[Dovecot] v1.1.8 released
http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz.sig
Most importantly mbox bugfixes. v1.1 should finally be as stable with mboxes as it was with v1.0. Hopefully we'll also soon have the first v1.2 beta release and the final v1.2.0 somewhat soon after that.
- mbox: Several bugfixes. Fixes "next message unexpectedly lost"
errors and perhaps some other problems as well.
- deliver: It wasn't possible to override boolean settings in
lda section by setting them to "no".
- Maildir++ quota didn't correctly check if maildirs had changed
during recalculation.
- kqueue notify: Fixed assert-crash in some situations
- dbox: Several fixes to handling Maildir migrations
- Logging/error message improvementsTimo,
Many months back you recommend that, when compiling on Mac OS X systems, that we pass the '--with-notify=kqueue' swith to 'configure'. For many, many months we've been configuring using this command...
./configure --with-ssldir=/System/Library/OpenSSL --with-ssl=openssl --with-notify=kqueue
...and all has been well.
Do you still consider that the appropriate command line for OS X systems?
Thank you, B. Bodger
On Wed, 2009-01-07 at 14:54 -0600, Bruce Bodger wrote:
Timo,
Many months back you recommend that, when compiling on Mac OS X systems, that we pass the '--with-notify=kqueue' swith to 'configure'. For many, many months we've been configuring using this command...
./configure --with-ssldir=/System/Library/OpenSSL --with-ssl=openssl --with-notify=kqueue
...and all has been well.
Do you still consider that the appropriate command line for OS X systems?
It's not necessary since it's used by default anyway (verify that it says "File change notification method .. kqueue" after configure).
On 1/7/09 3:02 PM, Timo Sirainen wrote:
Do you still consider that the appropriate command line for OS X systems?
It's not necessary since it's used by default anyway (verify that it says "File change notification method .. kqueue" after configure).
It did indeed. Thank you.
I just compiled 1.1.8. Should there be any concern regarding these errors in config.log...
configure:8771: checking how to run the C++ preprocessor configure:8807: g++ -E conftest.cpp configure:8813: $? = 0 configure:8844: g++ -E conftest.cpp conftest.cpp:21:28: error: ac_nonexistent.h: No such file or directory configure:8850: $? = 1 configure: failed program was: | /* confdefs.h. */ | #define PACKAGE_NAME "dovecot" | #define PACKAGE_TARNAME "dovecot" | #define PACKAGE_VERSION "1.1.8" | #define PACKAGE_STRING "dovecot 1.1.8" | #define PACKAGE_BUGREPORT "dovecot@dovecot.org" | #define PACKAGE "dovecot" | #define VERSION "1.1.8"
AND
configure:16884: checking how to link with libiconv configure:16886: result: -liconv configure:16897: checking for iconv declaration configure:16936: gcc -c -g -O2 conftest.c >&5 conftest.c:31: error: conflicting types for 'libiconv' /usr/include/iconv.h:82: error: previous declaration of 'libiconv' was here configure:16942: $? = 1 configure: failed program was: | /* confdefs.h. */ | #define PACKAGE_NAME "dovecot" | #define PACKAGE_TARNAME "dovecot" | #define PACKAGE_VERSION "1.1.8" | #define PACKAGE_STRING "dovecot 1.1.8"
AND
configure:21053: checking for uoff_t configure:21083: gcc -c -std=gnu99 -g -O2 -Wall -W -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wchar-subscripts -Wformat=2 -Wbad-function-ca st -Wstrict-aliasing=2 conftest.c >&5 conftest.c:105: error: parse error before 'ac__type_new_' conftest.c:105: warning: type defaults to 'int' in declaration of 'ac__type_new_' conftest.c:105: warning: data definition has no type or storage class conftest.c: In function 'main': conftest.c:109: error: parse error before ')' token
AND quite a few others pointing to confdefs.h
Thank you, B. Bodger
On Wed, 2009-01-07 at 15:31 -0600, Bruce Bodger wrote:
On 1/7/09 3:02 PM, Timo Sirainen wrote:
Do you still consider that the appropriate command line for OS X systems?
It's not necessary since it's used by default anyway (verify that it says "File change notification method .. kqueue" after configure).
It did indeed. Thank you.
I just compiled 1.1.8. Should there be any concern regarding these errors in config.log...
I think they're all normal. They're just figuring out what the build system looks like by testing what programs compile and what don't.
Hello Timo!
