[Dovecot] 1.0-test73
Two potentially dangerous changes:
- mbox parser should now be able to handle CR+LF linefeeds. Hopefully doesn't break anything.
- Some mail header parser changes. Hopefully doesn't break anything.
And other stuff:
- COPY shouldn't crash anymore with 64bit systems
- If PAM modifies username, keep it
- When auth workers died they leaked file descriptors in dovecot-auth
- Auth workers were leaking memory for each request
- Make MySQL connect abort in 10 seconds if it can't connect
- 32bit UID/GIDs were truncated to 31bit value
- dovecot-auth reports cache hits and misses on SIGUSR2
Timo Sirainen wrote:
Two potentially dangerous changes:
- mbox parser should now be able to handle CR+LF linefeeds. Hopefully doesn't break anything.
- Some mail header parser changes. Hopefully doesn't break anything.
And other stuff:
- COPY shouldn't crash anymore with 64bit systems
- If PAM modifies username, keep it
- When auth workers died they leaked file descriptors in dovecot-auth
- Auth workers were leaking memory for each request
- Make MySQL connect abort in 10 seconds if it can't connect
- 32bit UID/GIDs were truncated to 31bit value
- dovecot-auth reports cache hits and misses on SIGUSR2
This version seems to delete the subscription file.
When I try to change the subscription list (Mozilla 1.7.5 or through Squirrelmail) the file just disappears without a message or error. I'm not perfectly sure, but I think it worked in test72.
Martin
Martin Preen, Universitaet Freiburg, Institut fuer Informatik Georges-Koehler-Allee, Geb.52, Zi.00-006, 79110 Freiburg, Germany
phone: ++49 761 203-8250 preen@informatik.uni-freiburg.de fax: ++49 761 203-8242 www.informatik.uni-freiburg.de
Martin Preen wrote:
Timo Sirainen wrote:
Two potentially dangerous changes:
- mbox parser should now be able to handle CR+LF linefeeds. Hopefully doesn't break anything.
- Some mail header parser changes. Hopefully doesn't break anything.
And other stuff:
- COPY shouldn't crash anymore with 64bit systems
- If PAM modifies username, keep it
- When auth workers died they leaked file descriptors in dovecot-auth
- Auth workers were leaking memory for each request
- Make MySQL connect abort in 10 seconds if it can't connect
- 32bit UID/GIDs were truncated to 31bit value
- dovecot-auth reports cache hits and misses on SIGUSR2
This version seems to delete the subscription file.
Hmm, probably not a bug in dovecot. I think there is a strange network problem here. Creation of the new subscription file fails, due to a NFS problem.
Sorry for the noise.
Martin
When I try to change the subscription list (Mozilla 1.7.5 or through Squirrelmail) the file just disappears without a message or error. I'm not perfectly sure, but I think it worked in test72.
Martin Preen, Universitaet Freiburg, Institut fuer Informatik Georges-Koehler-Allee, Geb.52, Zi.00-006, 79110 Freiburg, Germany
phone: ++49 761 203-8250 preen@informatik.uni-freiburg.de fax: ++49 761 203-8242 www.informatik.uni-freiburg.de
Le 14.06.2005 23:52, Timo Sirainen a écrit :
Two potentially dangerous changes:
- mbox parser should now be able to handle CR+LF linefeeds. Hopefully doesn't break anything.
- Some mail header parser changes. Hopefully doesn't break anything.
And other stuff:
- COPY shouldn't crash anymore with 64bit systems
- If PAM modifies username, keep it
- When auth workers died they leaked file descriptors in dovecot-auth
- Auth workers were leaking memory for each request
- Make MySQL connect abort in 10 seconds if it can't connect
- 32bit UID/GIDs were truncated to 31bit value
- dovecot-auth reports cache hits and misses on SIGUSR2
Do any of these changes affect the 1.0-stable ?
Particularly, does the 1.0-stable version has a problem with COPY on 64bits systems ?
Thanks,
-- Nico L'enfer existe. Il est la part du plus grand nombre. -+- Guillaume Apollinaire -+-
On Wed, 2005-06-15 at 13:31 +0200, Nicolas STRANSKY wrote:
And other stuff:
- COPY shouldn't crash anymore with 64bit systems
- If PAM modifies username, keep it
- When auth workers died they leaked file descriptors in dovecot-auth
- Auth workers were leaking memory for each request
- Make MySQL connect abort in 10 seconds if it can't connect
- 32bit UID/GIDs were truncated to 31bit value
- dovecot-auth reports cache hits and misses on SIGUSR2
Do any of these changes affect the 1.0-stable ?
Particularly, does the 1.0-stable version has a problem with COPY on 64bits systems ?
No, the problem was with keyword code and 1.0-stable doesn't have that. 1.0-stable also doesn't have auth workers. MySQL connect aborting could be backported to it as well as that 32bit UID/GID truncation..
participants (3)
-
Martin Preen
-
Nicolas STRANSKY
-
Timo Sirainen