11/17/04 0:06:08 Signoff: deryck ("Leaving") 11/17/04 0:28:44 tridge_ (~tridge@bh02i525f01.au.ibm.com) has joined channel #samba-technical 11/17/04 0:37:16 Signoff: tridge (Read error: 60 (Operation timed out)) 11/17/04 0:41:16 porkpie is now known as ppsleep 11/17/04 0:42:10 lfeldman (~Lon@208.35.40.2) has joined channel #samba-technical 11/17/04 0:44:45 jazzyjames (~jazzy@ip-81-186.dot.net.au) has joined channel #samba-technical 11/17/04 0:59:29 Signoff: clonny2 (Read error: 110 (Connection timed out)) 11/17/04 1:03:20 pipitas is now known as pipitas|afk 11/17/04 1:10:52 Signoff: mimir (""goodnight"") 11/17/04 2:00:16 Signoff: Wbunaarf (" HydraIRC -> http://www.hydrairc.com <- Get hot chicks here!") 11/17/04 2:39:10 Signoff: idra ("Leaving") 11/17/04 3:42:50 daveemc2 (~Administr@fmgh8-13.2wcm.comporium.net) has joined channel #samba-technical 11/17/04 4:16:14 Signoff: jmcd ("Leaving") 11/17/04 4:19:07 Signoff: d0rt (Remote closed the connection) 11/17/04 5:09:49 Signoff: daveemc2 (Read error: 110 (Connection timed out)) 11/17/04 6:40:03 metze (~metze@p508D7569.dip.t-dialin.net) has joined channel #samba-technical 11/17/04 6:58:42 tridge_ is now known as tridge 11/17/04 6:58:49 yay! we have streams in pvfs 11/17/04 6:58:55 Mode change "+o tridge" on channel #samba-technical by ChanServ 11/17/04 7:02:25 hi tridge: cool 11/17/04 7:02:30 I need to go ... 11/17/04 7:02:32 bye 11/17/04 7:02:35 bye 11/17/04 7:03:01 Signoff: metze ("savIRC - http://savirc.sourceforge.net") 11/17/04 7:11:43 metze (~metze@p508D7569.dip.t-dialin.net) has joined channel #samba-technical 11/17/04 7:13:28 hi again 11/17/04 8:14:22 tmus (~tmus@blueice1n1.uk.ibm.com) has joined channel #samba-technical 11/17/04 8:14:49 good morning all! 11/17/04 8:18:13 * tridge rides home 11/17/04 8:18:16 Signoff: tridge ("Client exiting") 11/17/04 8:24:32 tpot_ (~tpot@ppp223-122.lns1.cbr1.internode.on.net) has joined channel #samba-technical 11/17/04 8:25:56 Signoff: acidkill ("Lost terminal") 11/17/04 8:27:01 vl (~vlendec@interner.SerNet.DE) has joined channel #samba-technical 11/17/04 8:28:20 hi vl 11/17/04 8:31:21 hi everyone 11/17/04 8:32:56 hi * 11/17/04 8:35:36 howdy! 11/17/04 8:43:55 tridge (~tridge@203-173-8-192.perm.iinet.net.au) has joined channel #samba-technical 11/17/04 8:43:56 Mode change "+o tridge" on channel #samba-technical by ChanServ 11/17/04 8:46:28 hi tridge 11/17/04 8:46:33 hi volker 11/17/04 8:46:34 hi tridge 11/17/04 8:46:52 Streams are in ... :-) 11/17/04 8:47:10 yep! 11/17/04 8:47:43 I've got another nasty problem to solve here. If you have profiles on a samba share, changing the icon of a folder in the start menu does not survive a logoff/logon... 11/17/04 8:48:00 Another 3000+ packet sniff to compare between windows and Samba... 11/17/04 8:48:20 try with samba4 11/17/04 8:48:35 btw, do you know how to enable xattr support in samba4? 11/17/04 8:48:42 Okay... Good hint. 11/17/04 8:48:52 No idea. 11/17/04 8:48:57 for ext3, you need to use the mount option user_xattr 11/17/04 8:50:45 thanks 11/17/04 8:51:48 tridge: are xattrs on xfs are automaticly used? 11/17/04 8:52:13 installing the "attr" package is also useful, so you can look at xattrs using the getfattr command 11/17/04 8:52:23 metze: i think so, but I'm not sure 11/17/04 8:52:31 metze: its been a while since I tested XFS 11/17/04 8:52:43 metze: if you have XFS, please give it a go 11/17/04 8:53:01 RAW-STREAMS is a good test to try 11/17/04 8:53:30 ok 11/17/04 8:59:17 yay, pvfs passes the IFSTEST "AlternateStreamTest" 11/17/04 9:00:16 hmm, but not StreamInformationTest .... 11/17/04 9:29:15 ppsleep is now known as porkpie 11/17/04 9:39:36 Hmmmm. How do I add a user to samdb.ldb? 11/17/04 9:39:46 Manual ldbadd? 11/17/04 9:39:53 yep 11/17/04 9:40:00 vl: ldbedit is easiest 11/17/04 9:40:08 or ldbedit name=administrator 11/17/04 9:40:17 and then just copy the user 11/17/04 9:40:23 in the editor 11/17/04 9:40:41 vl: you may not know that ldbedit allows the creation of new records, so copying works as metze describes 11/17/04 9:40:54 Ah, ok. That's probably best. 11/17/04 9:41:09 does usermanager works with samba4 yet? 11/17/04 9:41:10 vl: if you'd like to commit a scripts/adduser.pl then please do ;) 11/17/04 9:41:20 metze: should do, but i haven't tried 11/17/04 9:41:27 :-) 11/17/04 9:46:24 So if I just edit all fields corresponding to 'administrator' I get a new account? 11/17/04 9:47:54 Okay... Works :-) 11/17/04 9:48:22 Hmmm. How do I set the initial password? 11/17/04 9:48:53 add a file "unicodePwd" 11/17/04 9:48:57 field 11/17/04 9:49:07 vl: plaintext 11/17/04 9:50:01 you might also find the "unixName" field useful 11/17/04 9:50:19 vl: set unixName to the unix username you want to map to 11/17/04 9:50:35 Ok, works. This needs documentation. Online irc help does not scale too well... 11/17/04 9:50:39 ;-) 11/17/04 9:50:48 please edit prog_guide.txt 11/17/04 9:51:00 but really this is just a developer hack for the moment 11/17/04 9:51:09 i don't expect samba4 production users to do this! 11/17/04 9:51:28 People would literally toast us 11/17/04 9:51:58 Or simply ignore samba4 11/17/04 9:52:05 of course 11/17/04 9:52:14 admin tools are totally lacking now 11/17/04 9:52:24 while we concentrate on basic functionality 11/17/04 9:52:48 Probably I'll going to try samba3 'net rpc user add' later. 11/17/04 9:53:02 should work 11/17/04 9:53:19 BTW, samr for group memberships is completely missing. Something to work on? 11/17/04 9:53:26 Likewise aliases. 11/17/04 9:53:38 ? it does group memberships 11/17/04 9:53:44 vl: or code libnet_user_add() :-) 11/17/04 9:54:01 Only query I think. 11/17/04 9:54:16 no addmember/delmember/queryusergroups I think. 11/17/04 9:54:27 Have to look again. 11/17/04 9:54:32 hmm, possibly, i stopped working on samr when someone started complaining about lack of posix backend :) 11/17/04 9:54:39 *duck* 11/17/04 9:55:03 i think i'll finish pvfs before going back to samr again :) 11/17/04 9:55:11 good ! 11/17/04 9:55:30 samr is more or less just mapping rpc calls to ldap calls. 11/17/04 9:55:39 conceptually easier I think. 11/17/04 9:55:40 yes, plus access control 11/17/04 9:56:54 there will be a fairly major change when we add ACLs into ldb, pvfs and all rpc servers 11/17/04 9:57:20 it will be a fairly straightforward change, but will impact many functions 11/17/04 9:58:45 tridge: I'll will write a lib/security/ soon 11/17/04 9:59:08 metze: ok - core function will be a nt_access_check() function 11/17/04 9:59:36 metze: and a test suite that uses ntcreatex to test the access_check logic for random combinations of access masks and ACLs 11/17/04 10:00:05 no:-) security_check_acl() 11/17/04 10:00:12 ok 11/17/04 10:00:22 security_check_privilege() 11/17/04 10:01:03 we'll have struct security_token, security_descriptor, security_acl,... 11/17/04 10:01:23 should we name dom_sid => security_id then? 11/17/04 10:01:35 no, its too well known a name 11/17/04 10:01:42 ok 11/17/04 10:02:02 back to smbd vs. samba 11/17/04 10:02:48 I for 'samba' and 'sambadm' 11/17/04 10:03:02 argl. My kernel obviously does not support user_xattr... 11/17/04 10:03:10 vl: 2.6 ? 