#general
- b
bright-dinner-81701
10/09/2015, 6:21 PMso how did you make it work with source dependencies for go? - b
bright-dinner-81701
10/09/2015, 6:21 PMthe paths are different - b
bright-dinner-81701
10/09/2015, 6:21 PMin the flattened structure under .pants.d - e
enough-analyst-54434
10/09/2015, 6:22 PMMaybe rewind and state the problem, I'm a bit lost Go just makes symlink farms under .pants.d for local and or remote, all mixed together - b
bright-dinner-81701
10/09/2015, 6:22 PMuh hmmm ok - e
enough-analyst-54434
10/09/2015, 6:22 PMThe code is all in 2 classes if you want to read it - e
enough-analyst-54434
10/09/2015, 6:22 PMGoFetch GoWorkspaceTask - b
bright-dinner-81701
10/09/2015, 6:22 PMok ill check it out - w
witty-crayon-22786
10/09/2015, 6:24 PM(gist is: pretend everything is local) - e
enough-analyst-54434
10/09/2015, 6:24 PMyes - exactly - w
witty-crayon-22786
10/09/2015, 6:24 PMgot a bit lucky in that it was possible to do in go without rewriting
import statements*.go - e
enough-analyst-54434
10/09/2015, 6:24 PMresolve step maps remotes to local, prep step uses this map + real local to form link farm - b
bright-dinner-81701
10/09/2015, 6:25 PMthe problem is we want a reproducible build, and a normal npm install doesnt provide that, hence i looked into trying to construct the node_module targets bottom up by “cp -R”ing in the resolved 3rd party module targets into place (with their own full trees under them) and dedupin - b
bright-dinner-81701
10/09/2015, 6:25 PMg - b
bright-dinner-81701
10/09/2015, 6:25 PMand that doesnt work - b
bright-dinner-81701
10/09/2015, 6:26 PMso we’re looking at prebuilding node_modules for each target and having Pants re-use that same set subsequently, or having Pants work with shrinkwrap - e
enough-analyst-54434
10/09/2015, 6:27 PMshrinkwrap as a general approach - ie resolver is too smart, let it run once, lock its results if tests pass - has to be simplest I'd imagine - b
bright-dinner-81701
10/09/2015, 6:27 PMok - b
bright-dinner-81701
10/09/2015, 6:27 PMyeah - e
enough-analyst-54434
10/09/2015, 6:27 PMAnd just as good results since either way, the logic is ~broken or arbitrary, consistency is all we care about for the most part - I think - b
bright-dinner-81701
10/09/2015, 6:27 PMyeah that seems reasonable - b
bright-dinner-81701
10/09/2015, 6:28 PMso i thought id have to make Pants work with shrinkwrap in that case - modify npm-shrinkwrap.json to fix the paths for local sources to be what they are in .pants.d - e
enough-analyst-54434
10/09/2015, 6:28 PMcan't pants.d make a workspace to match the community expectations? - e
enough-analyst-54434
10/09/2015, 6:28 PMthats what we do in go - b
bright-dinner-81701
10/09/2015, 6:29 PMi suppose so if you symlink a tree together that matches what was in the source tree, so the relative paths from the shrinkwrap work - b
bright-dinner-81701
10/09/2015, 6:30 PMinside of .pants.d, with those symlinks pointing to the resolved targets - e
enough-analyst-54434
10/09/2015, 6:30 PMFor the go case the boon is plain old normal go command - resent and future - just work in the workspace - b
bright-dinner-81701
10/09/2015, 6:30 PMyes - b
bright-dinner-81701
10/09/2015, 6:30 PMthats what i wanted here too - e
enough-analyst-54434
10/09/2015, 6:30 PMThe downside for go is its slow as hell - compared to go itself