Re: [Dovecot] dovecot genesis v2.0.X
On Mon, 2010-10-18 at 09:47 -0700, Jim Pazarena wrote: So, my question is, respectfully, is dovecot V2.0.5 stable enough for prime-time on a busy ISP mail server? I'll install it myself independent of the FreeBSD ports tree if so. It's getting better :) I'll make 2.0.6 release this week again with more fixes, but I still have a few bug reports that I haven't managed to reproduce. I'm concerned with the very same question. I currently have 1.2.12 running very well and I see risk in an upgrade to ver 2.x. Unless of course the benefits outweigh the risks. I have been carefully watching the ver 2.x tree rev from week to week and I await the defacto signal from people that it is in fact at least as solid at 1.2.12, that signal being adoption by people in busy mail servers. So therefore I await ver 2.0.6 with a mix of conflicting anticipation and trepidation. :-) -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] dovecot genesis v2.0.X
I am running 2.0.5 (svn repository) in production for about a thousand users without major issues. Please Sir, on what OS and architecture ? -- Dennis
Re: [Dovecot] dovecot genesis v2.0.X
On Mon, 2010-10-18 at 13:46 -0400, Dennis Clarke wrote: I am running 2.0.5 (svn repository) in production for about a thousand users without major issues. Please Sir, on what OS and architecture ? I think the stability has less to do with that, and more to do with what parts of Dovecot you want to use. If your test setup works fine with the most common clients you use, and it survives a bit of stress testing, it should work fine. The things I still mostly worry about are some rare errors/crashes. There is still a rather large number of Solaris 8 Sparc servers out there running in production and based on the conversations I have had with some sysadmins they will upgrade them when the smoke comes out. That could be another five years. So, before you say stability has less to do with that you should see the horrendous work it takes to port an app from the open source world ( usually Linux people with tons of Linux-isms/GCC-isms in it ) over to an old POSIX/XPG4 ( POSIX.1-1990, POSIX.2-1992, POSIX.1b-1993, POSIX.1c-1996 etc ) compliant UNIX. My coffee cup needs whiskey some days ;-) -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
[Dovecot] dovecot-2.0.3 : Am I missing anything here ?
I'm just asking if there is anything else that needs to be added at all. I seem to have everything linked in from soup to nuts but I want to be sure that a basic or essential feature is not being overlooked for forgotten : [titan] elfdump -d .libs/auth | grep NEED [0] NEEDED 0x3084libdovecot.so.0 [1] NEEDED 0x3055libpam.so.1 [2] NEEDED 0x306alibpthread.so.1 [3] NEEDED 0x3094libkrb5.so.3 [4] NEEDED 0x30a1libk5crypto.so.3 [5] NEEDED 0x30b2libcom_err.so.3 [6] NEEDED 0x30c2libldap-2.4.so.2 [7] NEEDED 0x30d3libsasl2.so.2 [8] NEEDED 0x30e1libgssapi_krb5.so.2 [9] NEEDED 0x30f5libssl.so.0.9.8 [10] NEEDED 0x3105libcrypto.so.0.9.8 [11] NEEDED 0x3118liblber-2.4.so.2 [12] NEEDED 0x3129libresolv.so.2 [13] NEEDED 0x3138libgen.so.1 [14] NEEDED 0x3144libpq.so.5 [15] NEEDED 0x314flibsqlite3.so.0 [16] NEEDED 0x315flibz.so [17] NEEDED 0x3167libdl.so.1 [18] NEEDED 0x3172libnsl.so.1 [19] NEEDED 0x317elibsocket.so.1 [20] NEEDED 0x318dlibintl.so.8 [21] NEEDED 0x319alibiconv.so.2 [22] NEEDED 0x307alibc.so.1 [23] NEEDED 0x31a8librt.so.1 [24] NEEDED 0x31b3libmysqlclient.so.16 [35] VERNEED 0x8057770 [36] VERNEEDNUM 0x3 [titan] [titan] ldd .libs/auth libdovecot.so.0 = (file not found) libpam.so.1 = /usr/lib/libpam.so.1 libpthread.so.1 = /usr/lib/libpthread.so.1 libkrb5.so.3 = /opt/csw/lib/libkrb5.so.3 libk5crypto.so.3 = /opt/csw/lib/libk5crypto.so.3 libcom_err.so.3 = /opt/csw/lib/libcom_err.so.3 libldap-2.4.so.2 = /opt/csw/lib/libldap-2.4.so.2 libsasl2.so.2 = /opt/csw/lib/libsasl2.so.2 libgssapi_krb5.so.2 = /opt/csw/lib/libgssapi_krb5.so.2 libssl.so.0.9.8 = /opt/csw/lib/libssl.so.0.9.8 libcrypto.so.0.9.8 =/opt/csw/lib/libcrypto.so.0.9.8 liblber-2.4.so.2 = /opt/csw/lib/liblber-2.4.so.2 libresolv.so.2 =/usr/lib/libresolv.so.2 libgen.so.1 = /usr/lib/libgen.so.1 libpq.so.5 =/opt/csw/postgresql83/lib/i386/libpq.so.5 libsqlite3.so.0 = /opt/csw/lib/libsqlite3.so.0 libz.so = /opt/csw/lib/libz.so libdl.so.1 =/usr/lib/libdl.so.1 libnsl.so.1 = /usr/lib/libnsl.so.1 libsocket.so.1 =/usr/lib/libsocket.so.1 libintl.so.8 = /opt/csw/lib/libintl.so.8 libiconv.so.2 = /opt/csw/lib/libiconv.so.2 libc.so.1 = /usr/lib/libc.so.1 librt.so.1 =/usr/lib/librt.so.1 libmysqlclient.so.16 = /opt/csw/mysql51/lib/i386/mysql/libmysqlclient.so.16 libcmd.so.1 = /usr/lib/libcmd.so.1 libkrb5support.so.0 = /opt/csw/lib/libkrb5support.so.0 libintl.so.1 = /usr/lib/libintl.so.1 libgcc_s.so.1 = /opt/csw/lib/libgcc_s.so.1 libmp.so.2 =/usr/lib/libmp.so.2 libaio.so.1 = /usr/lib/libaio.so.1 libm.so.1 = /usr/lib/libm.so.1 libthread.so.1 =/usr/lib/libthread.so.1 That looks to be a massive list. I'm always worried that some feature is being forgotten by me. Any thoughts from the gurus ? -- Dennis
[Dovecot] Linux should be as easy as Windows