11/17/04 10:03:15 2.4.22 11/17/04 10:03:38 I'll boot a vmware with something else. 11/17/04 10:03:54 yeah, its an optional patch in 2.4 I think 11/17/04 10:04:03 sles8 probably has it. 11/17/04 10:05:03 Hmmm. No. 11/17/04 10:17:23 grr: I have a corrupted softwareraid array on my server... 11/17/04 10:18:16 metze: grr doesn't seem to be here right now... 11/17/04 10:18:18 ;-) 11/17/04 10:18:33 metze: in all seriousness, is it just loss of one disk or more than one? 11/17/04 10:18:35 Hi abartlet 11/17/04 10:18:41 * abartlet has unfortunetly had some experience in this area... 11/17/04 10:18:55 abartlet: the filesystem is broken 11/17/04 10:19:36 metze: I had a server that booted 2 disks out of the array 11/17/04 10:19:46 because it didn't like the way they spun up... 11/17/04 10:19:59 ? 11/17/04 10:20:26 anyway I'll switch over the monitor to my server, back later 11/17/04 10:20:29 well it went to run the array, and decided by a read test that the disks were bad... 11/17/04 10:20:56 they were not, as it turned out, and I got very much to know the mdadm and raid commands :-) 11/17/04 10:30:22 Hmmm. RAW-STREAMS should work? 11/17/04 10:30:26 I get 11/17/04 10:30:36 opening non-existant directory stream 11/17/04 10:30:36 (torture/raw/streams.c:121) Incorrect status NT_STATUS_ILLEGAL_CHARACTER - should be NT_STATUS_NOT_A_DIRECTORY 11/17/04 10:30:42 against samba4 11/17/04 10:30:56 ./smbtorture //127.0.0.1/tmp -Uvl%asdf RAW-STREAMS 11/17/04 10:31:28 hi vl 11/17/04 10:31:58 hi abartlet 11/17/04 10:32:02 again :-) 11/17/04 10:38:52 vl: RAW-STREAMS works if your target server supports xattrs 11/17/04 10:39:05 vl: that error means you haven't enabled xattrs right ;) 11/17/04 10:39:20 vl: or configure didn't find xattr support in the C library 11/17/04 10:39:33 vl: try "grep XATTR include/config.h" 11/17/04 10:40:16 vl: the ILLEGAL_CHARACTER error comes from the fact that on filesystems that don't support streams, the ':' stream separator is illegal 11/17/04 10:41:16 * tridge wonders if vl is still here 11/17/04 10:41:28 #define HAVE_ATTR_XATTR_H 1 11/17/04 10:41:28 #define HAVE_SYS_XATTR_H 1 11/17/04 10:41:50 /dev/mapper/rootvg-distslv on /dists type ext3 (rw,user_xattr) 11/17/04 10:41:51 there should be this: 11/17/04 10:41:53 #define HAVE_XATTR_SUPPORT 1 11/17/04 10:42:33 if you don't see that, then look in config.log and find out why it failed 11/17/04 10:43:01 you may need a package like "attr-dev" or "attr-devel" 11/17/04 10:43:08 depending on what distro this is 11/17/04 10:43:10 It's a sles9 back1:/usr/src/samba4/source # rpm -qa|grep attr 11/17/04 10:43:10 libattr-2.4.12-56.1 11/17/04 10:43:10 attr-2.4.12-56.1 11/17/04 10:43:10 libattr-devel-2.4.12-56.1 11/17/04 10:43:43 ok, look in config.log for "flistxattr" 11/17/04 10:43:49 thats the function it tests for 11/17/04 10:44:06 checking for library containing flistxattr 11/17/04 10:44:09 result: none required 11/17/04 10:44:41 AC_SEARCH_LIBS(flistxattr, [attr]) 11/17/04 10:44:41 if test x"$ac_cv_func_flistxattr" = x"yes"; then 11/17/04 10:44:41 AC_DEFINE(HAVE_XATTR_SUPPORT,1,[Whether we have xattr support]) 11/17/04 10:44:42 fi 11/17/04 10:44:48 thats the m4 test code 11/17/04 10:45:10 /* #undef HAVE_XATTR_SUPPORT */ 11/17/04 10:45:16 This is in config.h 11/17/04 10:45:18 try this: 11/17/04 10:45:26 grep flistxattr config.cache 11/17/04 10:45:33 you did use configure -C, yes? 11/17/04 10:45:43 no config.cache 11/17/04 10:45:51 ahh, run configure again wiith -C 11/17/04 10:45:52 No. ./configure --prefix=bla 11/17/04 10:45:55 the cache is useful :) 11/17/04 10:46:09 it will show us the string that the if statement is using 11/17/04 10:46:46 running configure... 11/17/04 10:47:13 ack1:/usr/src/samba4/source # grep flistxattr config.cache 11/17/04 10:47:13 ac_cv_search_flistxattr=${ac_cv_search_flistxattr='none required'} 11/17/04 10:47:20 Signoff: tpot_ (Remote closed the connection) 11/17/04 10:47:41 hmm, should be 2 lines 11/17/04 10:47:47 the other one should look like this: 11/17/04 10:47:50 ac_cv_func_flistxattr=${ac_cv_func_flistxattr=yes} 11/17/04 10:48:01 so, maybe autoconf version problem 11/17/04 10:48:02 nope, that's all. 11/17/04 10:48:10 2.59 11/17/04 10:48:11 maybe I didn't use a portable autoconf method 11/17/04 10:48:42 try using this one: 11/17/04 10:48:49 http://samba.org/~tridge/configure 11/17/04 10:48:54 thats my configure 11/17/04 10:49:06 if it works for you then we know its autoconf versions 11/17/04 10:51:00 running your configure 11/17/04 10:51:40 No change 11/17/04 10:52:48 gd_asleep is now known as gd 11/17/04 10:59:29 hi gd 11/17/04 11:00:59 hi 11/17/04 11:04:09 vl: its fixed now 11/17/04 11:04:15 thanks 11/17/04 11:06:14 it looks like people have been busy today :-) 11/17/04 11:07:18 tridge: I somehow think I'll be hearing 'try samba4' fairly often in future :-) 11/17/04 11:09:40 I just triggered a discussion on openldap-devel how to best support poll/epoll & friends. 11/17/04 11:10:08 I wonder if this also applies to Samba4. 11/17/04 11:10:52 Howard Chu seems to look at libevent which supports many scalable select variants out of the box. 11/17/04 11:14:45 brb 11/17/04 11:18:46 waider (~waider@itg-gw.cr008.cwt.esat.net) has joined channel #samba-technical 11/17/04 11:19:47 hi waider 11/17/04 11:25:55 RAW-STREAMS took 1.10801 secs 11/17/04 11:27:36 hola 11/17/04 11:27:53 vl: its a short test :) 11/17/04 11:28:08 vl: you can also manupulate them with smbclient 11/17/04 11:28:12 vl: try this: 11/17/04 11:28:19 smbclient //foo/blah .... 11/17/04 11:28:33 smb> put /etc/hosts x.txt 11/17/04 11:28:43 smb> put /etc/services x.txt:services 11/17/04 11:28:49 xp refuses to store its profile on samba4. It first wants to have a level 1005 netrShareGetInfo on the share 11/17/04 11:28:54 smb> allinfo x.txt 11/17/04 11:29:09 vl: can you add that? 11/17/04 11:29:37 lets see what it does :-) 11/17/04 11:35:55 case 1005: 11/17/04 11:35:55 r->out.info.info1005->dfs_flags = 0; 11/17/04 11:36:01 return WERR_OK; 11/17/04 11:36:11 :-) 11/17/04 11:36:16 rpc_server/srvsvc/dcesrv_srvsvc.c:611 11/17/04 11:36:17 looks tricky :) 11/17/04 11:36:21 this segfaults for me. 11/17/04 11:36:31 Do I have to talloc the 1005 struct? 11/17/04 11:36:35 vl: is it ref? 11/17/04 11:36:50 [out,switch_is(level)] srvsvc_NetShareInfo info 11/17/04 11:36:51 vl: yes, it think you do 11/17/04 11:37:24 [case(1005)] srvsvc_NetShareInfo1005 *info1005; 11/17/04 11:37:36 its a ptr, so you need to alloc it 11/17/04 11:37:57 talloc hanging from the dce_call? 11/17/04 11:38:14 from mem_ctx 11/17/04 11:38:20 see all the other calls above .... 11/17/04 11:38:29 Hmm. Thanks :-) 11/17/04 11:38:48 Not my system there... 11/17/04 11:39:08 and while you're there, changing all those "0", "512" etc to names from the IDL would be good 11/17/04 11:41:10 Oh great. It tries to create the profile directory itself using the NT Transact Request. 11/17/04 11:41:26 samba4 returns UNSUCCESSFUL 11/17/04 11:41:33 NTTRANS CREATE ? 11/17/04 11:41:40 Yes 11/17/04 11:41:46 i'm wokring on that right now 11/17/04 11:41:50 give me a few hours 11/17/04 11:41:50 :-) 11/17/04 11:41:54 :-) 11/17/04 11:41:59 its needed for IFSTEST 11/17/04 11:42:07 trying the 'mkdir on unix' workaround 11/17/04 11:42:10 basically its just ntcreatex, but with EAs and a SD 11/17/04 11:43:57 Now it wants a successful QUERY_SECDESC. 