[vchkpw] vdelolduser question

2003-11-14 Thread Tom Walsh
Recently I tried to remove the auth-logging feature from our vpopmail
configuration.

First some background: I had determined that our configuration was still
very master database dependant which is not one of our project goals.
Basically we have local databases on each client in the mail cluster we are
running. It works great because the master replicates the data to the
slaves, so we don't have to worry about moving the data around. Our goal was
to try and provide a reasonably available mail cluster for our clients, in
that if the master was down (which handles mysql and nfs duties) the mail
servers would queue messages and wait for the nfs to come backup. The
problem we encountered was that with the vpopmail options we had used
required constant update or insert queries (which are handled in the
libvpopmail and sent to the master server because of the replication
settings we used in vpopmail). So if the master was down... we had timeout
issues with pop3 authentication.

So I tried to recompile vpopmail without mysql-logging and auth-logging, but
ran into a problem when compiling vpopmail 5.3.20 under FreeBSD 4.8.

vdelolduser.c complained about an undefined function vget_lastauth and
failed to compile (with the option --enable-auth-logging=n).

Now... My question is this:

I modified the vdelolduser.c and moved the #endif that had commented out the
main() section of the vdelolduser.c if ENABLE_AUTH_LOGGING was not defined
to the end of the file... So basically... I made it so that if
ENABLE_AUTH_LOGGING wasn't enabled, none of the functions or the main
section of vdeloldusers would be compiled.

Would that have any negative effects? Other than the obvious that
vdelolduser is a useless binary.

We don't have a need for that anyways... So that is why I am asking here to
see if it has any internal uses that I don't see.

Any help or guidance is appreciated.

Tom Walsh
Network Administrator
http://www.ala.net/




Re: [vchkpw] vdelolduser question

2003-11-14 Thread Michael Bowe
You are correct, that is a bug in the vdeloldusers.c coding which is exposed
if
you

   ./configure --enable-mysql=y --enable-auth-logging=n
   make

Your proposed workaround of moving the ifdef will be fine

I will fix the problem in CVS

Michael.

- Original Message - 
From: Tom Walsh [EMAIL PROTECTED]
To: [EMAIL PROTECTED]
Sent: Saturday, November 15, 2003 2:15 AM
Subject: [vchkpw] vdelolduser question


 Recently I tried to remove the auth-logging feature from our vpopmail
 configuration.

 First some background: I had determined that our configuration was still
 very master database dependant which is not one of our project goals.
 Basically we have local databases on each client in the mail cluster we
are
 running. It works great because the master replicates the data to the
 slaves, so we don't have to worry about moving the data around. Our goal
was
 to try and provide a reasonably available mail cluster for our clients, in
 that if the master was down (which handles mysql and nfs duties) the mail
 servers would queue messages and wait for the nfs to come backup. The
 problem we encountered was that with the vpopmail options we had used
 required constant update or insert queries (which are handled in the
 libvpopmail and sent to the master server because of the replication
 settings we used in vpopmail). So if the master was down... we had timeout
 issues with pop3 authentication.

 So I tried to recompile vpopmail without mysql-logging and auth-logging,
but
 ran into a problem when compiling vpopmail 5.3.20 under FreeBSD 4.8.

 vdelolduser.c complained about an undefined function vget_lastauth and
 failed to compile (with the option --enable-auth-logging=n).

 Now... My question is this:

 I modified the vdelolduser.c and moved the #endif that had commented out
the
 main() section of the vdelolduser.c if ENABLE_AUTH_LOGGING was not defined
 to the end of the file... So basically... I made it so that if
 ENABLE_AUTH_LOGGING wasn't enabled, none of the functions or the main
 section of vdeloldusers would be compiled.

 Would that have any negative effects? Other than the obvious that
 vdelolduser is a useless binary.

 We don't have a need for that anyways... So that is why I am asking here
to
 see if it has any internal uses that I don't see.

 Any help or guidance is appreciated.

 Tom Walsh
 Network Administrator
 http://www.ala.net/