Create a share with unix extensions enabled and try to stat it.
Created attachment 15712 [details] patch for 4.11
Created attachment 15714 [details] patch for 4.10
Re-assigning to Karolin for inclusion in 4.11.next, 4.10.next
(In reply to Jeremy Allison from comment #4) Pushed to autobuild-v4-{11,10}-test.
(In reply to Karolin Seeger from comment #5) Pushed to both branches. Closing out bug report. Thanks!
See https://gitlab.com/samba-team/devel/samba/pipelines/120957076
(In reply to Volker Lendecke from comment #7) Is it 4.10 and 4.11 only (applies cleanly on 4.12)? And is the first patchset for 4.12, also?
Created attachment 15839 [details] additional patch for 4.12
Created attachment 15840 [details] additional patch for 4.11
Created attachment 15841 [details] additional patch for 4.10
Andreas - are these extra patches complete ? Volker did lots more work inside libsmbclient in master to fix the tests and make readdirplusXXX use the POSIX extensions. Do we need those changes also ?
I've just backported what was part of that Merge Request. I didn't look into the tests. I think we could backport the tests to 4.12 if you want but for lower version there are probably too many changes.
It's not the tests that actually matter, it's the code that changes cli_list() to use the POSIX extensions etc. Look at: git format-patch --stdout 9cc546956dcfdf7c898154aa79d0947eb4c9c578..a1856f07b39d4e8bc8f66cb49c7666b0d5ed99b8 I think that's the bare minimum on top of what you already have. Adding the tests as well will make sure you got it right.
Actually, don't merge those fixes yet. I have one thing to check with Volker first...
Jeremy, Andreas, any updates here? The code in master has a valgrind error. I've tried to fix this in 51551e0d53f, but this lead to an avalance of more and more posix functionality in libsmb, which eventually led to 9653a107389, re-introducing the original valgrind error.
Nope, working on other stuff. I had a call with Andreas, where we decided we needed an explicit smbc_I_want_posix() sort of call, but until then probably the best solution is just never set the try_posixinfo variable.
Until we have that call, I think we need a different solution for the valgrind error. Any idea?
What I mean: I think we should not just leave this "frame" vs "fname" error in the code, this needs fixing.
Sure, but why not just fix and comment this out as it should be falling back anyway ?
If I see it correctly, this would basically mean the patchset for this bug won't be used at all. But that means that this bug won't be fixed properly. Andreas, what do you think?
Lets comment out the code to disable try_posixinfo till we have a proper solution.
Sorry for the inconvenience here.
(In reply to Andreas Schneider from comment #22) > Lets comment out the code to disable try_posixinfo till we have a proper > solution. Can you explain a bit more what you want to say here? You want us to never disable try_posixinfo? This would mean that with every stat we run into the valgrind error. Do you have time to give me a quick hint or provide a simple patch? Thanks a *LOT* for your cooperation!
Created attachment 15901 [details] patch to remove it again for 4.12 and 4.11 and 4.10.
Comment on attachment 15901 [details] patch to remove it again for 4.12 and 4.11 and 4.10. I provided a test build to Fedora users and that solved a long-standing issue reported as https://bugzilla.redhat.com/show_bug.cgi?id=1801442#c28 which people already started reproducing in other distributions.
A revert was pushed out to Fedora stable updates yesterday: F32: https://bodhi.fedoraproject.org/updates/FEDORA-2020-7f0df2c653 F31: https://bodhi.fedoraproject.org/updates/FEDORA-2020-2655ce257f For the record, without this revert people were seeing some files shown as folders when talking to SMBv1 servers. This affects many Linux distributions already.
Karolin, please apply the patch to remove it please. Sorry for the inconvenience ... Thanks.
Comment on attachment 15901 [details] patch to remove it again for 4.12 and 4.11 and 4.10. Sorry for the delay.
(In reply to Jeremy Allison from comment #29) Pushed to autobuild-v4-{12,11,10}-test.
(In reply to Karolin Seeger from comment #30) Pushed to all branches. Closing out bug report. Thanks!
*** Bug 14379 has been marked as a duplicate of this bug. ***