11/17/04 11:44:17 I think XP profiles on samba4 are a bit off for now :-) 11/17/04 11:44:18 maybe come back next week :) 11/17/04 11:44:52 They might turn out to be one of the harder high-level tests to fulfil. 11/17/04 11:45:05 naah 11/17/04 11:45:37 Want to see the sniff of a profile stored on samba3 vs w2k3? 11/17/04 11:45:58 I'm back to staring at ethereal... 11/17/04 11:46:04 no, i'd rather ocontinue with the core stuff i'm doing 11/17/04 11:46:15 :-) 11/17/04 11:46:23 i know we need ACLs, and they are on the schedule for next week or so 11/17/04 11:46:29 This was not a serious request. 11/17/04 11:46:45 They just do all sorts of silly things while storing the profile. 11/17/04 11:48:06 I have just started with libcli/security/ 11/17/04 11:48:16 metze: good 11/17/04 11:48:32 metze: i'll do acl store/fetch soon in pvfs, so you have something to test with 11/17/04 11:50:11 essbee (~sean@nat-pool-brisbane.redhat.com) has joined channel #samba-technical 11/17/04 11:51:21 it looks like it's full steam ahead in samba4 :-) 11/17/04 11:52:52 * vl is really impressed 11/17/04 11:53:36 i haven't tested the WinXP SP2 "download security token" stuff in streams yet, but should work 11/17/04 11:53:43 vl: give it a try if you have time 11/17/04 11:54:39 I test it how? 11/17/04 11:54:53 use IE to download a executable 11/17/04 11:54:54 With IE save a downloaded file to a samba4 share? 11/17/04 11:55:00 then try and run it from samba4 11/17/04 11:55:09 should pop up a warning 11/17/04 11:55:21 and "allinfo" in smbclient should show the extra stream 11/17/04 11:56:44 Any .exe url available? 11/17/04 11:56:58 putty.exe... 11/17/04 11:57:50 what is the "download security token" ? 11/17/04 11:58:12 XP SP2 started using streams to store info on where a file came from 11/17/04 11:58:25 so, if you download a file, it marks it as "internet" 11/17/04 11:58:30 and warns on execute 11/17/04 11:58:42 vl: http://www.putty.nl/latest/x86/putty.exe 11/17/04 11:58:47 i'm expecting a PR effort from MS to point out that samba can't do this 11/17/04 12:00:04 yes, I'm supprised I've not seen it yet :-) 11/17/04 12:00:04 stream 1: 11/17/04 12:00:04 size 26 11/17/04 12:00:04 alloc size 512 11/17/04 12:00:04 name :Zone.Identifier:$DATA 11/17/04 12:00:07 stream 1: 11/17/04 12:00:07 size 26 11/17/04 12:00:07 alloc size 512 11/17/04 12:00:10 name :Zone.Identifier:$DATA 11/17/04 12:00:12 ups 11/17/04 12:00:31 And I get the warning 11/17/04 12:00:31 looks good 11/17/04 12:00:46 How do I list all xattrs from unix? 11/17/04 12:01:09 vl: great 11/17/04 12:01:30 vl: getfattr 11/17/04 12:01:51 yay, i got the security warning on open 11/17/04 12:01:54 so it works :) 11/17/04 12:01:55 # file: dists/samba4/putty.exe 11/17/04 12:01:55 user.DosAttrib 11/17/04 12:01:55 user.DosStream.Zone.Identifier 11/17/04 12:01:55 user.DosStreams 11/17/04 12:02:11 For me as well. 11/17/04 12:02:25 try getfattr -d 11/17/04 12:02:38 base64 stuff 11/17/04 12:03:21 Cool! Congrats! 11/17/04 12:03:27 thanks! 11/17/04 12:03:34 its limited to 64k of course 11/17/04 12:04:01 i've got a flags field in there so we can put large streams in .streams/ later 11/17/04 12:04:05 Yes, but the code to get them into a separate file should be doable. 11/17/04 12:04:16 yep 11/17/04 12:04:30 bbl (lunch) 11/17/04 12:10:53 tridge: what is your feeling on modification to a domain we are torturing? 11/17/04 12:11:08 i want to modify the domain object, so I can see the output 11/17/04 12:11:33 abartlet: if its a destructive modification, then condition those tests on -X 11/17/04 12:11:34 but so far most of the tests have just been adding users, ie easily spotted and deleted 11/17/04 12:11:46 just changing the comment 11/17/04 12:11:55 'tortured by samba 11/17/04 12:11:58 ' 11/17/04 12:12:04 non-destructive is fine I think 11/17/04 12:12:20 if you _really_ value your domain, don't run a tool called "torture" :) 11/17/04 12:12:54 :-) 11/17/04 12:13:00 thanks 11/17/04 12:13:10 tridge: is there already a plan how to support long filenames in samba4, i.e. filenames resulting in more than 256 bytes in utf-8 ? 11/17/04 12:13:37 bjacke: why should they be a problem? 11/17/04 12:14:40 windows support 256 UTF-16, better said, 256 UCS2 characters, which is in worst case more than twice the length we support when we store filenames in UTF-8 11/17/04 12:15:11 which unix filesystems enforce that limit? 11/17/04 12:15:54 it's a POSIX limit I think. 256 characters is the default limit of Linux for exmaple, I think the limit of all Unces 11/17/04 12:16:15 s/characters/bytes/ 11/17/04 12:17:21 is that really enforced in systems like ext3 ? 11/17/04 12:17:39 I filed a bugreport some time ago: https://bugzilla.samba.org/show_bug.cgi?id=257 11/17/04 12:17:50 yes, it's a hardcoded filesystem limit 11/17/04 12:17:51 for samba3, sure 11/17/04 12:18:22 well, samba4 doesn't have any specific name limits in the code, but if the underlying filesystem is limited then samba4 will be too 11/17/04 12:18:39 and I don't think doing mapping workarounds for this is a good idea 11/17/04 12:19:25 Signoff: grantb (Read error: 60 (Operation timed out)) 11/17/04 12:19:25 Signoff: jazzyjames (Read error: 60 (Operation timed out)) 11/17/04 12:20:41 in worst case it's in fact three times the length we support :-| 11/17/04 12:21:49 tridge: do we support to only generate the print functions for a struct in pidl 11/17/04 12:24:18 bjacke: thats a pretty unusual case I'd guess, but i could see this being a problem in some cases 11/17/04 12:25:19 I think it should be fixed in the filesystem 11/17/04 12:25:46 and till the release of samba 4, there a lot of time to put it in the linux kernel 11/17/04 12:26:05 yep 11/17/04 12:26:12 i'd be surprised if some filesystems on linux don't already support much longer names 11/17/04 12:28:28 no, it's limited by the OS: 11/17/04 12:28:28 linux/limits.h:#define NAME_MAX 255 /* # chars in a file name */ 11/17/04 12:28:42 sure, but does anybody use that define? 11/17/04 12:28:44 is it enforced? 11/17/04 12:28:47 (aside from the posix spec) 11/17/04 12:28:52 or is that just for sysctl 11/17/04 12:29:12 Signoff: nemisus (Read error: 110 (Connection timed out)) 11/17/04 12:31:21 there appear to be several header files which define it, checking to see if it's defined first. posix_1m.h uses 256, some others use 255, and I see one 4096. 11/17/04 12:31:48 /usr/include/linux/limits.h:#define PATH_MAX 4096 11/17/04 12:31:52 xfs has that limit, jfs has, ext2/3 has, I think all have 11/17/04 12:31:58 Signoff: [cat] (Read error: 60 (Operation timed out)) 11/17/04 12:31:58 so I guess it's like any other 'standard': pick your own :D 11/17/04 12:31:59 * tridge writes a test program 11/17/04 12:32:14 :-) 11/17/04 12:32:28 PATH_MAX is just for the actual pwd 11/17/04 12:32:32 #define PATH_MAX 640 /* should be enough for anyone */ 11/17/04 12:32:59 waider, MAX_PATH is not for file names 11/17/04 12:33:34 ah 11/17/04 12:33:49 hmm, limit is 255 on ext3 11/17/04 12:34:06 bjacke: but i thought PATH_MAX was the upper limit on filename + directory 11/17/04 12:34:14 which is obviously incorrect 11/17/04 12:34:23 * waider reboots his head 11/17/04 12:35:11 MAX_PATH limits you to go into a very deep nested directory, that doesn't hurt us so soon 11/17/04 12:35:43 hee. come back SVR4, all is forgiven: bits/posix1_lim.h:#define _POSIX_NAME_MAX 14 11/17/04 12:35:50 SVR3, rather 11/17/04 12:57:36 tridge: I've solved some more of the riddles you left in samr.idl :-) 11/17/04 12:57:47 oh good! 11/17/04 12:57:59 'oem_info', I've pinned down where it shows up, both in SAMR and SamSync 11/17/04 12:58:10 it looks very much like a comment to me :-) 11/17/04 12:58:22 nice 11/17/04 12:58:37 I've got an odd puzzle for secrets 11/17/04 12:59:00 the value in lsa and netlogon differ slightly 11/17/04 12:59:04 but only on some values 11/17/04 13:00:31 the comment? 