#general
- e
enough-analyst-54434
08/26/2015, 11:04 PMSo it violates annotation processor APIs/norms - e
enough-analyst-54434
08/26/2015, 11:04 PMdouble checking... - e
enough-analyst-54434
08/26/2015, 11:05 PMIn some sense, pants should still just work; however, if its the only case in practice... we could certainly issue strong warnings - e
enough-analyst-54434
08/26/2015, 11:05 PM - e
enough-analyst-54434
08/26/2015, 11:05 PMits an odd duck and does "illegal" things - e
enough-analyst-54434
08/26/2015, 11:06 PMI would personally be happy with being non-robust to non-compliant apt plugins, and adding a doc note for Lombok in particular - w
witty-crayon-22786
08/26/2015, 11:08 PMwell, this case isn't lombok actually - w
witty-crayon-22786
08/26/2015, 11:08 PMit's just that classfiles seem to have runtime annotations applied in different orders - e
enough-analyst-54434
08/26/2015, 11:08 PMIn any non-Lombok-style case order should not matter by definition - w
witty-crayon-22786
08/26/2015, 11:09 PM(or it's just a bug in annotation scraping) - w
witty-crayon-22786
08/26/2015, 11:09 PMok. - w
witty-crayon-22786
08/26/2015, 11:09 PMok. would not be super surprised if sbt bug 😃 - e
enough-analyst-54434
08/26/2015, 11:09 PMIff we wanted to be robust, the only way I can see is to run multiple compile rounds - in order, each with 1 apt processor enabled only - e
enough-analyst-54434
08/26/2015, 11:11 PMUnless javac ignores explicit whitelists, that would be a path to absolute control over order, thoug the user may need to give pants a config hint on plugin order - e
enough-analyst-54434
08/26/2015, 11:12 PMThst would be one of those cool intern projects - pants could probably then claim it's the only build tool to get this "right", but its a pretty esoteric feature - e
enough-analyst-54434
08/26/2015, 11:13 PMGood on the intern resume/life experience docket, Good bullet point, never used - w
witty-crayon-22786
08/26/2015, 11:13 PMhah - e
enough-analyst-54434
08/26/2015, 11:14 PMFWIW, @bright-dinner-81701 schooled me a bit on DM and we had a good conversation on npm install which I'll be posting tonight. I think we'll be in divide and conquer mode starting tomorrow am on all the rest - e
enough-analyst-54434
08/26/2015, 11:15 PMs/npm install/resolve.npm/ - w
witty-crayon-22786
08/26/2015, 11:24 PMawesome, thank you. - w
witty-crayon-22786
08/26/2015, 11:25 PMhoping this release doesn't explode horribly, at which point i'll be able to get on that. - e
enough-analyst-54434
08/26/2015, 11:26 PMI would be surprised if I get in more than resolve.npm from tonight's RB and then test.npm tomorrow - e
enough-analyst-54434
08/26/2015, 11:27 PMEnough to have a new distributed plugin, but not too much on the changelog front - e
enough-analyst-54434
08/26/2015, 11:27 PMChris may make your life miserable though implementing all the rest - he knows node/npm in ways that I don not and may be a rocket - b
bright-dinner-81701
08/27/2015, 12:03 AMill continue where we left off in DM here - b
bright-dinner-81701
08/27/2015, 12:03 AM@enough-analyst-54434 said “It seems to me we have to allow 3rdparty to declare multiple versions of any given dep since in the end, install will be dealing with this anyway" - b
bright-dinner-81701
08/27/2015, 12:04 AMthen i was saying - b
bright-dinner-81701
08/27/2015, 12:04 AMoh, we definitely need to allow 3rd party to declare multiple versions of transitive dependencies we decided early on that 3rd party transitive dependencies would have to be opaque to pants dependency trees in node are so enormous it would be pretty much impossible to enforce consistency there what ive been struggling with is how to get even a single isolated 3rd party, with its dependencies - b
bright-dinner-81701
08/27/2015, 12:05 AMan “npm install” of an arbitrary NPM module doesnt necessarily give you all its dependencies - b
bright-dinner-81701
08/27/2015, 12:05 AMit will skip anything that has already been installed at a higher level