Anyone here who uses Pants in conjunction with doctest? https://docs.python.org/3.9/library/doctest.html

Is there a way to combined

--changed-since

with actual listed targets in a single Pants command? Like I want the union of what the changed subsystem gives plus the things I specify. The changes subsystem doesn't seem to like this

❯ ./pants --changed-since=origin/main list tests/local/email/test_email.py 
10:26:05.80 [ERROR] You used `--changed-since` at the same time as using file arguments. You can only use `--changed-since` or use normal arguments.

you want the *intersection or the union?

nevermind I don't want no part of this 😅

I hadn't realized that there were so many edge cases around doing that, and so many potential interpretations of what it might mean. I'll write my own plugin to solve my use case 😤

if you're motivated, we'd definitely welcome a PR to improve the error message with some recipes of what users may be wanting to do

To clarify: our CI has a feature where we can inject "virtual changes" via messages in commits. This is mostly to let people run extra tests without making dummy changes, and then we can go back and audit CI logs for these and try to use that info to improve our dep graph. So I want to union those things we scrape from messages with the set of actual changes, calculate the dependant transtive closure, and use that as the input set to our next operations. I can easily do this with multiple Pants commands, but locally it triggers pantsd restarts more than I want so I end up waiting more than I'd like. Trying to think about how to combine this stuff into a single Pants command.

FYI, for Pants 2.14, it appears that the

remote_auth_plugin

is not deprecated but is in fact now required to be removed. I'm running through the upgrade now on a bunch of repos

Hi everybody - quick question on naming things. Please fill in the blank in the thread: GIven:

A imports B

• A depends on B • B is like the parent in the relationship • A is like the child in the relationship (think: things in A can subclass or inherit from things in B) • B is a

dependency

of A • A is a

___???___

of B What word/noun in English could replace

___???___

?

Hello. I find that pants rebuild a lot of things that to me, didn't change on my code. So, I was wondering if there is an easy way to understand what makes pants rebuild something. Like a magic command that could tell me that this "packaging" target would be rebuilt because those targets changed since the last time the package was built. Any ideas?

I'm testing 2.14 and in my repo, I have a plugins for pants. Now, pants tells me it can't infer owners for some imports like pants.engine.fs.Digest and so on. Should I add pantsbuild.pants in my requirements.txt ?

Do you have a target like

pants_requirements(name="pants", resolve="pants-plugins")

in your build file under your plugin directory?

I don't , just the regular python_requirements

Ah, yeah, I think that's one of the early steps of setting up plugins

See https://www.pantsbuild.org/docs/plugins-overview#building-in-repo-plugins-with-pants

Ah, I thought it wasn't mandatory

I'll give it a try, thanks!

definitely not mandatory - only a best practice to treat your plugin code like production quality, that you want to run linters etc on it

ok but if I'm not including that, how can I fix the error I got?

You'd have to set a whole bunch of

# pants: no-infer-dep

or whatever it is

Or likely change something in

pants.toml

to not warn about it

Since it's a one liner and also lets you set up a venv to have nice editor autocompletion when developing in that directory, I'm a big fan even if you don't use it for linting/tests/etc.

[python-infer].unowned_dependency_behavior = 'ignore'

I think

I'm a big fan of the unknown dependencies being warnings (with option to elevate to error) in the newest releases. It caught a few places in our repo where folks hadn't declared direct requirements properly and were grabbing things transitively.

it works with the pants_requirements. I'll keep it 🙂

thanks for the help!

I'm testing

[pytest].xdist_enabled = true

. I wonder if I should add pytest-xdist on

[pytest].extra_requirements

(I guess yes given it is on the default value). And also I don't see much difference. The execution seems to set

-n 1

on the pytest command line. But I have 2 cpus. Is it expected?

Pants 2.14 minor issues of note 1. In build_ignore the escaping appears to have changed. I had to change

build_ignore = ['\\{\\{ cookiecutter.__repo_name \\}\\}']

to

build_ignore = ['\{\{ cookiecutter.__repo_name \}\}']

. 2. For sh tests that sourced the original sh file, I had to add the original sh file as an explicit dependency to the sh_tests section of the BUILD file. (ie if

lib/foo_test.sh

has a line like

source lib/foo.sh

I had to add

:lib

to the shunit2_tests dependencies. I'm actually not sure how this worked previously, since I had a different repo where I had the explicit dependency, although that used

source "$(dirname "${BASH_SOURCE[0]}")/foo.sh"

Hi, I have a test that will use `packages`to fetch defined structs in a local file. however, it can't find anything when I ran it with pants 2.14. it works fine with

go test

though