Looks like the performance problems on opening/delivering mbox mailboxes and index updates/usage have been solved with dovecot 1.1.8 (your inbox subject was dovecot 1.1rc7 and cache with deliver). http://www.mail-archive.com/dovecot@dovecot.org/msg08274.html
Is this quite logically with the changes you made in 1.1.8?
Ciao, Gerhard
On Wed, 7 Jan 2009, Timo Sirainen wrote:
http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz.sig
Most importantly mbox bugfixes. v1.1 should finally be as stable with mboxes as it was with v1.0. Hopefully we'll also soon have the first v1.2 beta release and the final v1.2.0 somewhat soon after that.
- mbox: Several bugfixes. Fixes "next message unexpectedly lost" errors and perhaps some other problems as well.
- deliver: It wasn't possible to override boolean settings in lda section by setting them to "no".
- Maildir++ quota didn't correctly check if maildirs had changed during recalculation.
- kqueue notify: Fixed assert-crash in some situations
- dbox: Several fixes to handling Maildir migrations
- Logging/error message improvements
On Thu, 2009-01-08 at 13:09 +0100, Gerhard Wiesinger wrote:
Looks like the performance problems on opening/delivering mbox mailboxes and index updates/usage have been solved with dovecot 1.1.8 (your inbox subject was dovecot 1.1rc7 and cache with deliver). http://www.mail-archive.com/dovecot@dovecot.org/msg08274.html
That's good.
Is this quite logically with the changes you made in 1.1.8?
I'm not sure :) There were several different correctness problems I fixed that I think could have caused all kinds of problems.
On Wed, Jan 07, 2009 at 03:45:28PM -0500, Timo Sirainen wrote:
http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz.sig
Most importantly mbox bugfixes. v1.1 should finally be as stable with mboxes as it was with v1.0. Hopefully we'll also soon have the first v1.2 beta release and the final v1.2.0 somewhat soon after that.
Compiling Dovecot 1.1.8 under Solaris 8 using gcc 3.4.6 gives a final link error on dovecot-auth. Apparantely, unsetenv() isn't a universally available function?
libtool: link: gcc -std=gnu99 -O0 -Wall -W -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wchar-subscripts -Wformat=2 -Wbad-function-cast -o dovecot-auth auth.o auth-cache.o auth-client-connection.o auth-master-connection.o auth-master-listener.o auth-request.o auth-request-handler.o auth-stream.o auth-worker-client.o auth-worker-server.o db-sql.o db-passwd-file.o main.o mech.o mech-anonymous.o mech-plain.o mech-login.o mech-cram-md5.o mech-digest-md5.o mech-ntlm.o mech-otp.o mech-skey.o mech-rpa.o mech-apop.o mech-winbind.o otp-skey-common.o plain-common.o passdb.o passdb-blocking.o passdb-bsdauth.o passdb-cache.o passdb-checkpassword.o passdb-passwd.o passdb-passwd-file.o passdb-pam.o passdb-shadow.o passdb-sia.o passdb-vpopmail.o passdb-sql.o userdb.o userdb-blocking.o userdb-nss.o userdb-passwd.o userdb-passwd-file.o userdb-prefetch.o userdb-static.o userdb-vpopmail.o userdb-sql.o mech-gssapi.o db-ldap.o passdb-ldap.o userdb-ldap.o libpassword.a ../lib-settings/libsettings.a ../lib-ntlm/libntlm.a ../lib-otp/libotp.a ../lib-sql/libsql.a ../lib/liblib.a -lpam /usr/local/lib/libsqlite3.so -lz -ldl -lrt -lnsl -lsocket -R/usr/local/lib -R/usr/local/lib Undefined first referenced symbol in file unsetenv ../lib/liblib.a(env-util.o) ld: fatal: Symbol referencing errors. No output written to dovecot-auth collect2: ld returned 1 exit status
-- Dean Brooks dean@iglou.com
On Thu, 2009-01-08 at 11:00 -0500, Dean Brooks wrote:
On Wed, Jan 07, 2009 at 03:45:28PM -0500, Timo Sirainen wrote:
http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz.sig
Most importantly mbox bugfixes. v1.1 should finally be as stable with mboxes as it was with v1.0. Hopefully we'll also soon have the first v1.2 beta release and the final v1.2.0 somewhat soon after that.
Compiling Dovecot 1.1.8 under Solaris 8 using gcc 3.4.6 gives a final link error on dovecot-auth. Apparantely, unsetenv() isn't a universally available function?
