openldap-bugs June 2007

openldap-bugs@openldap.org

35 participants
170 discussions

Re: (ITS#5022) ldappasswd crash consumer slapd with some loglevels
by gao＠schrodinger.com 19 Jun '07

19 Jun '07

Pierangelo Masarati wrote: >> Sorry about the broken link. Here is the correct one: >> >> ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_20070618.ext >> >> >>>> When following command against a consumer slapd, it will crash slapd of the >>>> consumer when loglevel is set to any other value other than 1 or -1 on the >>>> consumer: >>>> >>>> ldappasswd -v -H ldap://consumer -D "uid=joe,ou=people,dc=example,dc=com" -W -S >>>> -x -A >>>> >>>> The crash happens to at least to loglevel 0, 256, 512. When loglevel is set to 1 >>>> or "-1", then no crash is experienced. In addition, when run consumer slapd >>>> manually as front process, >>>> >>>> >>> ^^^ what does this mean? >>> >>> >> If I start slapd manually as following commands: >> >> /usr/lib/openldap/slapd -d 0 -u ldap -g ldap -h 'ldap:// ldaps' >> /usr/lib/openldap/slapd -d 256 -u ldap -g ldap -h 'ldap:// ldaps' >> /usr/lib/openldap/slapd -d 512 -u ldap -g ldap -h 'ldap:// ldaps' >> >> Then I did not see the problem as I would when setting loglevel to 0, >> 256, 512 in slapd.conf and start slapd from /etc/init.d/slapd, which >> launch the daemon with the same options. When starting slapd from >> /etc/init.d/slapd as background daemon process, slapd will write to >> /var/log/ldap.log and /var/log/message. It seems that the difference >> exists between writing ldap logs to a file (slapd runs in background) or >> writing to console (slapd runs in foreground as started manually). >> >> >>> >>> >>>> then all debug level works without problem. >>>> >>>> >>> OpenLDAP re23 (in practice, 2.3.36 just released) doesn't show anything >>> like that. I note that differences related to the debug level usually >>> mean that a NULL or an invalid pointer is passed to a *printf(3) routine >>> on those systems that do not tolerate it (e.g. Solaris). But usually >>> the bug disappears when __decreasing__ the log level, while -1 means all. >>> >>> p. >>> >>> >> I would expect writing less log should help avoiding problem. In this >> case, somehow logging must be set at certain or above a level. Below >> that it will trigger problem. While I was trying to get pieces working, >> I always set loglevel to 1 or -1. So I did not realize this problem >> until very late. The error I got as a result of attempting change >> password is "can't contact LDAP server" which is very confusing (might >> be because slapd already crashed before serving last request). >> >> I am happy to run more tests and provide further information as needed. >> > > If slapd crashes, the only really useful info is a core dump, as > indicated in the FAQ I already pointed you to (that's basically why I > pointed you there: there's no point in always repeating instructions for > bug tracing, that's what the FAQ is intended for; apparently, it's > totally ignored). > > Not entirely. I thought system environment information might be useful too. Anyway. Here is the coredump. Let me know if this is enough. ==================================================================== # gdb /tmp/debug/usr/lib/openldap/slapd GNU gdb 6.6 Copyright (C) 2006 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "i686-pc-linux-gnu"... Using host libthread_db library "/lib/libthread_db.so.1". (gdb) run -d 0 -u ldap -g ldap -h 'ldap:// ldaps://' Starting program: /tmp/debug/usr/lib/openldap/slapd -d 0 -u ldap -g ldap -h 'ldap:// ldaps://' [Thread debugging using libthread_db enabled] [New Thread -1213843792 (LWP 7706)] [New Thread -1240826992 (LWP 7709)] [New Thread -1249219696 (LWP 7710)] [New Thread -1257612400 (LWP 7711)] slapd: result.c:364: send_ldap_response: Assertion `rs->sr_err != 0x0a' failed. Program received signal SIGABRT, Aborted. [Switching to Thread -1249219696 (LWP 7710)] 0xbfffe402 in __kernel_vsyscall () (gdb) bt full #0 0xbfffe402 in __kernel_vsyscall () No symbol table info available. #1 0xb7a8d429 in raise () from /lib/libc.so.6 No symbol table info available. #2 0xb7a8e9d1 in abort () from /lib/libc.so.6 No symbol table info available. #3 0xb7a86e51 in __assert_fail () from /lib/libc.so.6 No symbol table info available. #4 0x080809c5 in send_ldap_response () No symbol table info available. #5 0x0000000a in ?? () No symbol table info available. #6 0x08150731 in ?? () No symbol table info available. #7 0x08150731 in ?? () No symbol table info available. #8 0xb58a61b4 in ?? () No symbol table info available. #9 0xb58a5cac in ?? () No symbol table info available. #10 0x00000000 in ?? () No symbol table info available. ===================================================================== Simon

