Created attachment 7756 [details] Patch to properly free ctx in do_tar. Function do_tar tallocs() memory for ctx, but never frees it before returning. WHen running tarmode against a windows 2000 client I hit this when the first test 'Freed frame ../source3/libsmb/clilist.c:934, expected ../source3/client/clitar.c:821' evaluates to true. Doesn't happen against WinXP clients.
Created attachment 7757 [details] Patch to add a simple tarmode test. As per abartlett, I have attempted to create a test for tarmmode... I'm not too confident if this has any value.
This is against 4.0.0beta5, but it wasn't available as a version.
Whoops, the test is (strequal(finfo->name,"..") || strequal(finfo->name,".")).
This looks good. However we just need to avoid using /tmp, and instead use the selftest prefix. Please pass in $PREFIX like the previous example in tests.py Thanks!
The other thing that would be good is to actually check the contents of the tar file against the original directories, or the remote server (using $LOCAL_PATH).
I'll give it a shot. Would it be worthwile to also do the test in 'reverse' as well, ie. tar -x; mget?
Created attachment 7758 [details] 0002-Add-smbclient-tarmode-test.patch Same patch, exported with format-patch
Created attachment 7759 [details] 0003-Fix-smbclient-tarmode-panic-on-connecting-to-Windows.patch Same patch, exported with format-patch
Created attachment 7760 [details] 0004-Do-not-use-tmp-for-file-storage.patch Utilize LOCAL_PATH & PREFIX, verify data integrity, test both creation and extraction
Created attachment 7761 [details] 0001-Fix-copy-paste-error-in-test-usage-string.patch Quick typo fix I stumbled upon. Doesn't really belong in this bug, per-se.
I did not actuall *test* the changes to the test. Is there a way to selectively run a test, or must all of them run?
Run 'make test TESTS=tarmode' to test just tests with tarmode in the name.
Created attachment 7762 [details] Updated tests Updated (still not working) test
I can't seem to get this test to work... When I run a make test, I am getting NT_STATUS_LOGON_FAILURE.
Created attachment 7763 [details] make_test.log Output from make test TESTS="tarmode"
Comment on attachment 7759 [details] 0003-Fix-smbclient-tarmode-panic-on-connecting-to-Windows.patch Obviously correct, and applies cleanly to 3.6.next also. Please review and I'll get this re-assigned to Karolin for inclusion. Cheers, Jeremy.
This patch isn't needed except for developer builds on master, as the panic is the check that rusty introduced to check that we were using talloc_stackframe() correctly. The remaining patches are now on their way to master, after I worked out the issues with the testsuite.
Sure, but it is an obvious mistake, with a very clear fix - and it exists in 3.6.x. Jeremy.
Comment on attachment 7759 [details] 0003-Fix-smbclient-tarmode-panic-on-connecting-to-Windows.patch Looks goodb
Karolin, please pick for the next release
Pushed to v3-6-test. Closing out bug report. Thanks a lot!