#general
- b
bright-dinner-81701
08/27/2015, 12:49 AMno matter what ive done to make 3rd party cacheable npm always ends up re-installing them anyway and rebuilding any C they contain - e
enough-analyst-54434
08/27/2015, 12:49 AMto some extent, Wickman went down that rabbit hole with Python - e
enough-analyst-54434
08/27/2015, 12:50 AMHave you tried linkink a native module you know via external means you;ve already resolved and built ? - e
enough-analyst-54434
08/27/2015, 12:51 AMFrom the pants resolve.npm task side I don;t think that would be too hard to implement as a follow up RB - e
enough-analyst-54434
08/27/2015, 12:52 AMie: we could still use my -g trick - e
enough-analyst-54434
08/27/2015, 12:52 AMeven if we normally use local node_modules/ install under .pants.d/ - e
enough-analyst-54434
08/27/2015, 12:53 AMpants could always use a -g under .pants.d that all the local .pants.d/ installs see - e
enough-analyst-54434
08/27/2015, 12:54 AMmore than a bit of handwave there - e
enough-analyst-54434
08/27/2015, 12:54 AMreal code will be good to cure this - b
bored-art-40741
08/27/2015, 1:01 AMWe've dived quite deep into this problem too - b
bored-art-40741
08/27/2015, 1:02 AMMy conclusion is that unless you have tight control over both the platform and the API, you basically cannot make any guarantees - b
bored-art-40741
08/27/2015, 1:02 AMAnd that propagates all the way up to the top of any general API - e
enough-analyst-54434
08/27/2015, 1:10 AMAnd I think that's fine iff you set expecatations - e
enough-analyst-54434
08/27/2015, 1:10 AMPants does not set expectations wrt to non-Go dep issues downstream yet in clear concise docs - e
enough-analyst-54434
08/27/2015, 1:11 AMand of course the expectatins need to be honored by tasks not caching what they can't be sure they can cache - we miss alot there too - e
enough-analyst-54434
08/27/2015, 1:12 AMIts more and more clear to me Go cam from the g3 monorepo where there can only be 1 version ... thank god - e
enough-analyst-54434
08/27/2015, 1:17 AMWinding back up to @witty-crayon-22786's mention of go 1.5 `.so`s and python - I would definitely support an effort to link the python backend to the go backend (and c++ backend) via a 'c-shared-lib' product dependency and a standard installed noop provider of that product. - e
enough-analyst-54434
08/27/2015, 1:18 AM... and the implied new code, replacement code that could result in a controlled, incremental way - e
enough-analyst-54434
08/27/2015, 1:18 AMMy support is just that though, this definitely deserves wider consideration. - e
enough-analyst-54434
08/27/2015, 1:18 AMWe clearly have more important things to do - b
bored-art-40741
08/27/2015, 1:24 AMDoes python support wrapping go now? - b
bored-art-40741
08/27/2015, 1:25 AMLast I checked this was nearly impossible - e
enough-analyst-54434
08/27/2015, 1:25 AMThat post Stu linked shows not bad at all in 1.5 - e
enough-analyst-54434
08/27/2015, 1:25 AM - b
bored-art-40741
08/27/2015, 1:26 AMOh fun. This is pretty recent I assume? - b
bright-dinner-81701
08/27/2015, 1:26 AMok @enough-analyst-54434 so i just found the only way so far ive gotten npm to not recompile C - e
enough-analyst-54434
08/27/2015, 1:26 AM1.5 was released ~last week - b
bored-art-40741
08/27/2015, 1:26 AMWhen I say last I checked, it was like a year or more ago - b
bright-dinner-81701
08/27/2015, 1:26 AMits ugly tho - e
enough-analyst-54434
08/27/2015, 1:26 AMAnd pants defaults to 15