From a Standards thread I saw this : Linux should be as easy as Windows I would like to point out that this is a thinking problem. A shift in thinking is required. That should say : Open source and Free software should be as easy as Windows In this way we allow someone running BSD, Solaris or some port of Debian to an old Dec AlphaServer will have the same level of functionality provided standards are adhered to. Once programmers begin to see everything in the light of Linux then we can also kiss goodbye the opportunity for portability to other OS types. Dennis
Re: [Dovecot] Linux should be as easy as Windows
[ sorry for the slight OT ] On 18/08/2010 15:41, Dennis Clarke wrote: From a Standards thread I saw this : Linux should be as easy as Windows I would like to point out that this is a thinking problem. A shift in thinking is required. That should say : Open source and Free software should be as easy as Windows In this way we allow someone running BSD, Solaris or some port of Debian to an old Dec AlphaServer will have the same level of functionality provided standards are adhered to. Once programmers begin to see everything in the light of Linux then we can also kiss goodbye the opportunity for portability to other OS types. Dennis I think everything defined as easy is not (random order): - Light Sometimes things are not and that is okay. GNOME is not XFCE or Fluxbox but we have freedom to choose either. Perhaps by light you mean you don't need 32G of memory and a dual socket 12-core Opteron 6100 series machine with some hot NVidia graphics card and 10Gb ethernet to just *barely* install Windows 8. :-) - Scalable Nice if possible. Some things just don't scale well. A P-vs-NP type problem. - Secure Love dovecot for this reason. - Portable Totally. This is what open standards are supposed to be all about. Also why assembly tweaks to architecture locked modules should be avoided whereever possible. Sometimes there is no way out. - Customizable Nice but .. again not a hard need in my mind. As someone that *tried* to work on this disaster : http://www.computerworld.com/s/article/print/107625/Developers_put_OpenSolaris_on_PowerPC?taxonomyName=Open+SourcetaxonomyId=88 I watched as a large filthy corporation step in and say never ever on IBM hardware and the whole thing was taken inside to become this : http://labs.oracle.com/spotlight/2006/2006-06-14-SolarisPPC.html Where it was very quietly killed. Let's hope the big corporate world never calls up Mr. T ( dovecot author ) and tosses a billion dollars at him. A few million sure ... but not a billion. Otherwise dovecot would be wrecked by the greedy pigs at Oracle or IBM or Apple etc. Just my opinion. Sorry for the OT Dennis
Re: [Dovecot] Error: Index file too large ? 904 bytes ?
Something's just completely broken in your install. yep. I agree totally which is why I pkgrm'd it and went back to 1.2.12 which works flawlessly. I'll come back to 1.2.13 ( maybe ) after I get 2.0.0 working well Dennis
Re: [Dovecot] Linux should be as easy as Windows
- Original Message - From: Jerry dovecot.u...@seibercom.net Date: Wednesday, August 18, 2010 11:03 am Subject: Re: [Dovecot] Linux should be as easy as Windows To: dovecot@dovecot.org On Wed, 18 Aug 2010 10:07:27 -0400 Dennis Clarke dcla...@blastwave.org articulated: Sometimes things are not and that is okay. GNOME is not XFCE or Fluxbox but we have freedom to choose either. Perhaps by light you mean you don't need 32G of memory and a dual socket 12-core Opteron 6100 series machine with some hot NVidia graphics card and 10Gb ethernet to just *barely* install Windows 8. :-) I just installed Win7 on a 7 year old Dell 4550 with only 1G of RAM. I have a new shiney multicore 64-bit machine with an NVidia graphics card etc etc and Win 7 seems to work okay with 4G of memory. Performance still sucks compared to Solaris or Debian on the same box but it plays movies well. I installed Win 7 Ultimate 64-bit which I bought as a non-upgrade box kit. Obviously, since I do not have the latest video cards, etc. installed in the PC, certain functions were disabled by default; however, it definitely works. I only did the installation to see if it would really work, and was pleasantly surprised to see how well it performed. That was, of course, on a virgin system (all data erased from drives and reformatted). I'll try a old 32-bit box with baseline hardware sometime just to see ... it could be a waste of my time however. By the way, Win8 is not even scheduled to be released until at least July 1, 2010 and even that is not written in stone. I as joking. :-) Dennis
Re: [Dovecot] v2.0.0 released
- Original Message - From: Timo Sirainen t...