Restoring a idmap dump ('net idmap restore') created from 'net idmap dump' with Samba 3.5.x results in these error messages: Could not set USER HWM: WBC_ERR_DOMAIN_NOT_FOUND Could not set GROUP HWM: WBC_ERR_DOMAIN_NOT_FOUND Since the HWM doesn't get set, the tdb allocator ends up allocating previously-assigned UIDs and GIDs when new accounts are used.
Hi Justin, Thanks for you bug report. Could you please add your smb.conf ? Thanks - Michael
Created attachment 5858 [details] smb.conf
Created attachment 5870 [details] diff to git master I was able to track down the problem. _wbint_SetHWM() was switching on an uninitialized value. The attached patch fixes it for me. It applies to 3.5.x and master. I don't think it's relevant for 3.4.x.
Thanks Justin ! Sorry for you having to fix your own bug, but much appreciated :-). I'll review this asap and ensure it gets pushed where needed. Cheers, Jeremy.
Created attachment 5871 [details] Revised patch for master. Hi Justin, I actually think there are two typos here. Can you check this revised patch and I'll push if you ok it. Thanks, Jeremy.
Ok, I already pushed the first piece, sorry. Jeremy, go ahead, your bug. Volker
Comment on attachment 5871 [details] Revised patch for master. Well as you've already looked at it Volker, will you review for 3.5.5 and re-assign to Karolin if you're ok with this ? Thanks, Jeremy
Comment on attachment 5871 [details] Revised patch for master. Give it a nack. To me this does not look like a "git format-patch" style patch. I will upload one next.
Didn't know that was now a requirement :-). I'll update my internal procedures accordingly... :-) Jeremy.
Created attachment 5872 [details] Patch Jeremy, it is a lot easier for Karolin to apply a patch if it is generated by "git format-patch" instead of "git diff". For your convenience, I am uploading pretty much the same content of the patch that this replaces in "git format-patch" format. If you want to review it for correctness, feel free to do so and re-assign to Karolin for inclusion.
Comment on attachment 5872 [details] Patch +1 from me.
Well, it might not be a requirement, but I *try* to be as nice to our Release Manager as possible. And "git am -3" is simpler than "patch -p1; git commit -a --author= -m <invent-a-checkin-message>". At least *I* think that the step to invent a checkin message should be done by the patch author and not by Karolin. But we can certainly discuss this. If you see that as the duty of Karolin, then fine. Volker
Re-assigning to Karolin for inclusion in 3.5.x. Jeremy.
vl - no I'm not arguing. I simply take the path of least resistance, and this is what I always *used* to do with svn. I will update my practices accordingly. I agree with you that this should be the patch submitters job, not Karolins. Jeremy.
It still works fine for me with the second part of the patch, so we're good from my perspective.
Pushed to v3-5-test. Closing out bug report. Thanks!