11/17/04 13:00:41 no, the secret value 11/17/04 13:00:58 different encryption? 11/17/04 13:01:06 no, an extra header 11/17/04 13:01:28 I'm just looking at it again - it just might be my code... 11/17/04 13:16:42 jazzyjames (~jazzy@202.147.87.94) has joined channel #samba-technical 11/17/04 13:16:50 is there a way to find the name to an inode number, on a (xfs) filesystem 11/17/04 13:19:48 metze: is there a debugfs equivalent? 11/17/04 13:19:52 that's how i would do it on ext2 11/17/04 13:20:27 metze: or try find -inum 23123123, if you don't mind scanning hte whole disk 11/17/04 13:20:37 or, if it's open, lsof may help 11/17/04 13:20:49 mbp_: thx 11/17/04 13:22:04 a xfs_ncheck 11/17/04 13:22:12 ? 11/17/04 13:22:21 is that the debugfs thing? 11/17/04 13:23:59 it's 'debugfs ncheck' 11/17/04 13:24:11 I think 11/17/04 13:30:16 abartlet: What are you up to right now? Simply more vampire stuff? 11/17/04 13:30:20 Just curious 11/17/04 13:31:40 mbp_: xfs_db == debugfs 11/17/04 13:31:45 vl: yep 11/17/04 13:31:49 xfs_ncheck is just a wrapper 11/17/04 13:31:55 I'm hoping to have a test for the whole vampire area 11/17/04 13:32:06 Sounds good :-) 11/17/04 13:32:10 but at least, users, groups, domains, trusted domains and secrets 11/17/04 13:32:18 abartlet: do we get the password hashes via the samsync 11/17/04 13:32:23 yes 11/17/04 13:32:33 (for groups, i'll start with just 'the group exists') 11/17/04 13:38:19 [cat] (~cat@217.21.35.33) has joined channel #samba-technical 11/17/04 13:55:37 vl: so, anything in partiuclar you think I should torture next (in samlogon)? 11/17/04 13:55:52 the lsa session key? ;-)) 11/17/04 13:55:57 lm session key 11/17/04 13:56:14 actually, that's already tested 11/17/04 13:56:25 How does samba3 perform here? 11/17/04 13:56:29 badly :-) 11/17/04 13:56:35 Okay... 11/17/04 13:56:50 it's all quite easy to fix, gd should have little trouble 11/17/04 13:57:03 if he spends some 'quality time' with the testsuite :-) 11/17/04 13:57:09 vl: i've added NTTRANS CREATE, and seems to work, i'll do the setacl soon 11/17/04 13:57:14 * tridge gets some sleep 11/17/04 13:57:20 Good 11/17/04 13:58:45 dort (~d1rt@69.37.140.220.adsl.snet.net) has joined channel #samba-technical 11/17/04 13:58:57 I was actually looking for samsync tests - what things does samba3 do badly at now (other than trusted domains)? 11/17/04 13:59:14 Alias memberships. Not at all. 11/17/04 14:01:09 The ntuser.ini file, which format is that? 11/17/04 14:01:40 .ini or ntuser.dat? 11/17/04 14:01:49 .ini 11/17/04 14:02:32 * abartlet would suggest .ini format then... :-) 11/17/04 14:02:43 (possibly utf16 unicode, try iconv) 11/17/04 14:03:09 file is such a useful tool :-) 11/17/04 14:03:11 It looks like UCS-2, but iconv does not like the feff at the start 11/17/04 14:03:22 delphin:/data/tmp # file vl/ntuser.ini 11/17/04 14:03:22 vl/ntuser.ini: MP32, Stereo 11/17/04 14:03:23 :-) 11/17/04 14:03:28 :-) 11/17/04 14:03:42 mpg123 vn/ntuser.ini ;-) 11/17/04 14:06:42 jmcd (~jmcd@ptd-24-93-131-73.maine.rr.com) has joined channel #samba-technical 11/17/04 14:07:41 hi jmcd 11/17/04 14:13:37 pipitas|afk is now known as pipitas 11/17/04 14:17:22 abartlet: does the cifs module in linux support symlinks? 11/17/04 14:17:26 hi abartlet 11/17/04 14:25:49 tarjei_: yes, against a samba server 11/17/04 14:26:02 jmcd: look, i readied one of your bugs for closure ;-) 11/17/04 14:27:07 tmus: hehe, cool 11/17/04 14:27:28 lukeh (~lukeh@lester.off.padl.com) has joined channel #samba-technical 11/17/04 14:27:47 hiddenman (~andy@eva.a-teleport.com) has joined channel #samba-technical 11/17/04 14:27:51 hi all 11/17/04 14:29:46 ppl, why winbindd_idmap.tdb is so "breakable"? i often have broken mapping MS AD->unix uids ;-( 11/17/04 14:29:49 abartlet: thanks. 11/17/04 14:34:08 killahertz (~me@213.68.11.198) has joined channel #Samba-technical 11/17/04 14:34:25 afternoon all 11/17/04 14:35:18 just installed rpms from sernet.de from sles8... with compiled heimdal 0.4e with kadmin patch, and i'm getting the following error on "net join ads" net ads join -U adminag 11/17/04 14:35:18 adminag's password: 11/17/04 14:35:18 [2004/11/17 14:23:21, 0] libads/ldap.c:ads_add_machine_acct(1366) 11/17/04 14:35:18 ads_add_machine_acct: Host account for ottawa already exists - modifying old account 11/17/04 14:35:18 Using short domain name -- LHSFRA 11/17/04 14:35:20 [2004/11/17 14:23:21, 0] libads/kerberos.c:get_service_ticket(337) 11/17/04 14:35:21 get_service_ticket: kerberos_kinit_password OTTAWA$@LHS-SYSTEMS.COM@LHS-SYSTEMS.COM failed: Preauthentication failed 11/17/04 14:35:25 Segmentation fault 11/17/04 14:35:37 Signoff: killahertz (Excess Flood) 11/17/04 14:36:21 killahertz (~me@213.68.11.198) has joined channel #Samba-technical 11/17/04 14:37:03 the problem was not there in 3.0.7 or 3.0.4... is this a problem with 3.0.8 ?? 11/17/04 14:38:27 sorry for the previous flood 11/17/04 14:40:28 vl: ok, groups and alisas are now done (for the top level info, not members) 11/17/04 14:40:43 net ads join -U adminag 11/17/04 14:40:43 adminag's password: 11/17/04 14:40:45 [2004/11/17 14:23:21, 0] libads/ldap.c:ads_add_machine_acct(1366) 11/17/04 14:40:45 ads_add_machine_acct: Host account for ottawa already exists - modifying old account 11/17/04 14:40:45 Using short domain name -- LHSFRA 11/17/04 14:40:45 [2004/11/17 14:23:21, 0] libads/kerberos.c:get_service_ticket(337) 11/17/04 14:40:47 get_service_ticket: kerberos_kinit_password OTTAWA$@LHS-SYSTEMS.COM@LHS-SYSTEMS.COM failed: Preauthentication failed 11/17/04 14:40:49 Segmentation fault 11/17/04 14:41:20 killahertz: flooding here will nto get your bug fixed faster 11/17/04 14:41:43 sorry.. i know.. i keep pasting in the wrong window ;( 11/17/04 14:44:42 so is this a bug in libads or a kerberos problem? 11/17/04 14:47:59 lukeh has left channel #samba-technical because () 11/17/04 14:49:32 'night all! 11/17/04 15:14:38 Signoff: tmus (Read error: 104 (Connection reset by peer)) 11/17/04 15:23:27 Signoff: killahertz () 11/17/04 15:33:05 Spiekey (~spiekey@pD9E05391.dip.t-dialin.net) has joined channel #samba-technical 11/17/04 15:33:08 hi 11/17/04 15:33:14 has anyone got a good samba/ldap howto for me? the idealx howto is driving me nuts! 11/17/04 15:37:50 hiddenman has left channel #samba-technical because ("Client exiting") 11/17/04 15:37:55 Signoff: jmcd ("Leaving") 11/17/04 15:41:52 hm. one cannot enum shares with the rap-call on 3_0 at the moment. 11/17/04 15:41:58 anyone an idea? 11/17/04 15:42:07 jmcd (~jmcd@ptd-24-93-131-73.maine.rr.com) has joined channel #samba-technical 11/17/04 15:55:12 3_0 server or 3_0 client? 