1 0

Re: (ITS#4943) tpool.c pause vs. finish
by h.b.furuseth＠usit.uio.no 19 Jun '07

19 Jun '07

hyc(a)symas.com writes: >h.b.furuseth(a)usit.uio.no wrote: >> Eh? Yes, that's what I was talking about. But I'm not sure what you >> mean with them exiting immediately - or how they could. Also they stay >> around until there are no pending requests. A pause request could >> arrive more or less at any time, I presume. (I don't know if there is a >> pause while the max threads number is being modified.) > > Yes. As stated back in the distant past, cn=config fully serializes > all write operations. The pool is paused for each write op. Oh, right. > When the op completes and the pool is resumed, the condition variable > is broadcasted, so all the other threads will wake up, and the first > to do so will see that they're above the max and exit, before any > other tasks are started. No, they won't exit pool_wrapper() while there are pending requests, but if that is a bug it is easy to fix. And by that description the wait I inserted after context_reset() in pool_wrapper() should not be there, since it releases ltp_mutex before the ID is removed from thread_keys[]. However another thread could be waiting in pool_pause() and start another pause. And the thread which did pool_resume() will be active and could call pool_pause() again (but I think it won't in the same LDAP operation with the current bconfig.c code). So if pool_wrapper() does not go active for context_reset() we must ensure that a pause is not started while a thread in pool_wrapper() is dying. That leaves thread_keys[] protected by either a pause or one of two mutexes, depending on circumstances. I'll sleep on it before trying to wrap my head around that one:-) >> (...) We did have code which compared thread IDs with '==' for >> quite some time. Do you remember if anyone complained, or if the >> switch to using ldap_pvt_thread_equal() was a "just in case" fix? > > Yes, I needed to use pthread_equal() on OS/390 since their thread type > was a structure. Ouch. That kills my tls.c patch too. I so hoped not to hear of an example... >> BTW I don't know if the unsigned long needs to be the thread ID or just >> _some_ unique integer. A question to openssl-users(a)openssl.org got no >> answer. > > It just needs to be a unique integer. I've already had this > conversation with the OpenSSL folks... Got a reference to that which I can beat them with? (To get them to update their doc if nothing else.) >>>>>>> The best fix may be to just use the address of the thread's >>>>>>> stack as the thread ID. That's actually perfect for our purpose. >> (...) >> Hmm. I asked the wrong question, should have read some man pages first. >> Still, thr_cthreads.c & thr_lwp.c don't use LDAP_PVT_THREAD_STACK_SIZE. >> man pthread_attr_getstacksize says it sets the _minimum_ stack size, but >> perhaps several M is large enough that one can expect it won't get >> multiples of that. > > That kinda implies that no one has been using the cthreads or lwp support > recently, since I doubt their default stack size would work with back-bdb's > filterindex demands. Um, okay... could remove them from configure and tell anyone who needs them to shout. How about pool_context() called from subsystems which create their own threads (like Perl code or plugins)? I guess it's just that I don't see I why using a private, hopefully-OS-independent hack is an improvement over OS-provided thread-specific data (when that is supported). Any risk that there are systems around which support non-contiguous stacks? (Allocate a new stack chunk if the current one is used up, and chain it to the old one.) I don't have a clue how common they are or were, nor of the arguments for/against implementing them. -- Regards, Hallvard

