date:20090429

Re: [Dovecot] questions about compiling Sieve plugin

2009-04-29 Thread Steffen Kaiser


-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

On Tue, 28 Apr 2009, MercuryShadow wrote:


I figured that, as is the case with many Debian packages, the libraries
needed to compile would be included in a dovecot-dev package, but I
checked and there is no such package in the Debian repos.


Get Debian source packages of Dovecot,

cd /tmp
sudo apt-get source dovecot
sudo apt-get build-dep dovecot

cd dir

then:
dpkg-buildpackage -rfakeroot

This command does a lot more, but you will have the compiled Dovecot and 
anything else in the dir.


Bye,

- -- 
Steffen Kaiser

-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.6 (GNU/Linux)

iQEVAwUBSff6mnWSIuGy1ktrAQKB9wgAiFUIDiUhOBgsq6NwktbCokeUSx+o7oah
lO2Tfc7xbDNHzEynmDoc2AwRbsiiNelKLLZgihemlItQtKHJT2fYaUAd7RM1plQ8
c6wbseJ2oy3Gfls9sBQNNQt0dY1Eo5vjGn2J9e2Mx3CUO5VkJA+WWFJT7sVTNZXm
4WUJ8BaoLhwwYb7tRgUu0KyTOIAId6dk2hwSjeI4j6Amr89pdafODmgxULoE+prJ
jk5WWKVVh3lmUNSgpmJOiXpD3DqOIwgP4gWj8RVy3Px5tFCwBma8Rfeh0/UbS3B4
kgSs6ATBTDfaAbgVgxy/rLJvLlFpqeVpf8YearkH8KNEABCVjlnPFg==
=DcKH
-END PGP SIGNATURE-

[Dovecot] Quota/Dict Postgres Trigger

2009-04-29 Thread Warren Volz

I posted the trigger for v1.1 versions of Dovecot on the Wiki (http://wiki.dovecot.org/Quota/Dict 
) and while I understand the comment posted about two process  
inserting at the same time, I'm not sure I understand how this is  
fixed in v1.2 other than via the revised trigger in the Wiki. Does  
someone have a known working trigger that will handle a double insert  
correctly?


Also, out of curiosity why wasn't the code for dict written to do an  
update first and then an insert if this failed? That would eliminate  
the need for this trigger.


Thanks.

-Warren

Re: [Dovecot] Quota/Dict Postgres Trigger

2009-04-29 Thread Timo Sirainen


On Apr 29, 2009, at 3:25 AM, Warren Volz wrote:

I posted the trigger for v1.1 versions of Dovecot on the Wiki (http://wiki.dovecot.org/Quota/Dict 
) and while I understand the comment posted about two process  
inserting at the same time, I'm not sure I understand how this is  
fixed in v1.2 other than via the revised trigger in the Wiki. Does  
someone have a known working trigger that will handle a double  
insert correctly?


I don't think it's possible until PostgreSQL supports INSERT .. ON  
DUPLICATE KEY UPDATE .. like MySQL. Kind of annoying, I like  
PostgreSQL but this feature is really missing from it.


Also, out of curiosity why wasn't the code for dict written to do an  
update first and then an insert if this failed? That would eliminate  
the need for this trigger.


There would still be a race condition. It's still possible that two  
processes do the steps at the exact same time and still finally find  
out that one of the INSERTs fail because they did everything at the  
same time. You could do similar EXCEPTION WHEN unique_violation  
THEN .. check with v1.1 too, although that still breaks in the very  
rare (or practically non-existent) condition if you happen to delete  
the row while another process is inserting. Or something like that.


Anyway the v1.0/v1.1 code was mainly written for MySQL and then I  
realized PostgreSQL requires some more kludgeing to get it working  
properly in all situations and added that code to v1.2.

Re: [Dovecot] Quota/Dict Postgres Trigger

2009-04-29 Thread tomas

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

On Wed, Apr 29, 2009 at 03:38:27AM -0400, Timo Sirainen wrote:
On Apr 29, 2009, at 3:25 AM, Warren Volz wrote:

I posted the trigger for v1.1 versions of Dovecot on the Wiki
(http://wiki.dovecot.org/Quota/Dict) and while I understand the comment
posted about two process inserting at the same time, I'm not sure I
understand how this is fixed in v1.2 other than via the revised trigger in
the Wiki. Does someone have a known working trigger that will handle a
double insert correctly?

I don't think it's possible until PostgreSQL supports INSERT .. ON
DUPLICATE KEY UPDATE .. like MySQL. Kind of annoying, I like PostgreSQL but
this feature is really missing from it.

FWIW, this seems to be the canonical way of dealing with that in PostgreSQL:

http://www.postgresql.org/docs/current/static/plpgsql-control-structures.html#PLPGSQL-UPSERT-EXAMPLE

Also, out of curiosity why wasn't the code for dict written to do an
update first and then an insert if this failed? That would eliminate the
need for this trigger.

There would still be a race condition. It's still possible that two
processes do the steps at the exact same time and still finally find out
that one of the INSERTs fail because they did everything at the same time.
[...]

Right. The trick seems to be to wrap the thing in one plpgsql function
(which wraps the try-to-update-then-insert into one transaction), so the
client doesn't see anything of that.

45 matches

Mail list logo