pantsbuild #development

Channels

welcome

github-notifications

development

fancy-queen-20734

05/01/2017, 8:45 PM

My thought here is that it maps to what v1 does, and would be incrementally better than the current behavior.

fancy-queen-20734

05/01/2017, 8:46 PM

Right now, the graph trace is pretty internally facing

fancy-queen-20734

05/01/2017, 8:46 PM

--print-exception-stacktrace is also an internalish thing

witty-crayon-22786

05/01/2017, 9:27 PM

@rough-minister-58256 : src/python/pants/backend/jvm/tasks/bootstrap_jvm_tools.py is one of the last remaining users of

get_alternate_target_roots

and uses it to inject a bunch of targets

witty-crayon-22786

05/01/2017, 9:29 PM

...

changed

is the other, but it's deprecated.

witty-crayon-22786

05/01/2017, 9:29 PM

but the bootstrap task's usage looks like something that could move to your dep declaration task

rough-minister-58256

05/01/2017, 9:33 PM

witty-crayon-22786

05/02/2017, 5:56 PM

@rough-minister-58256 : re: https://github.com/pantsbuild/pants/issues/4541 ... would requesting older images in travis be sufficient?

rough-minister-58256

05/02/2017, 5:57 PM

hopefully, yeah

rough-minister-58256

05/02/2017, 6:00 PM

worst case, we should be able to use e.g. docker https://docs.travis-ci.com/user/docker/

rough-minister-58256

05/02/2017, 6:00 PM

and point that to an older OS image just to conduct the native engine build

rough-minister-58256

05/02/2017, 6:01 PM

it’s not clear what the OS library of travis is out of the box

rough-minister-58256

05/02/2017, 6:01 PM

but with docker we have lots of options: https://hub.docker.com/_/centos/

rough-minister-58256

05/02/2017, 6:02 PM

we could probably build the bin against the manylinux image, which is ~centos5

rough-minister-58256

05/02/2017, 6:02 PM

and designed for forward-compatibility-focused builds

witty-crayon-22786

05/02/2017, 6:13 PM

@rough-minister-58256 : is https://github.com/pantsbuild/pants/issues/3409 still a blocker? i'll trade you that one for https://github.com/pantsbuild/pants/issues/4541

witty-crayon-22786

05/02/2017, 6:17 PM

also, wondering what your strategy for target injection is going to look like

witty-crayon-22786

05/02/2017, 6:18 PM

it's possible i shouldn't touch https://github.com/pantsbuild/pants/issues/4539 if you're changing things such that bootstrap might not need to use

get_alternate_target_roots

rough-minister-58256

05/02/2017, 6:18 PM

I think I can restore the hack for #3409 in the new model

witty-crayon-22786

05/02/2017, 6:20 PM

rough-minister-58256

05/02/2017, 6:32 PM

have been in the weeds a bit, but I think the new interfaces are pretty solid now - was going to take a stab at target injection next

witty-crayon-22786

05/02/2017, 7:40 PM

ok, cool. don't rush... i think i have a workaround here.

rough-minister-58256

05/02/2017, 7:51 PM

witty-crayon-22786

05/02/2017, 8:05 PM

but i would be interested in seeing the direction you're taking, if you think you have time to post a preview.

rough-minister-58256

05/02/2017, 8:13 PM

posted https://github.com/pantsbuild/pants/pull/4542

rough-minister-58256

05/03/2017, 9:56 PM

https://github.com/pantsbuild/pants/issues/4547

rough-minister-58256

05/03/2017, 9:57 PM

looks like this is another case where explicit whitelisting for

pants_ignore

(e.g.

!.babelrc

) would be required.. wonder if it would make sense to more broadly tackle that for 1.{3,4}.0 tho?

rough-minister-58256

05/03/2017, 9:58 PM

maybe we could more concretely ignore [

..

.gitignore

, ?]

witty-crayon-22786

05/03/2017, 10:03 PM

another more reasonable default might be to ignore hidden directories by default, but not hidden files.

rough-minister-58256

05/03/2017, 10:03 PM

sgtm