Well, that's annoying. I was worried about that when I added it, but I couldn't find an OS where it wasn't supported so I thought it was safe. It is in Solaris 10 anyway. You could just comment out the call, it was added just to fix this bug:
- deliver: It wasn't possible to override boolean settings in
lda section by setting them to "no".I'm not really sure how this could be handled best. I guess it would be possible to implement our own environ variable handling but that would be pretty ugly.. Or perhaps I could modify the deliver config parsing code once again so that it wouldn't need to unset the environment.. With the upcoming v1.3 none of this would be necessary..
On Thu, Jan 08, 2009 at 11:15:55AM -0500, Timo Sirainen wrote:
On Thu, 2009-01-08 at 11:00 -0500, Dean Brooks wrote:
On Wed, Jan 07, 2009 at 03:45:28PM -0500, Timo Sirainen wrote:
http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz.sig
Most importantly mbox bugfixes. v1.1 should finally be as stable with mboxes as it was with v1.0. Hopefully we'll also soon have the first v1.2 beta release and the final v1.2.0 somewhat soon after that.
Compiling Dovecot 1.1.8 under Solaris 8 using gcc 3.4.6 gives a final link error on dovecot-auth. Apparantely, unsetenv() isn't a universally available function?
I'm not really sure how this could be handled best. I guess it would be possible to implement our own environ variable handling but that would be pretty ugly.. Or perhaps I could modify the deliver config parsing code once again so that it wouldn't need to unset the environment.. With the upcoming v1.3 none of this would be necessary..
I found similar code implementing unsetenv() in a perl module that seems to do the same thing. Completely untested, but example follows.
void env_remove(const char *name) { int name_len; extern char **environ; char **envp;
name_len = strlen(name);
for (envp = environ; *envp != NULL; envp++) {
if (strncmp(name, *envp, name_len) == 0 &&
(*envp)[name_len] == '=') {
free(*envp);
do {
envp[0] = envp[1];
} while (*envp++);
break;
}
}}
-- Dean Brooks dean@iglou.com
On Thu, 2009-01-08 at 11:38 -0500, Dean Brooks wrote:
I found similar code implementing unsetenv() in a perl module that seems to do the same thing. Completely untested, but example follows.
Hmm. Maybe I could add that to compat.c..
void env_remove(const char *name) { int name_len; extern char **environ; char **envp;
name_len = strlen(name); for (envp = environ; *envp != NULL; envp++) { if (strncmp(name, *envp, name_len) == 0 && (*envp)[name_len] == '=') { free(*envp);
I think the free() here shouldn't be done, because Dovecot allocates the memory from memory pools.
On Thu, 2009-01-08 at 11:45 -0500, Timo Sirainen wrote:
On Thu, 2009-01-08 at 11:38 -0500, Dean Brooks wrote:
I found similar code implementing unsetenv() in a perl module that seems to do the same thing. Completely untested, but example follows.
Hmm. Maybe I could add that to compat.c..
On Jan 8 2009, Dean Brooks wrote:
I found similar code implementing unsetenv() in a perl module that seems to do the same thing. Completely untested, but example follows.
void env_remove(const char *name) { int name_len; extern char **environ; char **envp;
name_len = strlen(name); for (envp = environ; *envp != NULL; envp++) { if (strncmp(name, *envp, name_len) == 0 && (*envp)[name_len] == '=') { free(*envp); do { envp[0] = envp[1]; } while (*envp++); break; } } }
At the UofMN we tried some things like this that didn't seem to quite do the trick, so in the end we just built on a Solaris 10 machine with proper unsetenv and ran that binary on 9. Might work for 8 too. Not recommended, but.... :)
-- Brian Hayden Minnesota Supercomputing Institute University of Minnesota
Timo Sirainen wrote:
http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz http://dovecot.org/releases/1.1/dovecot-1.1.8.tar.gz.sig
I refreshed the ManageSieve patch:
http://www.rename-it.nl/dovecot/1.1/dovecot-1.1.8-managesieve-0.10.4.diff.gz http://www.rename-it.nl/dovecot/1.1/dovecot-1.1.8-managesieve-0.10.4.diff.gz...
Regards,
Stephan.
participants (6)
-
Brian Hayden
-
Bruce Bodger
-
Dean Brooks
-
Gerhard Wiesinger
-
Stephan Bosch
-
Timo Sirainen