1 0

Re: (ITS#4943) tpool.c pause vs. finish
by hyc＠symas.com 19 Jun '07

19 Jun '07

h.b.furuseth(a)usit.uio.no wrote: > hyc(a)symas.com writes: >> h.b.furuseth(a)usit.uio.no wrote: >>> It's called from pool_wrapper(). pool_wrapper() could go active so >>> there can be no pause while that context_reset() call is running. >> Oh, ok. This is pretty much a non-issue because threads in the pool stay >> live until slapd shuts down. The one exception is for a cn=config modify >> that decreases the size of the pool to below the number of currently >> active threads, in which case some number of threads will exit >> immediately. I don't see any potential for conflict here. > > Eh? Yes, that's what I was talking about. But I'm not sure what you > mean with them exiting immediately - or how they could. Also they stay > around until there are no pending requests. A pause request could > arrive more or less at any time, I presume. (I don't know if there is a > pause while the max threads number is being modified.) Yes. As stated back in the distant past, cn=config fully serializes all write operations. The pool is paused for each write op. When the op completes and the pool is resumed, the condition variable is broadcasted, so all the other threads will wake up, and the first to do so will see that they're above the max and exit, before any other tasks are started. >> Casting to an int only works with simple types, the point of using the >> hash was to provide a solution that works for both simple and structured >> thread IDs. > > Yes, except it's not guaranteed to work. Which is why I'm thinking it > may be better to fail to compile and provide a fix if/when someone > complains. We did have code which compared thread IDs with '==' for > quite some time. Do you remember if anyone complained, or if the > switch to using ldap_pvt_thread_equal() was a "just in case" fix? Yes, I needed to use pthread_equal() on OS/390 since their thread type was a structure. > However note that we do > #define ldap_int_thread_equal(a, b) ((a) == (b)) > for all our supported thread implementations except pthreads, and > for pthreads we can test pthread_<set/get>specific() just fine > (as a replacement for thread_keys[]). > > OTOH, '==' also works for pointers that are wider than any integer type > (and for floating-point thread IDs:-) so I guess I'm arguing against my > own suggestion on that one if we add pthread_<set/get>specific() > support. > >> If you want to #ifdef the code so the two cases are handled >> differently, then just get rid of TID_HASH for the integral case. > > No, sould still hash to avoid clustering. > >> I'm still leary of the pthread TLS support, > > Me too. But the old code was broken on real-life systems I have access > to, not just on theoretical ones. Still waiting to hear from Kurt if my > fix to the first tls.c change fixed it for FreeBSD. > >> and making that change will >> require us to update all of the other thread package support, some of >> which I don't have access to build/test any more (e.g. SunOS4 LWP). > > What do you mean? The TLS problem is to extract an unsigned long from a > ldap_pvt_thread_t. That needs no support in the thr_*.c files, only to > know what kind of type ldap_pvt_thread_t is. Once configure knows what > ldap_pvt_thread_t will be typedeffed to it can test if that's a pointer, > integer or something worse - and whether it's too wide for a long. > > BTW I don't know if the unsigned long needs to be the thread ID or just > _some_ unique integer. A question to openssl-users(a)openssl.org got no > answer. It just needs to be a unique integer. I've already had this conversation with the OpenSSL folks... >>>>>> The best fix may be to just use the address of the thread's stack as the >>>>>> thread ID. That's actually perfect for our purpose. >>> (...) >>> I'm not sure how to determine _which_ multiple of 1MB - start two >>> threads and compare their stack addresses, maybe? In any case, I >>> agree it sounds rather more machine-dependent than the current code. >> That's what the LDAP_PVT_THREAD_STACK_SIZE macro tells you. It's really >> not hard. If we were to disallow increasing the number of threads at >> runtime, we could simply allocate a block of thread stacks all at once, >> which would then allow our thread IDs to be simple integers from 0-N. > > Hmm. I asked the wrong question, should have read some man pages first. > Still, thr_cthreads.c & thr_lwp.c don't use LDAP_PVT_THREAD_STACK_SIZE. > man pthread_attr_getstacksize says it sets the _minimum_ stack size, but > perhaps several M is large enough that one can expect it won't get > multiples of that. That kinda implies that no one has been using the cthreads or lwp support recently, since I doubt their default stack size would work with back-bdb's filterindex demands. -- -- Howard Chu Chief Architect, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc/ Chief Architect, OpenLDAP http://www.openldap.org/project/

