pantsbuild #pex

Channels

welcome

github-notifications

development

enough-analyst-54434

05/13/2016, 4:56 PM

I'm not sure exactly what's going on, but I think the non-pre-processing of MANIFEST.in is due to the on-the-surface odd invocation of setup.py here: https://github.com/pantsbuild/pex/blob/5aea7d0de5acbcd2881251facace208009d4afeb/pex/installer.py#L102-L111

enough-analyst-54434

05/13/2016, 4:57 PM

In case anyone has stared at this code, but I'll file an issue shortly

rough-minister-58256

05/13/2016, 5:16 PM

seems like

<http://MANIFEST.in|MANIFEST.in>

is a pre-processing thing handled in a

python setup.py sdist

- I believe pex currently expects to be handed a well formed sdist, which is usually the form most packages take with

pex .

rough-minister-58256

05/13/2016, 5:17 PM

also, MANIFEST.in afaict seems to be a distutils thing - pex forces setuptools

rough-minister-58256

05/13/2016, 5:17 PM

not sure tehre’s tons of merit in trying to support legacy distutils packaging

rough-minister-58256

05/13/2016, 5:18 PM

since most projects are readily abandoning that nowadays

rough-minister-58256

05/13/2016, 5:18 PM

seems like maybe this guy is cargo culting super old examples

enough-analyst-54434

05/13/2016, 5:22 PM

Sorry - talking about

pex . -o setup_py.pex

enough-analyst-54434

05/13/2016, 5:23 PM

since we actually run setup.py for local dir resolvables, we should get it right it seems to me

enough-analyst-54434

05/13/2016, 5:23 PM

no - MANIFEST.in definitely works with setuptools

enough-analyst-54434

05/13/2016, 5:23 PM

just tested this

enough-analyst-54434

05/13/2016, 5:24 PM

I guess it seems to me if we are gunning for the simplest and most complete compatibility by just running setup.py, which we do - weirdly - then it should fully work and not half work

enough-analyst-54434

05/13/2016, 5:25 PM

I'll tag this as an idea if a quick experiment says this can be fixed at the high level of just running setup.py less weirdly

rough-minister-58256

05/13/2016, 5:27 PM

ah k

rough-minister-58256

05/17/2016, 12:04 AM

thinking about starting to publish nightly dev snapshot packages of pex for easier development consumption by pants - anyone have any thoughts?

rough-minister-58256

05/17/2016, 12:05 AM

this makes me want to add support for VCS resolvables to pex

rough-minister-58256

05/17/2016, 4:47 AM

seems that’s not a good idea to publish such a thing to pypi

rough-minister-58256

05/17/2016, 4:48 AM

need a better way to test pex development changes in pants master without the whole release/consume cycle via pypi

rough-minister-58256

05/17/2016, 4:48 AM

I guess VCS resolvables is ideal

rough-minister-58256

05/17/2016, 6:00 PM

trying out a new label to make ready-to-be-reviewed pex PRs more discoverable:

rough-minister-58256

05/17/2016, 6:00 PM

Pasted image at 2016_05_17 11_00 AM.png

rough-minister-58256

05/25/2016, 6:15 PM

https://blog.twitter.com/2016/open-sourcing-twitter-heron

rough-minister-58256

05/25/2016, 6:15 PM

using bazel + pex.. which means that the first real pex+skylark impl just hit the public.

rough-minister-58256

05/25/2016, 6:15 PM

https://github.com/twitter/heron/blob/master/tools/rules/pex_rules.bzl

faint-holiday-3863

06/18/2016, 7:07 PM

kwlzn: I’ve been working on improving the bazel pex rules so they work with sandboxing enabled and are more consistent with some of the built-in rules

faint-holiday-3863

06/18/2016, 7:08 PM

kwlzn: do you know if there are plans to publish a standalone repository for those rules, a la https://github.com/bazelbuild/rules_go and various others?

faint-holiday-3863

06/18/2016, 7:09 PM

If not, I might take that on myself

faint-holiday-3863

06/18/2016, 11:57 PM

here’s what I’ve got so far: https://github.com/benley/bazel_rules_pex

faint-holiday-3863

06/19/2016, 12:59 AM

still have some work to do to make it work smoothly as an external repo

faint-holiday-3863

06/19/2016, 7:39 PM

okay, got that working too