@iki.fi Date: Monday, August 16, 2010 6:31 pm Subject: Re: [Dovecot] v2.0.0 released Cc: Dovecot Mailing List dovecot@dovecot.org On 16.8.2010, at 23.25, Dennis Clarke wrote: imap-match.c, line 199: identifier redeclared: imap_match_dup I think someone reported this problem earlier .. but I'm not sure. Anyway I've no idea why it would complain about this. So that stops me while I figure out what the issue is with imap-match.h and/or imap-match.c in the Solaris world while using Sun Studio 11. I heard it compiles ok with Sun Studio 12.1. Seems to compile fine with GCC 4.5.1 and binutils 2.20.1.20100303. However the install process then fails repeatedly. Any thoughts on this ? http://www.blastwave.org/docs/dovecot_2.0.0_install-FAIL.txt The same log with numbers lines is here : http://www.blastwave.org/docs/dovecot_2.0.0_install-FAIL-N.txt Where I see on line 354 : libtool: install: ../.././install-sh -c .libs/auth /opt/csw/libexec/dovecot/auth Which sure looks like the file auth being copied into place and then right after we get a failure : libtool: install: ../.././install-sh -c .libs/checkpassword-reply /opt/csw/libexec/dovecot/checkpassword-reply test -z /opt/csw/libexec/dovecot/auth || ../.././install-sh -c -d /opt/csw/libexec/dovecot/auth mkdir: /opt/csw/libexec/dovecot/auth: Exists but is not a directory mkdir: /opt/csw/libexec/dovecot/auth: Exists but is not a directory On line 372 onwards I try again with a whole new compile and install fails again. I think the compile went fine but getting the results into place on the filesystem is turning out to be a real chore. Certainly not make install simple. Dennis
Re: [Dovecot] v2.0.0 released
On Tue, 2010-08-17 at 09:34 -0400, Dennis Clarke wrote: libtool: install: ../.././install-sh -c .libs/auth /opt/csw/libexec/dovecot/auth Which sure looks like the file auth being copied into place Yes, that's correct. and then right after we get a failure : libtool: install: ../.././install-sh -c .libs/checkpassword-reply /opt/csw/libexec/dovecot/checkpassword-reply This is also correct. test -z /opt/csw/libexec/dovecot/auth || ../.././install-sh -c -d /opt/csw/libexec/dovecot/auth This || install part seems wrong, but it happens with me too. But the main problem is that why does test -z fail above? The auth binary should have been installed there already, so test -z should return success. Or is the auth file zero bytes long? Something's wrong with that.. No it is a real valid Sparc binary : # file /opt/csw/libexec/dovecot/auth /opt/csw/libexec/dovecot/auth: ELF 32-bit MSB executable SPARC Version 1, dynamically linked, not stripped I'm going to experiment a bit here ... I have a theory but it will take hours to track down. Dennis
Re: [Dovecot] v2.0.0 released
On Tue, 2010-08-17 at 17:48 +0100, Timo Sirainen wrote: On Tue, 2010-08-17 at 17:32 +0100, Timo Sirainen wrote: test -z /opt/csw/libexec/dovecot/auth || ../.././install-sh -c -d /opt/csw/libexec/dovecot/auth This || install part seems wrong, but it happens with me too. But the main problem is that why does test -z fail above? The auth binary should have been installed there already, so test -z should return success. Or is the auth file zero bytes long? Something's wrong with that.. Uh. No, I remembered wrong. test -z tests if the string is empty. And of course the above string isn't empty. And looks like the mkdir is run with Linux, but it doesn't report a failure. Umm.. Oh, right. The above should be: test -z /opt/csw/lib/dovecot/auth || /bin/mkdir -p /opt/csw/lib/dovecot/auth Why is it libexec and not lib? What parameters did you give to configure? Looks as if you gave --libdir=/opt/csw/libexec This may help : [titan] more ../../sparc/dovecot-2.0.0-sparcv8-002/config.log This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. It was created by Dovecot configure 2.0.0, which was generated by GNU Autoconf 2.65. Invocation command line was $ ./configure --build=sparc-sun-solaris2.8 --host=sparc-sun-solaris2.8 --prefi x=/opt/csw --sysconfdir=/etc/opt/csw/dovecot --localstatedir=/var/opt/csw/doveco t --enable-shared --enable-static --with-ioloop=poll --with-shadow --with-pam -- with-ldap=yes --with-sql=yes --with-mysql --with-sqlite --with-zlib --with-bzlib --with-ssl=openssl --with-storages=mbox,maildir --with-libiconv-prefix=/opt/csw --with-gssapi=yes --with-pgsql --with-libwrap --with-ssldir=/etc/opt/csw/ssl -- with-moduledir=/opt/csw/libexec/dovecot ## - ## ## Platform. ## ## - ## hostname = mimas uname -m = sun4u uname -r = 5.8 uname -s = SunOS uname -v = Generic_127721-03 /usr/bin/uname -p = sparc /bin/uname -X = System = SunOS Node = mimas Release = 5.8 KernelID = Generic_127721-03 Machine = sun4u BusType = unknown Serial = unknown Users = unknown OEM# = 0 Origin# = 1 NumCPU = 1 /bin/arch = sun4 /usr/bin/arch -k = sun4u /usr/convex/getsysinfo = unknown /usr/bin/hostinfo = unknown /bin/machine = unknown /usr/bin/oslevel = unknown /bin/universe = unknown PATH: /opt/csw/gcc4/bin PATH: /opt/csw/bin PATH: /usr/xpg4/bin PATH: /usr/ccs/bin PATH: /usr/bin PATH: /sbin PATH: /bin PATH: /usr/sbin PATH: /opt/schily/bin
Re: [Dovecot] v2.0.0 released
On Tue, 2010-08-17 at 13:15 -0400, Dennis Clarke wrote: [titan] more ../../sparc/dovecot-2.0.0-sparcv8-002/config.log This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. .. --with-moduledir=/opt/csw/libexec/dovecot This is wrong. Modules don't belong under libexec. Where would you like them ?
Re: [Dovecot] v2.0.0 released
On Tue, 2010-08-17 at 13:19 -0400, Dennis Clarke wrote: On Tue, 2010-08-17 at 13:15 -0400, Dennis Clarke wrote: [titan] more ../../sparc/dovecot-2.0.0-sparcv8-002/config.log This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. .. --with-moduledir=/opt/csw/libexec/dovecot This is wrong. Modules don't belong under libexec. Where would you like them ? Default location would be fine (= lib). Anything except libexec, since it creates a conflict. I'm doing a recompile and tossing --with-moduledir out :-)
[Dovecot] Error: Index file too large ? 904 bytes ?
