This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Trouble with Git 2.1.x pushing to repos over Samba

Hi John,

On Apr 30 18:44, John Orr wrote:
> > From: "Michael Enright" 
> > $ git push origin master
> > fatal: '//host/path/to/repo.git/' does not appear to be a git repository
> > fatal: Could not read from remote repository.
> [...]
> #: john@johndesktop:/cygdrive/l ; ls -ld .git/objects/
> drwxr-xr-x 1 john Unix_Group+1000 0 Nov 13 14:13 .git/objects/
> (albeit, Corinna, with my group issue still not yet resolved)

You tried the /etc/group tweak as I suggested in my latest mail in that
thread, I take it?

> access(/cygdrive/l/.git, R_OK) returned 0
> access(/cygdrive/l/.git, W_OK) returned 0
> access(/cygdrive/l/.git, X_OK) returned -1
> The last test is the one run by git, that makes it reject my /cygdrive/l/.git directory.
> Not sure if that's relevant, but just in case.

Thanks for the info.  I found a really dumb bug in my code.  The
access() function is using a Windows function for access checking under
the hood.  To account for the Samba account mapping in Cygwin, there's
a function converting the S-1-22-x-y SIDs in the file's ACL to Windows
SIDs if there *is* a mapping.  But I made a small mistake which has
a big result: The ACL is not completly copied over, thus the Windows
function has to deal with an incomplete ACL.

I fixed that in the git repo and uploaded new snapshots to  Please give them a try.  Don't use the
snapshots for anything else for the time being!


  I'll be unavailable for a few weeks starting tomorrow, so I'd like to
  do a bugfix Cygwin release, preferredly today, if this patch works as


P.S.: As a side-note: While this patch (hopefully) reverts this code to
      work as pre-1.7.34, it seems that the internal Windows access
      check function is not quite up to the task for Samba shares in
      scenarios as John's one.  It will always report back the access of
      the "others" part of POSIX permission bits.  Only with the new
      mapping of S-1-22-x-y SIDs to real WIndows accounts, or with
      winbindd-supported mapping, the Windows access check will really
      work as desired.

Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

Attachment: pgp1sulpSlPrr.pgp
Description: PGP signature

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]