A share defined in registry configuration without a path option is not marked unavailable as a text-defined share, but it is available with "path = /" instead. This bug applies to all Samba versions since 3.2. It has been fixed in 3.4 and master. Michael
Created attachment 4117 [details] patch for the bug Attacht find a git patch mbox (comprised of 4 individual commits) to be applied to the v3-3-test branch. I'd like these to get into the next 3.3 bugfix release. Michael
Isn't f7562e7bf7a7779b2f the only one that is really required to fix the bug? Volker
(In reply to comment #2) > Isn't f7562e7bf7a7779b2f the only one that is really required to fix the bug? > > Volker No, this just make testparm pretend the share is unavailable. The important other path is to convert smbd/service.c to use the higher level routines in loadparm that use libsmbconf and do the path checks. Michael
Ok, sorry. Going back to review the rest. Volker
Ok, we have to make a policy decision here. For the 3.3 release, do we want to have a "minimal patch" policy, or do we want to port back larger changes from master for fixes? If we want the former, then this patch is too large, if we want the latter, then this patch is certainly fine. This is something we need a community decision on, I can't decide this on my own. Volker
(In reply to comment #5) > Ok, we have to make a policy decision here. For the 3.3 release, do we want to > have a "minimal patch" policy, or do we want to port back larger changes from > master for fixes? > > If we want the former, then this patch is too large, if we want the latter, > then this patch is certainly fine. > > This is something we need a community decision on, I can't decide this on my > own. While I think that the patch I have attached is conceptually the right thing to do, it is admittedly more than a pure bugfix for 3.3, so I am going to make a more minimal patch that repeats the check done in loadparm in the server, and I am going to attach that patch for comparison. Michael
Uh, Crap - I just noticed an error in my patch (in the "obvious" part...) I need to figure out why it seems to work anyways. Please forget the attached patch until then... Michael
Created attachment 4120 [details] minimal version of the patch here is a minimal version of the patch. Also the loadparm portion is fixed (the original version did not use a return value of service_ok(), which was not bad in our case, but this one is correct)
That's kindof what I had expected, so +1 :-) Volker
Patch (min version) is upstream and will be included in 3.3.5. Closing out bug report.