My 1.2.13 build is giving me errors based on the same conf file that was working perfectly for 1.2.12 : Aug 18 03:28:27 IMAP(dclarke): Error: Index file too large: /var/mail/dclarke/dovecot.index Aug 18 03:28:27 IMAP(dclarke): Fatal: block_alloc(2147483648): Out of memory Aug 18 03:28:27 dovecot: Error: child 12560 (imap) returned error 83 (Out of memory - see mail_process_size setting) The file in question is 904 bytes in size : bash-4.1# ls -l /export/home/dclarke/mail/.Sent/dovecot.index -rw--- 1 dclarke csw 904 Aug 4 16:05 /export/home/dclarke/mail/.Sent/dovecot.index So this seems insane. SO I made sure the conf file had this in it : mail_process_size = 256 I then restart the daemon and see this : Aug 18 03:37:02 IMAP(dclarke): Fatal: block_alloc(1073741824): Out of memory Aug 18 03:37:02 dovecot: Error: child 12620 (imap) returned error 83 (Out of memory - see mail_process_size setting) Aug 18 03:37:07 IMAP(dclarke): Error: Index file too large: /export/home/dclarke/mail/.Sent/dovecot.index Aug 18 03:37:07 IMAP(dclarke): Warning: Created dotlock file's timestamp is different than current time (837089755 vs 1282102627): /export/home/dclarke/mail/.Sent/dovecot-uidlist Aug 18 03:37:07 IMAP(dclarke): Fatal: block_alloc(1073741824): Out of memory Aug 18 03:37:07 dovecot: Error: child 12622 (imap) returned error 83 (Out of memory - see mail_process_size setting) Aug 18 03:37:09 IMAP(dclarke): Error: Index file too large: /export/home/dclarke/mail/.Sent/dovecot.index Aug 18 03:37:09 IMAP(dclarke): Warning: Created dotlock file's timestamp is different than current time (527779909 vs 1282102629): /export/home/dclarke/mail/.Sent/dovecot-uidlist Aug 18 03:37:09 IMAP(dclarke): Fatal: block_alloc(1073741824): Out of memory Aug 18 03:37:09 dovecot: Error: child 12624 (imap) returned error 83 (Out of memory - see mail_process_size setting) Sure enough dovecot catches a timestamp issue : bash-4.1# date Wed Aug 18 03:39:24 GMT 2010 bash-4.1# ls -lapE /export/home/dclarke/mail/.Sent/dovecot-uidlist -rw--- 1 dclarke csw 1806 2010-08-16 13:52:16.442510077 + /export/home/dclarke/mail/.Sent/dovecot-uidlist bash-4.1# touch /export/home/dclarke/mail/.Sent/dovecot-uidlist bash-4.1# ls -lapE /export/home/dclarke/mail/.Sent/dovecot-uidlist -rw--- 1 dclarke csw 1806 2010-08-18 03:39:53.263345256 + /export/home/dclarke/mail/.Sent/dovecot-uidlist However the inability to deal with a 904 byte file is a bit of an issue here. Anyone know why mail_process_size = 256 is the wrong thing to look at ? Dennis
[Dovecot] reverting back to 1.2.12
ver 1.2.13 is also tossing these : Aug 18 03:37:49 dovecot: Warning: I/O leak: 805c600 (27) Aug 18 03:37:49 dovecot: Warning: I/O leak: 805c600 (25) Aug 18 03:37:49 dovecot: Warning: I/O leak: 805c600 (23) I have reverted back to 1.2.12 which just works. I will go back and recompile 1.2.13 and 2.0.0 again. Maybe the addition of postgressql libs linked in or the use of GCC is the issue on Solaris. Not sure but 1.2.12 to 1.2.13 did not go smoothly at all. Dennis
Re: [Dovecot] Upgrade to 2.0 - Not so smooth
On 2010-08-16 07:16:46 -0700, Marc Perkel wrote: Maybe dovecot could use the nobody user if dovenull doesn't exist? you could just read the warning message, create the user and stop worrying? and no ... it shouldnt fallback to nobody. a broken set up is a broken setup and needs to be fixed. Wow. Well said. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
http://dovecot.org/releases/2.0/dovecot-2.0.0.tar.gz http://dovecot.org/releases/2.0/dovecot-2.0.0.tar.gz.sig As promised last Friday, here's the v2.0.0 release finally. I'm cautiously optimistic that v2.0.1 won't (have to) be released for a few weeks, since there was quite a lot of testing and fixing going on in the RC stage. cool .. .however .. should I see this message at the end of the configure stage : . . . passdbs : static passwd passwd-file shadow pam checkpassword ldap sql : -bsdauth -sia -vpopmail userdbs : static prefetch passwd passwd-file checkpassword ldap sql : -vpopmail -nss SQL drivers : mysql sqlite : -pgsql NOTE: This is the UNSTABLE development branch of Dovecot v2.0. ^ That won't go well with the Solaris crowd. Just FYI. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
On 08/16/2010 06:46 PM Dennis Clarke wrote: ⦠NOTE: This is the UNSTABLE development branch of Dovecot v2.0. ^ See: http://dovecot.org/list/dovecot/2010-August/051736.html wow ... I was hoping I was the first :-\ -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
-2.0.0-sparcv8-001/src/lib-imap' gmake[2]: *** [all-recursive] Error 1 gmake[2]: Leaving directory `/export/medusa/dclarke/build/dovecot/sparc/dovecot-2.0.0-sparcv8-001/src' gmake[1]: *** [all-recursive] Error 1 gmake[1]: Leaving directory `/export/medusa/dclarke/build/dovecot/sparc/dovecot-2.0.0-sparcv8-001' gmake: *** [all] Error 2 [mimas] So that stops me while I figure out what the issue is with imap-match.h and/or imap-match.c in the Solaris world while using Sun Studio 11. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
