Add command line globbing to all utilities (for windows targets)

[rivy]

Command line globbing is added (for windows targets) to all utilities via 'uucore'.

This change increases portability and parity across machine targets.

This initial commit uses the published 'wild' crate. The second commit references a personal fork of 'wild' which updates the API and includes proper case insensitivity (again, for windows targets). I plan to update and re-reference the published 'wild' crate when it is updated to include the fixes.

[rivy]

It's not obvious how to fix the failing tests from my end.

I do believe they are due to bugs in wild and/or glob, but glob seems semi-deprecated and wild is hard for me to grok in order to issue a PR for repair of the problem (maybe the author of wild could offer a hand? @kornelski?).

Maybe the three failing tests should just be gated out for Windows for the time being until the globbing story becomes more polished (see rust-lang/glob#59).

[kornelski]

Note that Windows' quoting/escaping syntax is bizarre, and incompatible with any sensible Unix-like expectations. It has weird hacks to tolerate unescaped backslashes, and weird cases where:

exe """"""""

is parsed as "" (i.e. "\"\"").

Also keep in mind that cargo run doesn't preserve quoting, so if you test anything that way it'll give wrong results. Only executables launched directly without cargo have access to unparsed command line.

[rivy]

Agreed, CMD has a LOT of weird edge cases. But the initial failing test that I see is echo \\ returning just a single \. No quotes involved. In fact, when I tested the executable with various permutations of slashes, it will "compress" them.

C:> REM "compression" of slashes
C:> uutils echo \
\
C:> uutils echo \\
\
C:> uutils echo \\\\\\
\
C:> uutils echo \\.
\.
C:> uutils echo \\\\.
\.
C:> uutils echo /
/
C:> uutils echo ///////
/
C:> uutils echo //.
/.
C:> uutils echo ////.
/.
C:> REM but if not a possible PATH? ...
C:> uutils echo \\x //x
\\x //x

I think the wildcard/globbing process is being overly aggressive. If there aren't any glob characters in an argument (or if globbing returns an empty set), the argument should just be returned. Isn't that a wild/globbing issue?

[kornelski]

Oh, I see. Fair point about overly-aggressive glob. I'll check that out.

[kornelski]

Fixed in wild v1.1.1 and v2.0.1

[rivy]

👍

Looking better, and passing all tests, on my machine... Let's see how we do with AppVeyor.

[rivy]

Excellent, everything is passing, thanks @kornelski !

[vks]

This is problematic, because * is a valid file or folder name on Unix. If I understand correctly, this PR makes it impossible to remove such files and folders.

Only enabling the feature on Windows might work, because there * is not allowed in file and directory names.

[kornelski]

The wild crate is a no-op on everything but Windows (on Unix it re-exports std::env::args() and that's it).

[rivy]

@Arcterus , I've got a branch with rust@1.27 and revised / working Travis CI testing at https://github.com/rivy/rust.coreutils/tree/fix.ci, specifically rivy@deead6f. It includes AppVeyor changes as AppVeyor CI was failing for odd reasons; so I remodeled it based upon a newer template (see https://github.com/starkat99/appveyor-rust/blob/master/appveyor.yml).

The full comparison is master...rivy:fix.ci.

The Travis CI build is at https://travis-ci.org/rivy/rust.coreutils/builds/432010513.

The AppVeyor CI build is at https://ci.appveyor.com/project/rivy/rust-coreutils/build/2%20~%20fix.ci.

I can add all or part of it to this PR... let me know how you'd like me to proceed.

(( reposted from sub-discussion above ))

[Arcterus]

It’s fine if you include it in this PR.

[rivy]

I've used this updated PR as a base for further fixes, and it seems like a good new base.

All tests are working correctly for all of my newer commits (which I'll rebase and PR after this is merged).

[cnd]

@Arcterus can we merge this?

[rivy]

Ping @Arcterus.

Anything changes needed before a merge?

I've got a couple of other pending fixes based on this commit set.

[rivy]

@Arcterus ,

Just FYI, I've got several pending PRs based on this now (#1304, #1305, #1315, #1316, #1317, #1318).

[rivy]

@Arcterus @cnd

I just wanted to check in now that the holidays are over and see if we can get this merged (as well as the other PR's mentioned above) ...

[rivy]

@vks ?

[Arcterus]

Really sorry about being MIA for the past while. This seems good to me though!