While trying to build 4.3.0 on a Solaris 10 i386 box the link of smbtorture fails as follows: [3690/3885] Linking default/source4/torture/smbtorture Undefined first referenced symbol in file getgrouplist default/source4/torture/local/nss_tests_1.o ld: fatal: Symbol referencing errors. No output written to /home/projects/tools/samba/samba-4.3.0rc4.i386/bin/default/source4/torture/smbtorture collect2: error: ld returned 1 exit status Note that Solaris does not provide this function. In order to get the build to finish, I edited nss_tests.c to remove the call to getgrouplist. This can be made conditional with the use of #ifdef HAVE_GETGROUPLIST I am not submitting what I did as a patch as I probably did not do it in quite the proper way.
Created attachment 11442 [details] Don't test getgrouplist if we do not have it Here is a patch to source4/torture/local/nss_tests.c to not try to test getgrouplist on systems that do not supply that routine.
Comment on attachment 11442 [details] Don't test getgrouplist if we do not have it Never mind, has already been pushed by jra.
(In reply to Ralph Böhme from comment #2) > Comment on attachment 11442 [details] > Don't test getgrouplist if we do not have it > > Never mind, has already been pushed by jra. If it has, it has not reached master yet.
Created attachment 11479 [details] Patch for master Proper git patch for master. With Tom as author, his signoff, and with bug reference. Also included my an Ralph's review, so this can be pushed as is. Tom: OK?
OK. One of these days I will have to find out how to do a proper git patch. Right now I am using diff -u (actually gdiff -u on Solaris) to make patches. I see that the command you seem to be using is 'diff --git'. I expect that you have everything I need to know published someplace. Now to find the time to look it up. I assume that after the patch is pushed to master then it can be cherry picked for 3.0.next.
(In reply to Tom Schulz from comment #5) > > OK. One of these days I will have to find out how to do a proper git patch. :-) That would be useful if you continue to contribute patches, but as you see, we are quite patient. ;-) > Right now I am using diff -u (actually gdiff -u on Solaris) to make patches. > I see that the command you seem to be using is 'diff --git'. I expect that > you have everything I need to know published someplace. Now to find the time > to look it up. You do a local git clone, do changes, commit your changes to git locally, and then use "git format-patch" to create the git-patches from that. See https://wiki.samba.org/index.php/Using_Git_for_Samba_Development especially: https://wiki.samba.org/index.php/Using_Git_for_Samba_Development#Creating_patches_if_you_don.27t_have_write_access_to_git.samba.org_repositories > I assume that after the patch is pushed to master then it can be cherry > picked for 3.0.next. For 4.3.NEXT rather.
It looks like this will miss 4.3.1. How about for 4.3.2?
Created attachment 11501 [details] Patch for 4.3 cherry-picked from master
Patch doesn't apply to 4.2 as is. Are we ok with having this in only 4.3?
The patch is only needed in 4.3.*. The test with the problem is new in 4.3.
Comment on attachment 11501 [details] Patch for 4.3 cherry-picked from master LGTM.
Re-assigning to Karolin for inclusion in 4.3.next.
(In reply to Jeremy Allison from comment #12) Pushed to autobuild-v4-3-test.
(In reply to Karolin Seeger from comment #13) Pushed to v4-3-test. Closing out bug report. Thanks!