On 16.8.2010, at 23.25, Dennis Clarke wrote: imap-match.c, line 199: identifier redeclared: imap_match_dup I think someone reported this problem earlier .. but I'm not sure. Anyway I've no idea why it would complain about this. So that stops me while I figure out what the issue is with imap-match.h and/or imap-match.c in the Solaris world while using Sun Studio 11. I heard it compiles ok with Sun Studio 12.1. I will try GCC and let you know what I see : [mimas] gcc --version gcc (Blastwave.org Inc. Mon Aug 9 07:10:45 GMT 2010) 4.5.1 Copyright (C) 2010 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. No surprise that GCC works fine : [mimas] ls -lap ./src/master/.libs/dovecot -rwxr-xr-x 1 dclarke csw86388 Aug 17 00:09 ./src/master/.libs/dovecot [mimas] file ./src/master/.libs/dovecot ./src/master/.libs/dovecot: ELF 32-bit MSB executable SPARC Version 1, dynamically linked, not stripped [mimas] elfdump -d ./src/master/.libs/dovecot Dynamic Section: .dynamic index tag value [0] NEEDED 0xe21 libdovecot.so.0 [1] NEEDED 0xe31 libintl.so.8 [2] NEEDED 0xe3e libdl.so.1 [3] NEEDED 0xe49 libnsl.so.1 [4] NEEDED 0xdf6 libsocket.so.1 [5] NEEDED 0xe55 libiconv.so.2 [6] NEEDED 0xe0e libc.so.1 [7] NEEDED 0xe63 librt.so.1 [8] INIT 0x1b71c [9] FINI 0x1b738 etc etc ... I did see a few odd messages fly by like this : *** Warning: Linking the shared library lib30_imap_zlib_plugin.la against the loadable module *** lib20_zlib_plugin.so is not portable! Not too sure what is up with that but I'll look into it. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
So that stops me while I figure out what the issue is with imap-match.h and/or imap-match.c in the Solaris world while using Sun Studio 11. Works fine here with the 12.1 compiler. Solaris 10 on UltraSPARC, current patches, on both the OS and the compilers. My CFLAGS: -fast -xtarget=ultra3 -m32 -xarch=sparcvis2 The is ye ol'production Solaris 8 and thus Sun Studio 11 is needed. Problem solved simply by using GCC 4.5.1 from Blastwave ( of course ). Question for you: why would you use -m32 with an -xarch=sparcvis2 ? Why not go for greatest possible portability and range of processors with something like this : CFLAGS=-m32 -Qy -Xa -xbuiltin=%none -xnolibmil -xnolibmopt -xcode=pic32 -xildoff -xregs=no%appl -xs -xstrconst -D_TS_ERRNO If you feel like some optimization then chuck in -xO3 and -xdepend however I'd test the resultant bins from a basic compile first. Just my 0.02 -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
On 17.8.2010, at 1.28, Dennis Clarke wrote: *** Warning: Linking the shared library lib30_imap_zlib_plugin.la against the loadable module *** lib20_zlib_plugin.so is not portable! Not too sure what is up with that but I'll look into it. That's normal. It doesn't need to be portable. In that case ... awesome :-) I'll let you know when I get packages into testing QC and then for release. However I am going to rename the package to dovecot2 in order to separate it from dovecot 1.x installs. My idea there is to allow the user to continue with the dovecot 1.x line in the event of patch updates and further releases from that tree ( like dovecot 1.2.13 ) while also maintaining the ver 2.x family. However I don't think I'll support multiple instances of dovecot ( regardless of version ) in the same Solaris zone. If the user wants that then they can do all the uid/pid hackery themselves. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
On 8/16/10 8:38 PM, Dennis Clarke wrote: Works fine here with the 12.1 compiler. Solaris 10 on UltraSPARC, current patches, on both the OS and the compilers. My CFLAGS: -fast -xtarget=ultra3 -m32 -xarch=sparcvis2 The is ye ol'production Solaris 8 and thus Sun Studio 11 is needed. Oh. Eeeew. :) Hey, there are banks and telco places all over that are running it and they *refuse* to upgrade. I know of government sites that are still on old E450's because they have not died yet. No kidding. Some of that old Sun gear just WILL NOT bloody die. The way to justify the capital expenditure on a new Sun Sparc server is to spill a coffee on/in it. 8-P Problem solved simply by using GCC 4.5.1 from Blastwave ( of course ). I wonder how GCCFSS would do. Will it even run under Solaris 8? Why not ? It is a rock solid kernel and you can build it and see :-) Question for you: why would you use -m32 with an -xarch=sparcvis2 ? Why not go for greatest possible portability and range of processors with something like this : CFLAGS=-m32 -Qy -Xa -xbuiltin=%none -xnolibmil -xnolibmopt -xcode=pic32 -xildoff -xregs=no%appl -xs -xstrconst -D_TS_ERRNO Because I don't intend to run it on a range of processors. =) ah ... I am in a different world. If you feel like some optimization then chuck in -xO3 and -xdepend however I'd test the resultant bins from a basic compile first. Just my 0.02 Fair enough, and I appreciate your advice. My goal in this case is to generate the fastest binaries possible on the machine it's being built on, which is identical to the machine it's going to be run on. There's zero need for portable binaries in this particular installation. In that case run fpversion : fpversion A SPARC-based CPU is available. Kernel says CPU's clock rate is 900.0 MHz. Kernel says main memory's clock rate is 150.0 MHz. Sun-4 floating-point controller version 0 found. An UltraSPARC chip is available. Use -xtarget=ultra3cu -xcache=64/32/4:8192/512/2 code-generation option. See those compiler options ? Use them. Also put in -xO3 and -xdepend and then the compile will take a lot longer but you will be happy you waited. Sun Studio compilers can really rock if you know how to push them. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] v2.0.0 released
