Can't delete read-only files on Windows

[staticfloat]

By default, windows balks at rm'ing files that have the read-only attribute set. This is causing the LibGit2.jl tests to fail on Windows. My favorite solution is to ignore read-only files inside of libuv when removing files on windows. This might involve just chmod'ing them before calling unlink or something, but @jakebolewski and I wanted to run it past the rest of you to see what you thought; is there a better way to do this?

[tkelman]

Ah so that's what that was, good catch and apologies to @jakebolewski for my wrong guess in https://github.com/jakebolewski/LibGit2.jl/pull/11.

Judging by Linux behavior here, I think we should only do a force removal of anything read-only via a new force=true keyword argument to rm. Hopefully read-only files should be a fairly rare occurrence on Windows, what was making them show up in LibGit2?

[staticfloat]

libgit object files are read-only by default. Objects like those .git/objects/XX/<hash> files.

Our linux behavior removes read-only files by default:

$ mkdir test
$ cd test/
$ touch blah
$ chmod -w blah 
$ ll
total 0
drwxr-xr-x   3 sabae  staff   102 Jul 11 18:02 .
drwxr-xr-x+ 80 sabae  staff  2720 Jul 11 18:02 ..
-r--r--r--   1 sabae  staff     0 Jul 11 18:02 blah
$ cd ..
$ julia
               _
   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: http://docs.julialang.org
   _ _   _| |_  __ _   |  Type "help()" to list help topics
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.3.0-prerelease+4134 (2014-07-11 00:01 UTC)
 _/ |\__'_|_|_|\__'_|  |  Commit d42b141* (0 days old master)
|__/                   |  x86_64-apple-darwin13.3.0

julia> rm("test", recursive=true)

julia> 

EDIT: Oh, this is OSX, not linux, but I would be surprised if Linux doesn't do this, as the LibGit2.jl tests have no problem with this kind of operation.

[tkelman]

Oh, so it does. Here I was assuming Julia's rm would/should be more consistent with coreutils rm in that respect. Julia's rm is more like unlink then, which doesn't seem to care as much about permissions.

[staticfloat]

Yeah, rm is basically a recursive wrapper around unlink

[tkelman]

I believe I fixed this in #7590