pantsbuild #general

Channels

welcome

github-notifications

announce

limited-country-20626

05/06/2016, 10:56 PM

Yeah - I was testing out how well the port process worked and what I needed to do to publish to PyPI.

enough-analyst-54434

05/06/2016, 10:56 PM

Cool!

limited-country-20626

05/06/2016, 10:57 PM

The build files will enforce ' = ' style in BUILD files - so we would need to do some sort of mass transform against the Pants repo or make it configurable.

limited-country-20626

05/06/2016, 10:58 PM

For Pants itself there are still two outstanding hiccups

limited-country-20626

05/06/2016, 10:58 PM

a) does not play nice when being run alongside go buildgen (don't know enough about that goal to know why)

limited-country-20626

05/06/2016, 10:59 PM

b) Needs to understand two broken tests, thrift/antlr importing under pants_test/python

enough-analyst-54434

05/06/2016, 10:59 PM

currently the GoBuildgen makes heavy use of target injection in the graph

enough-analyst-54434

05/06/2016, 10:59 PM

ie, you can hand it 1 go_binary

enough-analyst-54434

05/06/2016, 11:00 PM

And it derives N targets from its transitive graph

enough-analyst-54434

05/06/2016, 11:00 PM

I suspect that may cause issues with your stuff? Maybe not

enough-analyst-54434

05/06/2016, 11:01 PM

iow, it doesn't just fix deps, it builds graphs

enough-analyst-54434

05/06/2016, 11:01 PM

since the go tooling supports this natively and speedily

enough-analyst-54434

05/06/2016, 11:01 PM

Not sure if your stuff builds graphs or just fixes deps in existing BUILD files

limited-country-20626

05/06/2016, 11:03 PM

I am trying to trigger the go failure - but something I merged from upstream is breaking (PARALLEL TESTS changes...)

enough-analyst-54434

05/06/2016, 11:04 PM

I'll be interested in this if you file an issue with the backtrace

limited-country-20626

05/06/2016, 11:04 PM

yes - can do. Just have to get my checkout in a decent state - I think I badly rebased - not a big issue

limited-country-20626

05/06/2016, 11:05 PM

I also have some thoughts about publishing independent contrib modules, now that one has been created

limited-country-20626

05/06/2016, 11:07 PM

Okay, the go conflict looks simple enough - Fsqio buildgen operates over a list of source roots, but passing those source roots to the go buildgen task is against the rules of that task:

Copy code

23:05:32 00:12     [go]
FAILURE: Can only BUILD gen for a single Go local sources source root, found:
	contrib/go/3rdparty/go
	contrib/go/examples/src/go
	contrib/go/testprojects/src/go

limited-country-20626

05/06/2016, 11:07 PM

I will open an issue - although it may just be better fixed from our end.

rough-dog-96241

05/06/2016, 11:08 PM

@enough-analyst-54434: is this http://www.pantsbuild.org/howto_plugin.html still accurate, or does that not work with virtualenvs? (I’d think that setting PYTHONPATH in the ./pants script wouldn’t do much if pants is installed in a virtualenv.)

enough-analyst-54434

05/06/2016, 11:10 PM

Looks out of date, instead see pants dogfood: https://github.com/pantsbuild/pants/blob/master/pants.ini#L25-L36

enough-analyst-54434

05/06/2016, 11:10 PM

This is re: you developing your own plugin(s)?

enough-analyst-54434

05/06/2016, 11:11 PM

An RB to fix those docs would be great if

[GLOBAL] pythonpah

works for you and you can find time

rough-dog-96241

05/06/2016, 11:12 PM

👍

enough-analyst-54434

05/06/2016, 11:12 PM

src/docs/howto_plugin.md

enough-analyst-54434

05/06/2016, 11:13 PM

Apropos to trogdor ascii art

HadoopBinary

in those docs!

rough-dog-96241

05/06/2016, 11:20 PM

@enough-analyst-54434: is

%(buildroot)

my pants build root, or pant’s build root?

enough-analyst-54434

05/06/2016, 11:20 PM

Its the current repo build root

enough-analyst-54434

05/06/2016, 11:21 PM

where the pants script and/or pants.ini live

rough-dog-96241

05/06/2016, 11:21 PM

And what you linked me to is pant’s pythonpath override.