1 0

(ITS#5025) back_ldap crashes if it gets a search reference from remote server
by mhardin＠symas.com 19 Jun '07

19 Jun '07

Full_Name: Matthew Hardin Version: 2.3.35 OS: Linux 2.6.9 URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (63.193.240.128) The back_ldap backend crashes if it gets a search reference from the remote server. The following patch corrects the problem. --- search.c 2007-06-19 11:01:13.000000000 -0700 +++ search.c.patched 2007-06-19 10:50:38.000000000 -0700 @@ -383,6 +383,7 @@ BER_BVZERO( &rs->sr_ref[ cnt ] ); /* ignore return value by now */ + rs->sr_entry = NULL; ( void )send_search_reference( op, rs ); } else {

1 0

Re: (ITS#4943) tpool.c pause vs. finish
by h.b.furuseth＠usit.uio.no 19 Jun '07

19 Jun '07

hyc(a)symas.com writes: >h.b.furuseth(a)usit.uio.no wrote: >> It's called from pool_wrapper(). pool_wrapper() could go active so >> there can be no pause while that context_reset() call is running. > > Oh, ok. This is pretty much a non-issue because threads in the pool stay > live until slapd shuts down. The one exception is for a cn=config modify > that decreases the size of the pool to below the number of currently > active threads, in which case some number of threads will exit > immediately. I don't see any potential for conflict here. Eh? Yes, that's what I was talking about. But I'm not sure what you mean with them exiting immediately - or how they could. Also they stay around until there are no pending requests. A pause request could arrive more or less at any time, I presume. (I don't know if there is a pause while the max threads number is being modified.) >>>>> How can a single thread ID be represented by multiple bit patterns? >>>> A struct/union with padding bytes seems the least unlikely possibility. >> (...) >> So in that respect, my preference (not for RE23) is to cast to an >> integer for the hash, and also to add pthread_<set/get>specific() so >> thread_keys[] and its hash will only be a fallback solution. > > Casting to an int only works with simple types, the point of using the > hash was to provide a solution that works for both simple and structured > thread IDs. Yes, except it's not guaranteed to work. Which is why I'm thinking it may be better to fail to compile and provide a fix if/when someone complains. We did have code which compared thread IDs with '==' for quite some time. Do you remember if anyone complained, or if the switch to using ldap_pvt_thread_equal() was a "just in case" fix? However note that we do #define ldap_int_thread_equal(a, b) ((a) == (b)) for all our supported thread implementations except pthreads, and for pthreads we can test pthread_<set/get>specific() just fine (as a replacement for thread_keys[]). OTOH, '==' also works for pointers that are wider than any integer type (and for floating-point thread IDs:-) so I guess I'm arguing against my own suggestion on that one if we add pthread_<set/get>specific() support. > If you want to #ifdef the code so the two cases are handled > differently, then just get rid of TID_HASH for the integral case. No, sould still hash to avoid clustering. > I'm still leary of the pthread TLS support, Me too. But the old code was broken on real-life systems I have access to, not just on theoretical ones. Still waiting to hear from Kurt if my fix to the first tls.c change fixed it for FreeBSD. > and making that change will > require us to update all of the other thread package support, some of > which I don't have access to build/test any more (e.g. SunOS4 LWP). What do you mean? The TLS problem is to extract an unsigned long from a ldap_pvt_thread_t. That needs no support in the thr_*.c files, only to know what kind of type ldap_pvt_thread_t is. Once configure knows what ldap_pvt_thread_t will be typedeffed to it can test if that's a pointer, integer or something worse - and whether it's too wide for a long. BTW I don't know if the unsigned long needs to be the thread ID or just _some_ unique integer. A question to openssl-users(a)openssl.org got no answer. >>>>> The best fix may be to just use the address of the thread's stack as the >>>>> thread ID. That's actually perfect for our purpose. >> (...) >> I'm not sure how to determine _which_ multiple of 1MB - start two >> threads and compare their stack addresses, maybe? In any case, I >> agree it sounds rather more machine-dependent than the current code. > > That's what the LDAP_PVT_THREAD_STACK_SIZE macro tells you. It's really > not hard. If we were to disallow increasing the number of threads at > runtime, we could simply allocate a block of thread stacks all at once, > which would then allow our thread IDs to be simple integers from 0-N. Hmm. I asked the wrong question, should have read some man pages first. Still, thr_cthreads.c & thr_lwp.c don't use LDAP_PVT_THREAD_STACK_SIZE. man pthread_attr_getstacksize says it sets the _minimum_ stack size, but perhaps several M is large enough that one can expect it won't get multiples of that. -- Regards, Hallvard

