait another patch release.
glass cave
Does any of this put into question the design itself?
west stone
Nah, this just feels like tech debt starting to catch up to us 🙂
Things have been "broken" for a bit, this is just the first thing to do where we would have noticed
glass cave
Just checking in, is anyone actively working on this? I worry about bitrot