I wonder how GCCFSS would do. Will it even run under Solaris 8? # which gcc /opt/csw/gcc4/bin/gcc # gcc --version gcc (Blastwave.org Inc. Mon Aug 9 07:10:45 GMT 2010) 4.5.1 Copyright (C) 2010 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. # uname -a SunOS mimas 5.8 Generic_127721-03 sun4u sparc SUNW,UltraAX-i2 # psrinfo -v Status of virtual processor 0 as of: 08/17/10 00:58:04 on-line since 07/06/10 12:03:37. The sparcv9 processor operates at 500 MHz, and has a sparcv9 floating point processor. Is that baseline Sparc enough for you ? ;-) -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
Re: [Dovecot] Upgrade to 2.0 - Not so smooth
On 15.8.2010, at 17.03, Marc Perkel wrote: Starting Dovecot Imap: Fatal: service(pop3-login) User doesn't exist: dovenull Just some real time feedback. I don't know what dovenull user is or why it is necessary. I added now to http://wiki2.dovecot.org/Upgrading/2.0 : - Dovecot uses two internal users now by default: dovenull and dovecot. You need to create the dovenull user or change default_login_user setting. Dovecot 2.x seems to be very close to release grade. Could you provide a guess as to when it will be released so that I can plan downtime for the upgrade on my production servers? -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
[Dovecot] Dovecot 1.2.12 released for Solaris 8, 9 and 10
Just an FYI, there are SMF aware packages for dovecot 1.2.12 avail from the Blastwave mirrors now. All world wide mirrors will sync within hours but the primary ( at ISC ) is at http://download.blastwave.org Please use pkgutil to fetch software packages. # pkginfo -l CSWdovecot PKGINST: CSWdovecot NAME: dovecot - Dovecot is an open source IMAP and POP3 server CATEGORY: application ARCH: i386 VERSION: 1.2.12,REV=2010.07.06 BASEDIR: /opt/csw VENDOR: http://www.dovecot.org/ packaged by Blastwave.org Inc. DESC: CSWbind - Dovecot is an open source IMAP and POP3 server PSTAMP: titan20100706031415 INSTDATE: Jul 06 2010 01:58 HOTLINE: http://www.blastwave.org/ EMAIL: supp...@blastwave.org STATUS: completely installed FILES: 573 installed pathnames 6 shared pathnames 37 directories 91 executables 130828 blocks used (approx) Also avail on all Sparc servers of course. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris dcla...@blastwave.org - Email related to open source for Solaris
[Dovecot] I can receive but not send
I'm not too sure where to look here. I can access my mail via Dovecot and there is an initial requirement to accept an SSL key ( self signed ) however after I look in my inbox and read a piece of email ( via SeaMonkey eMail client ) I can not send. I simply get told that the server would not accept the smtp connection. I was hoping for a smtps connection but I'll take what I can get. Am I missing something obvious here ? -- Dennis
Re: [Dovecot] I can receive but not send
Am I missing something obvious here ? Dovecot isn't a SMTP-Server/MTA. See also: http://wiki.dovecot.org/MailServerOverview That figures ... it had to be my sendmail config of all things. Geez .. I hate sendmail configuration. OKay .. thanks. -- Dennis
Re: [Dovecot] I can receive but not send
Am I missing something obvious here ? Yes, Dovecot is not an SMTP server, it provides POP3 and IMAP which are for mail retrieval. Common SMTP servers are Exim, Postfix, Sendmail and Qmail. Well that tells me dovecot is working fine but sendmail is not listening to roaming mail clients. Sending of message failed. The message could not be sent because connecting to SMTP server foo.domainname.ca failed. The server may be unavailable or is refusing SMTP connections. Please verify that your SMTP server settings are correct and try again, or contact the server administrator. However I can telnet to the port 25 just fine ... bizarre. -- Dennis
[Dovecot] Jul 06 13:00:50 dovecot: Warning: I/O leak: 806ef10 (36)