11/17/04 16:01:11 3_0 server 11/17/04 16:03:24 okay... Also have it here. 11/17/04 16:03:43 Later :-) 11/17/04 16:08:30 :) it's not in release, btw. 11/17/04 16:09:29 deryck (~deryck@131.204.73.24) has joined channel #samba-technical 11/17/04 16:09:45 morning, all. 11/17/04 16:10:23 Hmmm. One of my changes? 11/17/04 16:40:47 Signoff: Spiekey () 11/17/04 16:44:22 YESS -- solved the start menu folder icon problem :-) 11/17/04 16:44:36 what was it? 11/17/04 16:45:03 The folder in the start menu has to have the 'read only' flag set, otherwise the desktop.ini that it contains is ignored. 11/17/04 16:45:43 oh my god... 11/17/04 16:45:44 The trans2_queryfileinfo is ignored, the trans2_findfirst returning the folder name including the stat info is relevant. 11/17/04 16:46:15 cool, huh? Another hack for 'profile acls = yes'. 11/17/04 16:47:33 nice one, vl 11/17/04 16:56:29 anybody seen cupsd crash when a printer is deleted and samba reloads the printer list? :) 11/17/04 16:57:17 Adam has left channel #samba-technical because ("Client Exiting") 11/17/04 17:01:08 Signoff: metze ("savIRC - http://savirc.sourceforge.net") 11/17/04 17:13:32 hm, looks like there's a fix in 1.1.22rc1 11/17/04 17:17:49 poseiden_ (~poseiden@leviathan.inethouston.com) has joined channel #samba-technical 11/17/04 17:22:05 Signoff: tridge (Read error: 60 (Operation timed out)) 11/17/04 17:25:32 everyone must be working hard on samba code in here :-P 11/17/04 17:26:01 Signoff: jmcd ("Leaving") 11/17/04 17:26:15 Abood (~BLACKFAIT@212.38.144.157) has joined channel #samba-technical 11/17/04 17:35:11 jmcd (~jmcd@ptd-24-93-131-73.maine.rr.com) has joined channel #samba-technical 11/17/04 17:40:47 jmcd: can you take a look at the first two patches in http://www.loonybin.net/~rob/ ? 11/17/04 17:40:57 both are pretty short 11/17/04 17:42:36 Rwf: just headed out the door for a bit...maybe in ~90mi 11/17/04 17:42:37 min 11/17/04 17:44:05 Rwf: ooh you should have jerry look at these 11/17/04 17:44:16 the 2nd one at least is a bit sensitive... 11/17/04 17:46:20 how so? 11/17/04 17:47:16 oh wait, I didn't look closely enough 11/17/04 17:47:17 hehe 11/17/04 17:47:24 I think i'll need to take a look after lunch 11/17/04 17:47:28 take your time :) 11/17/04 17:47:37 printer deletions are really nasty right now 11/17/04 17:47:57 i delete a printer in cups, samba gets really confused, stuff gets left all over the disk, and cupsd crashes 11/17/04 17:48:32 not necessarily in that order ;) 11/17/04 17:49:09 the cupsd crash should be fixed in newer releases, and that patch takes care of cleaning out the tdb 11/17/04 17:50:21 i might try to figure out why samba needs an excessive length of time to really delete the printer share after i find some lunch myself 11/17/04 18:02:46 Abood has left channel #samba-technical because () 11/17/04 18:04:49 stevef (~stevef@pixpat.austin.ibm.com) has joined channel #samba-technical 11/17/04 18:20:18 Rwf: 3,0.8? 11/17/04 18:21:05 they'll probably apply 11/17/04 18:21:18 those are svn diffs against latest SAMBA_3_0 11/17/04 18:21:57 Rwf: Ok. Saw the URL. 11/17/04 18:23:12 samba-dns-name-param.patch is more of an experiment, and probably won't be the final nor correct solution to that particular problem 11/17/04 18:23:22 Rwf: printer-make-and-model is not there at all? 11/17/04 18:23:41 hm? 11/17/04 18:24:08 it was only being used as a default printer comment when cups didn't supply a description 11/17/04 18:27:30 Rwf: deleting a printer's security descriptor should have an own function and it would be nice to have it accessible from a command-line-tool. 11/17/04 18:27:57 should be doable 11/17/04 18:27:57 In samba-delete-printer-cleanliness.patch you call tdb_delete() twice. 11/17/04 18:28:07 lieschen: yes, on two different keys :) 11/17/04 18:28:17 i just noticed something else that that didnt fix 11/17/04 18:28:22 AD listings aren't being unpublished 11/17/04 18:29:17 Rwf: I should have applied your patch ... 11/17/04 18:29:29 hold off on -cleanliness for now 11/17/04 18:29:39 i' 11/17/04 18:29:53 Rwf: And the fix in smbd/service.c is to remove_stale_printers()? That's my fault. :( 11/17/04 18:30:06 er i'll redo some of that 11/17/04 18:30:17 lieschen: it's still not correct, AD listings need to be addressed as well 11/17/04 18:31:35 gd: i see what you mean, regarding a commandline tool 11/17/04 18:33:22 i really need to eat something, bbl 11/17/04 18:33:32 Signoff: porkpie (Remote closed the connection) 11/17/04 18:41:05 gd? 11/17/04 18:41:24 yes. 11/17/04 18:41:42 rap shareenum is a segfault... 11/17/04 18:41:53 exactly :) 11/17/04 18:42:01 You already found it? 11/17/04 18:42:05 just who invented it? 11/17/04 18:42:12 no, not yet. 11/17/04 18:43:35 If you don't have printers in your smb.conf, it works... :-) 11/17/04 18:44:28 I think it's in loadparm.c 11/17/04 18:47:14 Hmmm. Or not? That looks innocent. 11/17/04 18:47:19 Dig further :-) 11/17/04 18:48:50 Ok, leaving from the net til tomorrow. 11/17/04 18:49:21 Hmm. My english is getting worse.. Really have to leave :-) 11/17/04 18:49:24 bye 11/17/04 18:49:25 Signoff: vl ("Leaving") 11/17/04 19:13:16 dort has left channel #samba-technical because ("Leaving") 11/17/04 19:23:02 Signoff: waider ("outta here") 11/17/04 20:02:25 qyang (26219ff2b1@198.147.128.71) has joined channel #samba-technical 11/17/04 20:04:35 idra (~simo@81-174-11-173.f5.ngi.it) has joined channel #samba-technical 11/17/04 20:04:35 Mode change "+o idra" on channel #samba-technical by ChanServ 11/17/04 20:04:47 hi * 11/17/04 20:08:33 hi idra. 11/17/04 20:09:05 deryck, today I looked at some mirrors, and seem hu.samba.org is not responding correctly 11/17/04 20:09:08 hi 11/17/04 20:09:15 deryck, there was also another one looping 11/17/04 20:09:20 jfm, hello, how are you? 11/17/04 20:09:27 idra: what is the trick to make a good polenta ? 11/17/04 20:09:59 idra: I was aware of the looping one. I'll take a look at hu.samba.org. 11/17/04 20:09:59 idra: I'm fine 11/17/04 20:10:16 idra: thanks for helping with the checking. 11/17/04 20:11:00 jfm, don't ask me, I only eat it sometime :-P 11/17/04 20:11:16 deryck, no porblems 11/17/04 20:11:34 deryck, do you have statistic on how much trafic a mirror usually get ? 11/17/04 20:11:41 idra: just checked hu.samba.org... our scripts should catch that one when they run again. Not sure how often the mirror update script runs. 11/17/04 20:11:52 idra: no, sorry. 11/17/04 20:12:06 I have stats on our traffic, but not the mirrors. 