Created attachment 6545 [details] patchset for v3-6-test idmap_autorid is going to be released first with 3.6.0. This is why I collect two little improvements to autorid in this single bug report. If we need to split, then I can still do it. (1) autorid initialization should fail if configured for a non-default domain (i.e. for domain != "*"): It only makes sense for the default config. (2) The configuration should be made more systematic in the sense that I would like to use "idmap config * : rangesize = autorid" instead of "autorid : rangesize". Attached find a patchset (already applied to master) that fixes the two issues for v3-6-test. It contains the doc fix for (2) and the actual code fix for (2) consists of two patches, one preparatory and the actual config change.
Comment on attachment 6545 [details] patchset for v3-6-test Ambi, please review.
Comment on attachment 6545 [details] patchset for v3-6-test adding the stackframe and freeing it leads to access to freed memory as the config that is allocated in the stackframe is stored as private data additionally, the dynamic calculation if the parameter with config_option = talloc_asprintf(frame, "idmap config %s", dom->name); becomes unnecessary with the next patch that limits the module to the default config, so this can be done with a fixed string
(In reply to comment #2) > Comment on attachment 6545 [details] > patchset for v3-6-test > > adding the stackframe and freeing it leads to access to freed memory as the > config that is allocated in the stackframe is stored as private data > > additionally, the dynamic calculation if the parameter with config_option = > talloc_asprintf(frame, "idmap config %s", dom->name); > > becomes unnecessary with the next patch that limits the module to the default > config, so this can be done with a fixed string As discussed already: Thanks for catching the talloc error! It was a bad oversight in my patch. Update patch following...
Created attachment 6556 [details] patch for master This is a patchset for master that first reverts the old patches (apart from the doc fix) and then re-does the patches the right way. (So that the proposed v3-6-test patch that I am going to attach next can be identical to the top master commits).
Created attachment 6557 [details] updated patchset for v3-6-test Updated patchset for v3-6-test this consists of the top commits of the master patch (less the revert commits)
Created attachment 6559 [details] updated patchset for master The master patchset did not apply any more after the change of TALLOC_ZERO_P to talloc_zero. ==> updated patchset attached.
Ambi, did you have a chance to review/test the new patches? Thanks - Michael
Comment on attachment 6557 [details] updated patchset for v3-6-test reviewed the patches, but did not have time to test them out. But from pure looking, they seem to be good
I just pushed the (master) patches to autobuild. Assigning to Karolin for inclusion into 3.6.0. Cheers
Pushed to v3-6-test. Closing out bug report. Thanks!
There was a patch missing to the idmap_autorid manpage to reflect the changes of this bug. I did not want to open an new bug for this...
Created attachment 6617 [details] patch for the 3.6.0 idmap_autorid manpage patch attached. this patch is already in master.
Karo, please add the manpage patch to 3.6.0. Thanks!
Pushed. Closing out bug report. Thanks!