I'm currently fighting my way through the dovecot configuration pages and endless options to see if I can just get my mail from the server. Thus far no joy but dovecot seems to be working if I could figure out the configuration. My process thus far has been : 1) read docs 2) try a config change 3) fail 4) goto (1) I just saw an error while trying to read my /var/mail/dclarke good 'ol mbox in the message.log : Jul 06 12:58:50 IMAP(dclarke): Error: mbox root directory can't be a file: /var/mail/dclarke/ (http://wiki.dovecot.org/MailLocation/Mbox) So I figure I better go read that page again and see what I did wrong this time. I guess this is wrong : mail_location: mbox:/var/mail/%u Even though that is where the mail actually is : # ls -lap /var/mail total 34 drwxrwxrwt 3 root mail 6 Jul 6 02:19 ./ drwxr-xr-x 47 root sys 47 May 30 02:52 ../ drwxrwxr-x 2 root mail 2 May 29 23:47 :saved/ -rw-rw 1 dclarke mail 0 Jul 5 16:01 dclarke -rw-rw 1 lucone mail 657 Jul 2 17:55 lucone -rw-rw 1 root mail8705 Jul 2 05:40 root # I shutdown the dovecot service on Solaris via svcadm and saw this in the error log : Jul 06 13:00:50 dovecot: Warning: I/O leak: 806ef10 (36) Jul 06 13:00:50 dovecot: Warning: I/O leak: 806ef10 (30) Jul 06 13:00:50 dovecot: Warning: I/O leak: 806ef10 (26) Jul 06 13:00:50 dovecot: Warning: I/O leak: 806ef10 (24) I'm not too sure what that means but it doesn't look good. Here is some info : # /opt/csw/sbin/dovecot -n # 1.2.12: /etc/opt/csw/dovecot/dovecot.conf Warning: fd limit 256 is lower than what Dovecot can use under full load (more than 640). Either grow the limit or change login_max_processes_count and max_mail_processes settings # OS: SunOS 5.10 i86pc zfs base_dir: /var/opt/csw/dovecot/ log_path: /var/opt/csw/dovecot/error.log info_log_path: /var/opt/csw/dovecot/message.log protocols: imap pop3 imaps pop3s ssl: required ssl_cert_file: /etc/opt/csw/dovecot/certs/dovecot.pem ssl_key_file: /etc/opt/csw/dovecot/private/dovecot.pem ssl_cipher_list: ALL:!LOW:!SSLv2 verbose_ssl: yes login_dir: /var/opt/csw/dovecot//login login_executable(default): /opt/csw/libexec/dovecot/imap-login login_executable(imap): /opt/csw/libexec/dovecot/imap-login login_executable(pop3): /opt/csw/libexec/dovecot/pop3-login login_greeting: Dovecot from Blastwave.org is ready. login_max_processes_count: 64 verbose_proctitle: yes mail_location: mbox:/var/mail/%u mail_debug: yes mail_executable(default): /opt/csw/libexec/dovecot/imap mail_executable(imap): /opt/csw/libexec/dovecot/imap mail_executable(pop3): /opt/csw/libexec/dovecot/pop3 mail_plugin_dir(default): /opt/csw/lib/dovecot/imap mail_plugin_dir(imap): /opt/csw/lib/dovecot/imap mail_plugin_dir(pop3): /opt/csw/lib/dovecot/pop3 auth default: passdb: driver: pam userdb: driver: passwd Any thoughts before I enter the infinite loop noted at the top of this message ? :-) -- Dennis
Re: [Dovecot] dovecot-1.2.12 fails to build on Solaris 8 and 10
On Sun, Jul 04, 2010 at 05:56:54PM -0400, Dennis Clarke wrote: Makes me wonder. However my simple test code works fine. Something very fishy going on here. Just as a datapoint. Here is 1.2.12 on opensolaris 2009.06 While that is very far away from production Solaris 8 it is a valid datapoint. However I did manage to find the ID-Ten-T error that I was having related to a CFLAGS switch of all things. # uname -a SunOS titan 5.8 Generic_127722-03 i86pc i386 i86pc # /opt/csw/sbin/dovecot --version 1.2.12 # /opt/csw/sbin/dovecot --build-options Build options: ioloop=poll ipv6 openssl Mail storages: maildir mbox raw shared SQL drivers: Passdb: checkpassword pam passwd passwd-file shadow Userdb: passwd passwd-file prefetch static If you wonder why it is compiled on a ten year old OS it is because of the Solaris ABI ( Application Binary Interface ) stability in which *anything* that compiles and runs on Solaris 8 will run everywhere else flawlessly. However, your dovecot compiled on that OpenSolaris beta release won't run anywhere else but on that machine. In fact, no promises on that[1] OpenSolaris release either. :-) Dennis [1] keep in mind that there ha never been a prod ready release of OpenSolaris. Maybe someday but Oracle isn't talking. :-(
[Dovecot] Jul 06 00:06:15 dict: Error: dict client: Broken handshake
After building and install dovecot I then made my own self signed SSL certs and placed them carefully into the correct places : Thus : # grep -v ^# dovecot-openssl.cnf | grep -v ^$ [ req ] default_bits = 1024 encrypt_key = yes distinguished_name = req_dn x509_extensions = cert_type prompt = no [ req_dn ] C=CA ST=Ontario L=Toronto O=Blastwave OU=IMAP server CN=titan.blastwave.org emailaddress=postmas...@blastwave.org [ cert_type ] nsCertType = server # /opt/csw/bin/openssl req -new -x509 -nodes -config ./dovecot-openssl.cnf -out /etc/opt/csw/dovecot/certs/dovecot.pem -keyout /etc/opt/csw/dovecot/private/dovecot.pem -days 365 Generating a 1024 bit RSA private key ...++ ++ writing new private key to '/etc/opt/csw/dovecot/private/dovecot.pem' - # ls -l /etc/opt/csw/dovecot/certs/dovecot.pem /etc/opt/csw/dovecot/private/dovecot.pem -rw-r--r-- 1 root other 1050 Jul 5 23:58 /etc/opt/csw/dovecot/certs/dovecot.pem -rw-r--r-- 1 root other887 Jul 5 23:58 /etc/opt/csw/dovecot/private/dovecot.pem # I then checked things out with a dovecot -n : # /opt/csw/sbin/dovecot -n # 1.2.12: /etc/opt/csw/dovecot/dovecot.conf Warning: fd limit 256 is lower than what Dovecot can use under full load (more than 576). Either grow the limit or change login_max_processes_count and max_mail_processes settings # OS: SunOS 5.8 i86pc base_dir: /var/opt/csw/dovecot/ log_path: /var/opt/csw/dovecot/error.log info_log_path: /var/opt/csw/dovecot/message.log ssl_cert_file: /etc/opt/csw/dovecot/certs/dovecot.pem ssl_key_file: /etc/opt/csw/dovecot/private/dovecot.pem login_dir: /var/opt/csw/dovecot login_executable: /opt/csw/libexec/dovecot/imap-login login_greeting: Dovecot from Blastwave.org is ready. login_max_processes_count: 64 mail_location: mbox:%h/mail:INBOX=/var/mail/%u auth default: passdb: driver: pam userdb: driver: passwd Other than the warning it looks ready to run. I fire up the init script : # /etc/opt/csw/init.d/cswdovecot start dovecot service starting. ps -ef confirms that things are happening : # ps -ef UID PID PPID CSTIME TTY TIME CMD root 0 0 0 Jun 15 ?0:16 sched root 1 0 0 Jun 15 ?0:24 /etc/init - root 2 0 0 Jun 15 ?0:00 pageout root 3 0 0 Jun 15 ? 