11/17/04 20:12:17 deryck, yeah, and I have mine, now :-) 11/17/04 20:12:26 :-) 11/17/04 20:12:30 deryck, I think yesterday I had near 1GB traffic 11/17/04 20:12:36 what's your mirror url? 11/17/04 20:12:40 wow. 11/17/04 20:12:41 it.samba.org 11/17/04 20:12:54 ah, okay. that's you now. 11/17/04 20:14:30 Signoff: poseiden_ ("leaving") 11/17/04 21:04:47 tridge (~tridge@203-173-8-192.perm.iinet.net.au) has joined channel #samba-technical 11/17/04 21:04:47 Mode change "+o tridge" on channel #samba-technical by ChanServ 11/17/04 21:10:13 misty has left channel #samba-technical because () 11/17/04 21:16:56 hello tridge 11/17/04 21:17:11 hi 11/17/04 21:22:44 hi Tridge, Simo 11/17/04 21:41:34 "smbclient //machine/share -U user password -N" does not behave the same as "smbclient //machine/share -U user password", even though the man page implies it should. Doc bug or code bug? 11/17/04 21:43:55 dajt: the docs don't document "smbclient //machine/share -U user password" as a valid invocation 11/17/04 21:44:12 (you have to specify the options after the password) 11/17/04 21:46:00 ctrlsoft: Heh. But "smbclient //machine/share password -U user -N" acts differently than "smbclient //machine/share password -U user" so that isn't the problem. :-( 11/17/04 21:46:44 The man page says that including a password on the cmd line is the equiv of saying -N 11/17/04 21:47:30 -N means "don 11/17/04 21:47:35 't ask for password" 11/17/04 21:47:44 * ctrlsoft guesses the docs might need a fix here.. 11/17/04 21:48:11 Yeah. It seems to implement "Don't use a password", which is somewhat different. 11/17/04 21:48:29 ctrlsoft: Yes. I think I stumbled once ... 11/17/04 21:48:54 dajt: You know https://bugzilla.Samba.org/ ? 11/17/04 21:49:10 * ctrlsoft cleans the dust off his old Samba 3 tree and takes a look at client/client.c 11/17/04 21:49:16 lieschen: far too well :-) 11/17/04 21:50:07 dajt: From my experience it's best to document it. That's how I do it even with stuff I've fixed. 11/17/04 21:51:05 Before going to bugzilla I wanted to know if the doc is wrong, the code is wrong, or both. :-) 11/17/04 21:52:07 dajt: Yes. But don't miss it. ;) 11/17/04 21:52:53 Not to worry. I need to point https://bugzilla.redhat.com/beta2/show_bug.cgi?id=106483 at it :-) 11/17/04 21:54:33 dajt: it's a code bug 11/17/04 21:54:41 dajt: Are you the rh Samba mainatiner? 11/17/04 21:54:48 yep. 11/17/04 21:54:58 Good to know. 11/17/04 21:55:17 Do you have an extra package for the cifs mount helper? 11/17/04 21:56:21 /sbin/mount.cifs ? I include it in the samba-client package. 11/17/04 21:56:36 lieschen, do you have any time now? 11/17/04 21:56:49 idra: Yes. 11/17/04 21:57:04 lieschen, about sles9 .. 11/17/04 21:57:31 lieschen, I have some "bugs" in the yast module that distrub me, may be you are interested in 11/17/04 21:57:53 idra: Just send me a mail and I file bugs. 11/17/04 21:58:11 lieschen, let's first see if they are actually bugs :) 11/17/04 21:58:36 lieschen, one, easy, you can add trusting domains but not trusted domains, why ? 11/17/04 21:59:33 lieschen, second, I read the script that's provided to add machines into a sles9 pased samba PDC ... and well it is ... hackish 11/17/04 22:00:11 idra: For the YaST Samba server moduke I file a bug. 11/17/04 22:00:25 lieschen, you grep the binary file secrets.tdb in search for the password, but that file format may change in future, wouldn't it be better to have a binary provided by samba that give you back the password if you are root ? 11/17/04 22:01:09 idra: Yes. Point me to the binary. And we'll use it. ;) 11/17/04 22:01:12 lieschen, last and most important, do you know that you cannot chose the SID of groups and users? they are forced by the old algorithmic mapping inside the interface 11/17/04 22:01:29 lieschen, I may intorduce it as i think it is needed :) 11/17/04 22:01:49 I'M not sure how to handle the password stuff. 11/17/04 22:02:08 idra: user and group mapping -> already a bug. 11/17/04 22:02:27 lieschen, the biggest problem is that you cannot make the groups Domain Admins, Domain Users and Domain Guest with their special SID, as the yast user&group interface will change the SID back anyt time you manage those groups 11/17/04 22:02:34 lieschen, ah ok fine 11/17/04 22:02:57 lieschen, well just pay attention at subtle tdb format changes :-) 11/17/04 22:03:14 as the hack you use is smart but may breack way too easily 11/17/04 22:03:42 lieschen, anyway the SID mapping is really bad, usrmgr.exe do not work without the right SIDs 11/17/04 22:05:04 lieschen, beside that the system works well enough :) 11/17/04 22:06:28 idra: trusted domain stuff filed. 11/17/04 22:11:29 idra: Thx for the feedback. 11/17/04 22:12:33 dpang (~dpang@209-101-18-214.pao-e100.cust.gw.epoch.net) has joined channel #samba-technical 11/17/04 22:15:32 dajt: Here we also have /sbin/mount.cifs in the client package. But bjacke convinced me to move it in an extra package. 11/17/04 22:16:30 dajt: With at least as possible dependencies. 11/17/04 22:16:50 lieschen, I'm planning to do the same for debian packages ... 11/17/04 22:17:25 dajt: I had also a short talk with Steve about this and he appreciated it. 11/17/04 22:17:43 So what I would like to see is a common package name for this piece. 11/17/04 22:18:18 cifs-mount; cifs-utils (even if it is only one) 11/17/04 22:18:24 mount.cifs 11/17/04 22:18:45 Hm. Are dots mossible an a package name? 11/17/04 22:18:54 s/an/in/ 11/17/04 22:19:06 s/moss/poss/ 11/17/04 22:19:14 no 11/17/04 22:19:35 cifsvfs-utils 11/17/04 22:19:37 ? 11/17/04 22:20:40 Björn (bjacke) currently uses samba-cifsmount. 11/17/04 22:20:50 nice name 11/17/04 22:21:00 So how about cifsmount? 11/17/04 22:21:04 cifsmount is also a good name 11/17/04 22:21:07 cifsmount is also a good name:) 11/17/04 22:21:10 :) 11/17/04 22:21:15 lieschen, I like it :) 11/17/04 22:21:23 idra: I saw it. ;) 11/17/04 22:21:53 lieschen, mistype on keyboard sorry :) 11/17/04 22:21:59 openoffice.org is a package name, so I don't see why not. 11/17/04 22:22:24 dajt: I 11/17/04 22:22:43 * dajt doesn't like samba-cifsmount, since part of the idea is to make it clear that mount.cifs is not part of samba. 11/17/04 22:22:43 uhmm right 11/17/04 22:22:48 dajt: I'm not sure if this is a limitation introduced by the brain of a colleague. ;) 11/17/04 22:23:17 dajt: Yes. And that there is _no_ dependency to any other Samb apckage. 11/17/04 22:24:31 dajt/ idra: I would like to see a common name across the vendors. 11/17/04 22:24:40 Signoff: [cat] ("leaving") 11/17/04 22:24:49 This makes it easier for the users. 11/17/04 22:29:34 Il ike cifsmount, but you should also ask to Eloy and Vorlon for debian unstable packages ... 11/17/04 22:32:36 wow, apparently i need to be a lot more careful when cleaning up after deleted printers 11/17/04 22:33:03 cupsd crashed again, and pcap_printername_ok returned false for every printer 11/17/04 22:33:09 and wiped all of them from the db 11/17/04 22:33:11 oops :) 11/17/04 22:33:21 ouch 11/17/04 22:33:27 Any name is fine with me, as long as it doesn't have "samba" in it. :-) 11/17/04 22:33:30 you need to validate the response from cups :) 11/17/04 22:33:32 yeah 11/17/04 22:33:51 i think i'm going to change cups_printername_ok to return true when it fails 11/17/04 22:34:08 dajt: Fine. I'll check if ther is any limitation in the suse build system and inform you and idra. 11/17/04 22:34:34 lieschen: Sure. If I'm not here, you can find me in #fedora-devel 11/17/04 22:34:58 dajt: Ah. foo.bar works. So mount.cifs should also ... :) 11/17/04 22:35:15 dajt: Ok. 11/17/04 22:36:42 vance (~vance@Toronto-HSE-ppp3867230.sympatico.ca) has joined channel #samba-technical 11/17/04 22:36:42 Mode change "+o vance" on channel #samba-technical by ChanServ 11/17/04 22:38:20 'morning vance 11/17/04 22:38:31 'allo 11/17/04 22:38:47 hi abartlet, vance. 