Dont assume everyone ace exists

[telamonian]

Fixes #9

Pretty much does what I suggested in #9 (comment):

New suggestion:

- If an ACE does not exist, set its mask to `0x0`

- Elevate all rights to at least those of "everyone" by combining masks using bitwise OR `|`

[codecov]

Codecov Report

Merging #11 into master will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master       #11   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            5         5           
  Lines          366       367    +1     
=========================================
+ Hits           366       367    +1     

Impacted Files	Coverage Δ
fs/smbfs/smbfs.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4881635...cf4f966. Read the comment docs.

[telamonian]

@althonos This one is going to be kind of tricky to set up tests for. What would you suggest?

Update

I can at least confirm that this PR does indeed solve the permissions issue I was running into on my CI:

https://ci.appveyor.com/project/telamonian/jupyter-fs/builds/32177461

Fully sucessful example of using a tool based on jupyter -> fs.smbfs -> pysmb on Windows. Wooo!

[telamonian]

Hey @althonos. So it turns out that I can't do the next release of jupyter-fs until I resolve this PR; I can't release a package on pypi that contains a dependency on a github branch (ie this PR branch).

What are your thoughts on this PR? Any chance we can get it pulled in? What's the current blocker?

[telamonian]

Also, I recently had a chance to chat with one of the Gnome VFS devs. They too have a SMB backend, and they've been working on it for quite a while. Sadly, he didn't have any deep advice to offer about the ACL -> posix-like permissions translation issue. There doesn't seem to be any standard lib for this, and, even worse, apparently different implementations of SMB treat ACLs somewhat differently. So one way or the other I think we're stuck with inferring what we can and rolling our own permissions translation implementation.

[althonos]

@telamonian : sorry, real work was hitting hard so I didn't check this. I don't have much ways to check this, since I don't known a lot about Windows permissions (I only know they allow too much combinations to be modelled by UNIX permissions), so I guess I can merge this; I thought this was still WIP.

[telamonian]

sorry, real work was hitting hard so I didn't check this

No worries. I am deeply familiar with the ways of the PhD candidate life.

At least w.r.t. my own code this PR is fairly well battled tested at this point. I'm looking into adding windows ci to this repo. It looks like it will be a fairly complex process, so I think it's best saved for another PR. Once that's in place we'll be able to add some windows-specific permissions tests to the ci

[telamonian]

Thanks for merging! Could I possibly impose on you to also do a new release?

[althonos]

The release is uploaded (v0.6.2)!

I have experience with setting up an AppVeyor in some other projects, but the current test harness relies on Docker so they would probably not play well together. But if AppVeyor VMs spawn a SMB server, that could be feasible!

[telamonian]

But if AppVeyor VMs spawn a SMB server

Yeah, I already learned how to set up that exact thing in appveyor for the ci in jupyter-fs. There's a couple of powershell scripts that make the windows smb magic happen and expose a live share:

https://github.com/jpmorganchase/jupyter-fs/blob/master/ci/enable_os_share_windows.ps1
https://github.com/jpmorganchase/jupyter-fs/blob/master/ci/create_os_share_windows.ps1