1 0

Re: (ITS#5010) broken ber_encode_oid/ber_decode_oid()
by hyc＠symas.com 19 Jun '07

19 Jun '07

h.b.furuseth(a)usit.uio.no wrote: > I wrote: >> Finally, libldap/tls.c does not check if ber_decode_oid() fails. > > Fixed that. Waiting for the other stuff, depends on the code's intent. > Go ahead and fix whatever problems you see. Assume that maliciously constructed cert DNs are present, because undoubtedly there will be. -- -- Howard Chu Chief Architect, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc/ Chief Architect, OpenLDAP http://www.openldap.org/project/

1 0

Re: (ITS#5005) test017 fails
by hyc＠symas.com 19 Jun '07

19 Jun '07

ando(a)sys-net.it wrote: > dieter(a)dkluenter.de wrote: > >> the patch fails, this is the patch reject file: > > That's probably because you cut'n'pasted from the email, where the tabs > were probably replaced by whitespaces. I'll apply that fix if anyone > confirms that cookie is legal at all. Yes. There's no hard definition of the cookie requirements. "rid=xxx" without a CSN is the same as no cookie at all. The patch seems to work here, go ahead and commit it. -- -- Howard Chu Chief Architect, Symas Corp. http://www.symas.com Director, Highland Sun http://highlandsun.com/hyc/ Chief Architect, OpenLDAP http://www.openldap.org/project/

1 0

Re: (ITS#5022) ldappasswd crash consumer slapd with some loglevels
by ando＠sys-net.it 19 Jun '07

19 Jun '07

Simon Gao wrote: > Pierangelo Masarati wrote: >> gao(a)schrodinger.com wrote: >> >> >>> URL: ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_2.3.35.its.ext >>> >> ^^^ This link is unreachable >> >> > Sorry about the broken link. Here is the correct one: > > ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_20070618.ext > >>> When following command against a consumer slapd, it will crash slapd of the >>> consumer when loglevel is set to any other value other than 1 or -1 on the >>> consumer: >>> >>> ldappasswd -v -H ldap://consumer -D "uid=joe,ou=people,dc=example,dc=com" -W -S >>> -x -A >>> >>> The crash happens to at least to loglevel 0, 256, 512. When loglevel is set to 1 >>> or "-1", then no crash is experienced. In addition, when run consumer slapd >>> manually as front process, >>> >> ^^^ what does this mean? >> > If I start slapd manually as following commands: > > /usr/lib/openldap/slapd -d 0 -u ldap -g ldap -h 'ldap:// ldaps' > /usr/lib/openldap/slapd -d 256 -u ldap -g ldap -h 'ldap:// ldaps' > /usr/lib/openldap/slapd -d 512 -u ldap -g ldap -h 'ldap:// ldaps' > > Then I did not see the problem as I would when setting loglevel to 0, > 256, 512 in slapd.conf and start slapd from /etc/init.d/slapd, which > launch the daemon with the same options. When starting slapd from > /etc/init.d/slapd as background daemon process, slapd will write to > /var/log/ldap.log and /var/log/message. It seems that the difference > exists between writing ldap logs to a file (slapd runs in background) or > writing to console (slapd runs in foreground as started manually). > >> >>> then all debug level works without problem. >>> >> OpenLDAP re23 (in practice, 2.3.36 just released) doesn't show anything >> like that. I note that differences related to the debug level usually >> mean that a NULL or an invalid pointer is passed to a *printf(3) routine >> on those systems that do not tolerate it (e.g. Solaris). But usually >> the bug disappears when __decreasing__ the log level, while -1 means all. >> >> p. >> > > I would expect writing less log should help avoiding problem. In this > case, somehow logging must be set at certain or above a level. Below > that it will trigger problem. While I was trying to get pieces working, > I always set loglevel to 1 or -1. So I did not realize this problem > until very late. The error I got as a result of attempting change > password is "can't contact LDAP server" which is very confusing (might > be because slapd already crashed before serving last request). > > I am happy to run more tests and provide further information as needed. If slapd crashes, the only really useful info is a core dump, as indicated in the FAQ I already pointed you to (that's basically why I pointed you there: there's no point in always repeating instructions for bug tracing, that's what the FAQ is intended for; apparently, it's totally ignored). p. Ing. Pierangelo Masarati OpenLDAP Core Team SysNet s.r.l. via Dossi, 8 - 27100 Pavia - ITALIA http://www.sys-net.it --------------------------------------- Office: +39 02 23998309 Mobile: +39 333 4963172 Email: pierangelo.masarati(a)sys-net.it ---------------------------------------

