me and @hundreds-father-404 are considering the possibility that having

.*/

as a default ignore pattern in the

--build-ignore

and

--pants-ignore

might be undesirable

why, what is getting ignored that shouldn’t be?

specifically now that we read from your

.gitignore

. My argument is that it will already ignore things like

.idea/

that you don’t want The other argument is that it’s much easier to add to

pants_ignore

than to take away

pants_ignore.add = ['!.ci']

right?

part of the problem here is that the behavior of the gitignore matching library we use and how it interacts with various things pants does with the filesystem is confusing

but yea, i think the “reading the gitignore” support does maybe change the equation there.

one issue there though

Yeah, they don’t. We’d for sure want to add

.git

explicitly, and other popular VCS directories

another issue is that if a pants user is using a non-git vcs, some stuff might break unexpectedly

is that if a pants user is using a non-git vcs, some stuff might break unexpectedly

My argument here is that it’s easier for us to help a user fix this than it is to un-ignore something. Also, that it’s less likely of a problem than the problem Thales is having of “WTF is Pants not reading this. It exists!”

we also need 

build_ignore = ['!.ci']

we should definitely remove

.*

from the default

build_ignore

if it is already in

pants_ignore

that might actually entirely solve his problem

although I think we might say that we only explicitly support git?

I don’t think so.

--changed-since

won’t work, and you’ll have to use a bit more work for

pants_ignore

. But otherwise no reason to require Git

that’s how gitignore syntax works, afaik

it actually might make sense to try to run pants in an environment without git, just to make sure there's nothing unexpected going on

it actually might make sense to try to run pants in an environment without git, just to make sure there’s nothing unexpected going on

This is the type of stress testing we need to do for 2.0. Try to break Pants as much as we can

anyway yeah if we remove

.*/

from teh --build-ignore defaults, then all we have to do to solve thales' problem is make sure

"!/.ci",

is in the --pants-ignore, which is sensible enough

Definitely better. I’d still recommend removing

.*

from the default for

pants_ignore

to be an explicitly enumerated list like

[.git, .svn]

But yeah, at least that’s an improvement

I'm gonna put a PR for the first one. the second one might or might not be a good idea, I think we should think a bit more if we can come up with a good explicitly enumerated list of common dot-dirs