11/17/04 22:39:04 * vance tips his hat 11/17/04 22:39:49 nemisus (~nem@203-173-17-113.dyn.iinet.net.au) has joined channel #samba-technical 11/17/04 22:41:56 Hi vance. 11/17/04 22:42:12 greetings 11/17/04 22:42:27 vance: I didn't find any time to work on the build farm stuff we're talking about at the cifs conference. I have to excuse me. 11/17/04 22:42:54 'sok - I haven't found much time since then either :) 11/17/04 22:43:26 lieschen: if you do find time, just let me know and I'll set up accounts for you 11/17/04 22:48:12 where's jerry been hiding lately? 11/17/04 22:50:06 pcap_printername_ok and friends return False on every failure case 11/17/04 22:50:31 which is arguably not correct behavior, since that return value is used in deleting printers, among other things 11/17/04 22:50:42 (it just didn't matter until i started trying to make deletions more aggressive) 11/17/04 22:51:29 vance: Thx. It's again on my record. As I have a good week vacation, this time it might happen. 11/17/04 22:51:48 :) 11/17/04 22:52:09 don't spend your entire vacation on it, though :) 11/17/04 22:52:15 Signoff: idra ("Leaving") 11/17/04 22:52:31 vance: No. I travell to Goettingen again. 11/17/04 22:53:10 vance: And there I have my special social correction unit. 11/17/04 22:53:20 :) 11/17/04 22:55:21 Rwf: jerry is at a conference this week. 11/17/04 22:55:38 ahh 11/17/04 22:58:35 Rwf: You should better have a look at the Team calendar. ;) 11/17/04 23:03:11 Signoff: dpang (Read error: 60 (Operation timed out)) 11/17/04 23:05:46 * deryck likes to hear the Team calendar mentioned. 11/17/04 23:08:42 All done for today. 11/17/04 23:12:40 ab (~ab@ppp83-237-19-212.pppoe.mtu-net.ru) has joined channel #samba-technical 11/17/04 23:14:53 tridge: Have you seen my reply to your email? 11/17/04 23:15:08 ctrlsoft: yes 11/17/04 23:15:41 why use constructors if we are going to have code that doesn't use them as well 11/17/04 23:15:49 that just means duplication? 11/17/04 23:16:23 it also means that nearly all team members will not be testing without constructors 11/17/04 23:16:29 so that code will surely break 11/17/04 23:17:00 Using constructors has two advantages: no need to recompile all the .o files for the binaries when adding a subsystem/module 11/17/04 23:17:22 and no need for calling a whole bunch of init functions for 3rd-party developers using Samba libs 11/17/04 23:18:05 so more implicit magic? 11/17/04 23:18:09 i understand the 2nd point, but not the first 11/17/04 23:18:32 tridge: currently, if you add a subsystem (say named foo with an init function foo_init) 11/17/04 23:18:32 why would we need to do a clean build without constructors? 11/17/04 23:18:54 because smb_build_h changes (gets a new init function in it's list) 11/17/04 23:19:15 so torture/torture.o, smbd/rewrite.o use in invalid define 11/17/04 23:19:18 ok, so we'd need to recompile only modules with main() in them, yes? 11/17/04 23:19:28 correct 11/17/04 23:20:00 ctrlsoft: how is the ordering determined with constructors? 11/17/04 23:20:10 abartlet: that was my next question! 11/17/04 23:20:22 * abartlet reads tridge's mind :-) 11/17/04 23:20:52 twinheads 11/17/04 23:20:52 abartlet: afaik it's in the order of the dependencies, so if foo.a depends on bar.a, bar.a's init functions are called first, then foo.a's functions 11/17/04 23:20:56 though I'd have to check 11/17/04 23:20:58 also, how do we make constructors run _after_ we parse the command line? (we need that) 11/17/04 23:21:12 that won 11/17/04 23:21:17 't work.. 11/17/04 23:21:21 why do we need it? 11/17/04 23:21:27 several reasons ... 11/17/04 23:21:35 killing krb5 :-) 11/17/04 23:21:42 for example, the talloc leak code is enabled by command line option 11/17/04 23:21:55 so, any constructor that allocates memory will not be seen by talloc leak code 11/17/04 23:22:04 ah, right.. 11/17/04 23:22:16 also krb5 as andrew says 11/17/04 23:22:33 and any other register fn that looks at command line or smb.conf options (lots do) 11/17/04 23:23:05 so how about optional constructor support for shared libraries that will be used by 3rd party developers? 11/17/04 23:23:31 I now understand they're now very useful for internal stuff... 11/17/04 23:23:52 s/now/not/ 11/17/04 23:24:25 ctrlsoft: I think users of shared libs would want even more control 11/17/04 23:24:33 yes, possibly when we start designing libs for external use we could have a single constructor per lib that calls all the necessary fns 11/17/04 23:24:50 I can imagine that having some function run almost arbitarily would be nasty - i prefer tridge's single constructor 11/17/04 23:25:12 ctrlsoft: but we are a long way from that 11/17/04 23:25:18 tridge: Yeah, I understand 11/17/04 23:25:25 ctrlsoft: imagine the following scenario with libs based on our current code 11/17/04 23:25:29 ctrlsoft: sorry to pull your plan to bits... 11/17/04 23:25:32 say emacs wants smb support 11/17/04 23:25:41 so, it links to libsambastuff.so 11/17/04 23:25:49 then when emacs starts the constructor gets called 11/17/04 23:26:01 and that reads smb.conf, and spews err msgs 11/17/04 23:26:14 but the user hasn't even asked emacs to do anything with smb yet 11/17/04 23:26:20 tridge: the idea would be to use the init functions for registering only.. 11/17/04 23:26:36 ctrlsoft: yeah, but the registration fns call lp_*() fns 11/17/04 23:26:37 e.g. if you want to use the mgmt, epmapper and samr pipes 11/17/04 23:26:49 you link against msrpc_epmapper, msrpc_mgmt and msrpc_samr 11/17/04 23:27:13 ctrlsoft: ok, where does it get its netbios name from? 11/17/04 23:27:17 tridge: oh, ok - I thought most of them just did a DLIST_ADD... 11/17/04 23:27:31 even DLIST_ADD() usually implies a talloc 11/17/04 23:27:50 i'm not totally opposed to constructors, but perhaps its premature 11/17/04 23:28:00 we need to design some good library interfaces first 11/17/04 23:28:16 interfaces that allow application developers to control needed params without smb.conf 11/17/04 23:28:25 yeah, I can see now why 11/17/04 23:28:50 samba4 is in much better shape than samba3 for becoming a library source, but its not there yet 11/17/04 23:28:59 Signoff: _Kaa (Client Quit) 11/17/04 23:29:08 someone needs to put some serious thought and effort into doing libs right 11/17/04 23:29:26 otherwise samba libs will get the same bad reputation as so many other libs out there :) 11/17/04 23:29:35 :-) 11/17/04 23:30:52 hmm, we only have about 20 lp_*() calls in librpc now 11/17/04 23:31:13 so its quite doable to remove those and instead have a parameter setup api 11/17/04 23:31:24 there are a few more (about 60) in libcli 11/17/04 23:32:06 actually, its less than that 11/17/04 23:32:15 i was grepping for lp_ and hit "help_" 11/17/04 23:32:33 its only 4 left in librpc 11/17/04 23:32:35 tridge: I'm seeing only lp_netbios_name() and lp_ncalrpc_path() in librpc/ 11/17/04 23:32:47 ctrlsoft: yep 11/17/04 23:32:50 oh, and lp_lockdir 11/17/04 23:32:56 tridge: hi 11/17/04 23:33:04 tridge: what do i have to do to reset tsums? 11/17/04 23:33:12 i thought tsums -u /etc/group would do it 11/17/04 23:33:15 but apparently not? 11/17/04 23:36:14 eborn (~eborn@linda.fambus.nl) has joined channel #samba-technical 11/17/04 23:36:19 tridge: what did you think about using dlsym() for checking if a subsystem is present? 