When running samba 3.0.0rc2 the contents of my music folder as displayed on a Windows XP box are 4333 files 18.7GB, but when running samba 2.2.8a my windows box sees all of the contents of that folder which are 26,417 Files 124.3 GB
Created attachment 128 [details] screenshot of folder with samba 2.2.8
Created attachment 129 [details] Screenshot of Samba 3.0.0rc2
I bet those files are full of unusual characters. Can we get a sample of the filenames? I'm guessing this is a problem with 8.3 filename generation as this code has changed a lot during the 3.0 development cycle.
complete files list is avaible at http://qnetalpha.com/samba_3_0_0rc2_file_list.txt
Ok - can you send me a tar file containing all the files - but set to zero length. ie. a tar file that when untarred will create a directory containing all the names but not all the mp3 file contents (don't want to upset the RIAA :-). That way I can test this and see if my recent changes have fixed it. Thanks, Jeremy.
Jeremy, can't you create this from the file list in comment 4? That's why I asked for it.
Do you still need me to create that tar file and if so how do i do it?
Created attachment 133 [details] Tar file of empty filenames Here is a tar file for you Jeremy. Go Python!
there are indeed all kinds of accented chracters in it, which are not ascii. If you did not set the "unix charset" ist's UTF-8 per default and each file which is not validly UTF-8 encoded will not be displayed by samba. You can either set "unix charset" to something like iso8859-1 (that will make every file valid because iso8859-1 is plain 8bit single byte (but valid does not mean correct in this case)) or you can convert the files to UTF-8 with convmv for example, that way, you are really ready for the world and you can also have Chinese song titles ;-). However to make sure everything I wrote is correct I would need a non-0-byte long versionaf all the test files :->
The current SAMBA_3_0 CVS correctly displays 26,417 files now. Closing as fixed. Jeremy.
on the other hand - it doesn't look like over 20000 files are non-ascii and non-utf-8. So there seems to be another problem. First of all, I converted all filenames from latin1 to utf8 to get sane filenames. Taking a closer look at the files you can see that there are newline characters in filenames like in Francis Cabrel - Je l'aime <85> mourir.mp3 An "ls" on the files on the local disk is without errors. An "ls" on the samba3 mounted smbfs gives me a bunch of error messages on some files like the above one. The above one is served by Samba as "Francis Cabrel - Je l'aime ", so everything after the <newlin> is removed, which would explain, why ls complains with "file not found". This probably is a Samba bug. Maybe there are more such unfriendly characters ...
Thanks you guys are the greatest, just remember that my collection is open just ask
originally reported against one of the 3.0.0rc[1-4] releases. Cleaning up non-production versions.
sorry for the same, cleaning up the database to prevent unecessary reopens of bugs.
database cleanup