10:37 fsflush root 331 1 0 Jun 15 ?0:00 /usr/lib/saf/sac -t 300 root 334 331 0 Jun 15 ?0:00 /usr/lib/saf/ttymon root 153 1 0 Jun 15 ?0:00 /usr/sbin/rpcbind root 392 196 0 Jun 16 ?0:00 in.telnetd root75 1 0 Jun 15 ?0:00 /usr/lib/sysevent/syseventd root 149 1 0 Jun 15 ?0:43 /usr/lib/ldap/ldap_cachemgr root 199 1 0 Jun 15 ?0:00 /usr/lib/autofs/automountd root 156 1 0 Jun 15 ?0:00 /usr/sbin/keyserv root 196 1 0 Jun 15 ?0:00 /usr/sbin/inetd -s daemon 198 1 0 Jun 15 ?0:00 /usr/lib/nfs/statd root 234 1 0 Jun 15 ?0:08 /usr/lib/inet/xntpd root 226 1 0 Jun 15 ?0:00 /usr/sbin/cron root 202 199 0 Jun 15 ?0:42 /usr/lib/autofs/automountd root 194 1 0 Jun 15 ?0:00 /usr/lib/nfs/lockd root 239 1 0 Jun 15 ?0:00 /usr/lib/lpsched root 266 265 0 Jun 15 ?0:00 /usr/sbin/cs00 root 251 1 0 Jun 15 ?0:00 /usr/lib/power/powerd root 265 1 0 Jun 15 ?0:00 /usr/sbin/cssd root 225 1 0 Jun 15 ?0:00 /usr/sbin/syslogd root 267 1 0 Jun 15 ?0:00 /usr/bin/fgd root 264 1 0 Jun 15 ?0:00 /usr/lib/utmpd root 290 288 0 Jun 15 ?0:00 htt_server -port 9010 -syslog -message_locale C root 275 1 0 Jun 15 ?0:00 /usr/lib/locale/ja/wnn/dpkeyserv root 281 1 0 Jun 15 ?0:00 /usr/lib/locale/ja/wnn/jserver root 282 281 0 Jun 15 ?0:00 /usr/lib/locale/ja/wnn/jserver_m root 288 1 0 Jun 15 ?0:00 /usr/lib/im/htt -port 9010 -syslog -message_locale C root 297 1 0 Jun 15 ?0:00 /usr/lib/locale/ja/atokserver/atokmngdaemon root 320 317 0 Jun 15 ?6:33 mibiisa -r -p 3200 root 5500 1 0 Jun 16 console 0:00 -sh root 317 1 0 Jun 15 ?0:00 /usr/lib/snmp/snmpdx -y -c /etc/snmp/conf root 325 1 0 Jun 15 ?0:00 /usr/lib/dmi/snmpXdmid -s titan root 324 1 0 Jun 15 ?0:00 /usr/lib/dmi/dmispd root 11633 11624 0 00:02:40 ?0:00 dovecot-auth -w root 11624 1 1 00:02:39 ?0:00 /opt/csw/sbin/dovecot -c /etc/opt/csw/dovecot/dovecot.conf sysadmin 394 392 0 Jun 16 pts/10:09 -sh root 10182 5500 0 Jun 19 console 0:00 /opt/csw/bin/bash dclarke 10759 10717 0 18:02:33 pts/20:01
Re: [Dovecot] Jul 06 00:06:15 dict: Error: dict client: Broken handshake
On 6.7.2010, at 1.15, Dennis Clarke wrote: base_dir: /var/opt/csw/dovecot/ login_dir: /var/opt/csw/dovecot These must not be the same. Just use the default for login_dir, it defaults under base_dir. (I thought I already added a check against this .. oh well, v2.0 won't let you specify login_dir at all anymore.) fixed ! awesome software ... now I'll test this in a production machine and see how things go :-) -- Dennis
[Dovecot] dovecot-1.2.12 fails to build on Solaris 8 and 10
configure fails with : configure: error: Unsupported off_t type I am not too sure what off_t is needed but this is a pretty standard thing. My configure line : ./configure --build=i386-pc-solaris2.10 --host=i386-pc-solaris2.10 --prefix=/opt/csw --with-zlib --with-ssl=openssl --with-storages=mbox,maildir --with-libiconv-prefix=/opt/csw Yes, the openssl libs are in /opt/csw/lib Not sure what Dovecot needs to build or what other options are required. Any thoughts would be appreciated. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris
Re: [Dovecot] dovecot-1.2.12 fails to build on Solaris 8 and 10
) to FILE f sizeof off_t = 4 $ $ ls -lap total 80 drwxr-xr-x 2 sysadmin sysadmin 277 Jul 4 21:43 ./ drwxrwxrwt 7 root sys 538 Jul 4 21:42 ../ -rwx-- 1 sysadmin sysadmin 311 Jul 4 21:33 compile.sh -rw-r--r-- 1 sysadmin sysadmin 2 Jul 4 21:42 conftestval -rwxr-xr-x 1 sysadmin sysadmin9424 Jul 4 21:42 sizeof -rw-r--r-- 1 sysadmin sysadmin1626 Jul 4 21:41 sizeof.c -rw-r--r-- 1 sysadmin sysadmin6152 Jul 4 21:42 sizeof.o $ cat conftestval 4 So I'm at a loss why the dovecot configure script fails due to anything regarding off_t. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris
Re: [Dovecot] dovecot-1.2.12 fails to build on Solaris 8 and 10
Have you read the compilation wiki http://wiki.dovecot.org/CompilingSource ? Yep .. been there. This is configure that fails. We never even get to make. :-( I even tried a real simple ./configure and that fails at the same point. I just posted a really verbose test to the maillist also that confirms that the conf.c code should work as is. -- Dennis Clarke dcla...@opensolaris.ca - Email related to the open source Solaris
Re: [Dovecot] dovecot-1.2.12 fails to build on Solaris 8 and 10
Interesting, it is almost like the fopen function fails to open conftestval for write. Makes me wonder. However my simple test code works fine. Something very fishy going on here. Dennis