11/17/04 23:37:05 (as opposed to register_subsystem() + register_module()?) 11/17/04 23:37:18 ctrlsoft: well, that only works on platforms with dlsym 11/17/04 23:37:33 ctrlsoft: if we use string names for subsystems then we can use a fn 11/17/04 23:37:42 tridge: that's not a problem, because those are the only platforms that can actually load plugins 11/17/04 23:37:46 ctrlsoft: like is_registered(const char *name); 11/17/04 23:38:06 yes, but it would be nice for the same code to work builtin or separate 11/17/04 23:38:31 true, yeah, we could have a function/macro that uses dlsym() or just returns true if dlsym() is not available.. 11/17/04 23:38:39 no .... 11/17/04 23:38:44 thats not what I meant 11/17/04 23:38:51 :-) 11/17/04 23:38:53 i meant to make register_subsystem() take a string name 11/17/04 23:39:04 and there be a global list of registered subsystems 11/17/04 23:39:15 so is_registered() would work on all systems 11/17/04 23:39:53 i'm trying to keep as much code as possible the same in the module/non-module cases 11/17/04 23:39:56 tridge: what about using AcceleratedX code for loading objects like Xorg does? 11/17/04 23:40:05 tridge: that would mean that subsystems would have to be registered before modules for those subsystems can be registered 11/17/04 23:40:10 ab: i don't know anything about it 11/17/04 23:40:35 ctrlsoft: yes, it means an ordering 11/17/04 23:40:54 ctrlsoft: but we already need ordering for other reasons (like above with talloc and krb5) 11/17/04 23:41:03 tridge: AcceleratedX has had a code to load modules even on dynamically-disabled platforms. They donated that to XFree86 some time ago. It is in use since XFree86 4.0 11/17/04 23:41:05 so there would be problems with circular dependencies between subsystems (like LIBRPC and LIBSMB) 11/17/04 23:41:07 ctrlsoft: so a sort in the build system perl should take care of that 11/17/04 23:41:21 ab: but how does it _work_ 11/17/04 23:41:39 tridge: I can dig up an article if you interested 11/17/04 23:41:49 ab: i know there are packages to do this, but they tend to force lowest common denoninator behaviour 11/17/04 23:42:36 ab: we don't desperately need modules in samba, so i'm more than happy for them to be only available on sane platforms 11/17/04 23:42:39 _Kaa (XtXtX@lemahp6.epfl.ch) has joined channel #samba-technical 11/17/04 23:42:52 everybody runs linux anyway! ;-) 11/17/04 23:42:57 tridge: http://www.usenix.org/publications/library/proceedings/als00/2000papers/papers/full_papers/hohndel/hohndel_html/node3.html 11/17/04 23:43:11 tridge: it was MetroLink, actually, I messed up the name 11/17/04 23:43:22 ab: if you want this to be used, i'd rather you read it, evaluate it, and argue why its good 11/17/04 23:43:39 ab: i'd prefer to be coding pvfs :) 11/17/04 23:43:53 tridge: so, what's left in pvfs? 11/17/04 23:44:04 abartlet: i was going to do ACLs in xattrs today 11/17/04 23:44:13 tridge: I'll get back with it if it worth. Don't get away from pvfs :-) 11/17/04 23:44:22 mixed with tpot's kernel module, or? 11/17/04 23:44:29 abartlet: not yet 11/17/04 23:44:30 or just storing, but not enforcing them? 11/17/04 23:44:43 abartlet: yes, first storing, then eval in user space 11/17/04 23:44:48 abartlet: then later integrate with lsm 11/17/04 23:44:49 tridge: I ran into trouble with that before. For example, LIBRPC depends on GENSEC which uses GENSEC_KRB5 which uses GEN_NDR_KRB5PAC which uses LIBRPC which ... 11/17/04 23:45:24 Signoff: _Kaa (Client Quit) 11/17/04 23:45:26 ctrlsoft: right, and this needs to be fixed or at least made to work - it doesn't go away by using linker magic :) 11/17/04 23:45:31 heh.. i'd rather be working on anything but printing ;) 11/17/04 23:48:00 tridge: I think the right behaviour here is to initialise the modules before their subsystem, which is why I would like to replace register_{subsystem,module}() with have_function() 11/17/04 23:48:37 what would have_function() do? 11/17/04 23:49:01 do the dlsym() trick to see if the specified function is resolved at the moment 11/17/04 23:49:23 and return TRUE if dlsym is not available (in which case the function just /has/ to be available) 11/17/04 23:49:43 and if its not available, what does the caller do? 11/17/04 23:49:52 doesn't run the init function 11/17/04 23:49:57 s/init/register/ 11/17/04 23:50:06 doesn't run its own init fn, or the target one? 11/17/04 23:50:21 sorry, I meant the register function. So we could have something like: 11/17/04 23:50:31 NTSTATUS dcerpc_epmapper_init() 11/17/04 23:50:51 { if (has_function("dcerpc_register")) { dcerpc_register(&my_table); } } 11/17/04 23:52:02 so what happens then? 11/17/04 23:52:12 we just end up with no epmapper capability? 11/17/04 23:52:26 indeed. which was the case in the old situation as well 11/17/04 23:52:36 or the dcerpc_epmapper_init() function could return NT_STATUS_NOT_SUPPORTED or something 11/17/04 23:52:42 if has_function() returns FALSE 11/17/04 23:53:06 so every app would need to check for not-supported? 11/17/04 23:53:19 no, every module 11/17/04 23:53:32 ok, so what does the application do? 11/17/04 23:53:41 it will get the error too 11/17/04 23:53:45 nothing, just call a list of init functions - like it does now 11/17/04 23:53:59 yes, but when it tries to do some workk it will fail 11/17/04 23:54:38 the generated *_init_subsystems macros check for the return value of the init functions 11/17/04 23:54:47 and exit(1) if it is a STATUS_ERR 11/17/04 23:55:06 so emacs exits if the link order ain't right? 11/17/04 23:55:10 (the init functions currently print error messages if they actually do return a !NT_STATUS_OK) 11/17/04 23:55:20 tridge: emacs doesn't have a list of init functions 11/17/04 23:55:20 and where do they print it? 11/17/04 23:55:28 stdout could be some valuable file 11/17/04 23:55:48 tridge: they use DEBUG() - it's just like in the current tree 11/17/04 23:56:05 think some more ... 11/17/04 23:56:09 where does DEBUG() go? 11/17/04 23:56:15 it needs to read smb.conf 11/17/04 23:56:22 or default to stdout 11/17/04 23:56:38 so, some server app gets stuff written to file descriptor 1 11/17/04 23:56:50 which could easily be some valuable file or database 11/17/04 23:57:05 You have a point there... 11/17/04 23:57:17 however, that's already broken at the moment (and always has been...) 11/17/04 23:57:20 what you seem to be doing is reinventing ld.so 11/17/04 23:57:58 when a application loads a shared lib it can say what should be done for symbol resoolution 11/17/04 23:58:08 it can choose "resolve all now" or "resolve at runtime" 11/17/04 23:58:17 so, the application chooses how it will deal wiith this 11/17/04 23:58:22 instead, you are forcing one way 11/17/04 23:58:49 so, to _really_ solve this we need to do one of two things: 11/17/04 23:59:00 1) use a single shared lib, so no problem at all 11/17/04 23:59:12 2) document the necessary load order of our shared libs 11/17/04 23:59:32 in the case of (2) the code needs to be able to init in the right order 11/17/04 23:59:39 ok, we seem to be talking about two different things :-) 11/17/04 23:59:58 if two libs are circular deps, then we either need to break the dependency, or merge them into one lib