1 0

Re: (ITS#5022) ldappasswd crash consumer slapd with some loglevels
by gao＠schrodinger.com 19 Jun '07

19 Jun '07

Pierangelo Masarati wrote: > gao(a)schrodinger.com wrote: > > >> URL: ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_2.3.35.its.ext >> > > ^^^ This link is unreachable > > Sorry about the broken link. Here is the correct one: ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_20070618.ext >> When following command against a consumer slapd, it will crash slapd of the >> consumer when loglevel is set to any other value other than 1 or -1 on the >> consumer: >> >> ldappasswd -v -H ldap://consumer -D "uid=joe,ou=people,dc=example,dc=com" -W -S >> -x -A >> >> The crash happens to at least to loglevel 0, 256, 512. When loglevel is set to 1 >> or "-1", then no crash is experienced. In addition, when run consumer slapd >> manually as front process, >> > > ^^^ what does this mean? > If I start slapd manually as following commands: /usr/lib/openldap/slapd -d 0 -u ldap -g ldap -h 'ldap:// ldaps' /usr/lib/openldap/slapd -d 256 -u ldap -g ldap -h 'ldap:// ldaps' /usr/lib/openldap/slapd -d 512 -u ldap -g ldap -h 'ldap:// ldaps' Then I did not see the problem as I would when setting loglevel to 0, 256, 512 in slapd.conf and start slapd from /etc/init.d/slapd, which launch the daemon with the same options. When starting slapd from /etc/init.d/slapd as background daemon process, slapd will write to /var/log/ldap.log and /var/log/message. It seems that the difference exists between writing ldap logs to a file (slapd runs in background) or writing to console (slapd runs in foreground as started manually). > >> then all debug level works without problem. >> > > OpenLDAP re23 (in practice, 2.3.36 just released) doesn't show anything > like that. I note that differences related to the debug level usually > mean that a NULL or an invalid pointer is passed to a *printf(3) routine > on those systems that do not tolerate it (e.g. Solaris). But usually > the bug disappears when __decreasing__ the log level, while -1 means all. > > p. > I would expect writing less log should help avoiding problem. In this case, somehow logging must be set at certain or above a level. Below that it will trigger problem. While I was trying to get pieces working, I always set loglevel to 1 or -1. So I did not realize this problem until very late. The error I got as a result of attempting change password is "can't contact LDAP server" which is very confusing (might be because slapd already crashed before serving last request). I am happy to run more tests and provide further information as needed. Thanks, Simon

1 0

Re: (ITS#5022) ldappasswd crash consumer slapd with some loglevels
by quanah＠zimbra.com 18 Jun '07

18 Jun '07

--On Tuesday, June 19, 2007 5:37 AM +0000 ando(a)sys-net.it wrote: > gao(a)schrodinger.com wrote: > >> URL: >> ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_2.3.35.its >> .ext <ftp://ftp.schrodinger.com/support/openldap/simon.gao.openldap_20070618.ext> is the actual file. --Quanah -- Quanah Gibson-Mount Principal Software Engineer Zimbra, Inc -------------------- Zimbra :: the leader in open source messaging and collaboration

1 0

← Newer
1
2
3
4
5
6
7
8
...
17
Older →

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

openldap-bugs June 2007