stat.writable? incorrectly reporting false for some directories on Windows 7

[aschmied]

In 1.7.22:

irb(main):031:0> dir = File.expand_path(ENV['TMP'])
"C:/Users/ANTHON~1/AppData/Local/Temp"
irb(main):032:0> stat = File.stat(dir)
#<File::Stat dev=0x2, ino=0, mode=040755, nlink=1, uid=0, gid=0, rdev=0x2, size=393216, blksize=512, blocks=0, atime=2015-11-30 18:19:32 -0500, mtime=2015-11-30 18:19:32 -0500, ctime=2015-11-16 11:18:42 -0500>
irb(main):033:0> stat.writable?
true

In 1.7.23:

irb(main):039:0> dir = File.expand_path(ENV['TMP'])
"C:/Users/ANTHON~1/AppData/Local/Temp"
irb(main):040:0> stat = File.stat(dir)
#<File::Stat dev=0x2, ino=0, mode=040555, nlink=1, uid=-1, gid=, rdev=0x2, size=393216, blksize=0, blocks=, atime=2015-11-30 06:45:17 -0500, mtime=2015-11-30 06:45:17 -0500, ctime=2015-11-15 23:43:54 -0500>
irb(main):041:0> stat.writable?
false

I noticed this because it broke Dir.tmpdir on this system: every candidate temp directory is skipped.

[nirvdrum]

@enebo It looks like we have more work.

[enebo]

@aschmied @nirvdrum crud. I spent the most time on mode logic with the new stat code. I guess there is some missing logic for directories in particular.

[atambo]

I'm also seeing this pop up when upgrading from jruby-1.7.22 to jruby-1.7.23. This is what the error looks like to a user when running rspec or anything that uses Dir.tmpdir:

ArgumentError: Unable to find a non world-writable directory for Dir::tmpdir. Consider setting ENV['TMPDIR'], ENV['TMP'] or ENV['TEMP'] to a non world-writable directory.
  occurred at C:/jruby-1.7.23/lib/ruby/shared/tmpdir.rb:41:in `tmpdir'

[perlun]

@olleolleolle and @rallenecraft - this may be the issue we've been seeing also, or related to it.

[perlun]

@nirvdrum, @aschmied, @enebo - I did a bit of R&D on the subject in jnr/jnr-posix#67. Unfortunately far from closing this yet, but we are getting started. More details there.

I guess I may have to clone the whole JRuby code and add some more debugging there... it could actually be an idea. Thus far, I've only tried debugging it on the jnr-posix side.

[enebo]

@perlun I mentioned in a comment in that PR that I think this is a generic mode issue and not platform specific. I also thought (cannot 100% remember for sure :| ) that I saw this while only step-debugging through jnr-posix itself. It would make sense that this would be jnr-posix only issue since the JRuby side is not modifying the mode value. getting mode value to match should fix all issues...

[headius]

@perlun Now that we're past the holidays, do you have any updates on this? We can get your jnr-posix PR merged in, but it sounds like we have more work to do.

[perlun]

@perlun Now that we're past the holidays, do you have any updates on this?

Unfortunately not, yet.

We can get your jnr-posix PR merged in, but it sounds like we have more work to do.

I think so yes. My main problem was that I wasn't able to reproduce the problem, in jnr-posix so I think I would more or less have to run JRuby against my "local jnr-posix code". I'm rather unfamiliar with Maven projects etc. so any pointers on how to get this done would be helpful.

I have a Windows 8 and a Windows 10 VM so testing any suggested fix is rather easy.

[headius]

Ahh I can help you run against an updated jnr-posix.

First, in jnr-posix, run:

mvn clean install

This should install a SNAPSHOT of jnr-posix in your local repository.

Then in JRuby, edit core/pom.rb so it references the same version of jnr-posix.

Do a full mvn clean package of JRuby and you should be running on the updated version.

[enebo]

This is crazy. 1.7.23 was not released with the latest version of jnr-posix. I was hurrying because of my impending travel but there is no excuse for this. I am sorry I did not update this as jnr-posix 3.0.22 had some serious issues. Strangely none of our acceptance tests hit it and we still do not have CI properly working on appveyor yet. Appveyor will happen soon since I don't ever want to repeat this...

@perlun can I ask you to build and verify that things work on your various windows boxen? It is possible there is still an issue but the basic issue I can see was fixed between 3.0.22 and 3.0.23 of jnr-posix.

[perlun]

Great @enebo - that would explain things, since I was absolutely unable to repro any of this w/ jnr-posix in itself. 😉

I rebuilt now from the latest jruby-1_7 HEAD and it worked fine. 👏 Both on Windows 8.1 and Windows 10.

Any chance we can have a JRuby 1.7 released w/ this fix included in the near future? Many thanks.

irb(main):004:0> require 'tmpdir'
=> true
irb(main):005:0> Dir.tmpdir
=> "C:/Users/PLUNDB~2/AppData/Local/Temp"
irb(main):006:0> File.stat(Dir.tmpdir)
=> #<File::Stat dev=0x2, ino=0, mode=040755, nlink=1, uid=0, gid=0, rdev=0x2, si
ze=70074368, blksize=nil, blocks=nil, atime=2016-01-12 08:44:21 +0200, mtime=201
6-01-12 08:44:21 +0200, ctime=2013-10-22 18:00:23 +0300>
irb(main):007:0> File.stat(Dir.tmpdir).writable?
=> true

[enebo]

@perlun one more fix involving stat on windows and we should be good for another release. Looking at this week.