gmane.mail.exim.user

Re: Exim 4.80 RC5 uploaded

Phil Pennock — 2012-05-25T14:56:16

Yeah, I wasn't aware there was a fight and so naively picked the better
option.

I'll toggle the default value of the variable in the build script now
and disable .lz generation.

-Phil

Re: Exim4 Satellite Mode for two smtp servers

Andrew — 2012-05-25T14:28:12

To do this you will need to configure exim using exim4.conf ( see http://www.exim.org/exim-html-current/doc/html/spec_html/ for more information ) rather than using the debian specific configuration you have included.


This information looks very wrong for anything that would send email that would be accepted (unless you really have registered mydomain.com the genuine registered owner of that domain would probably get quite upset)

If you want to continue using the debian config options in update-exim4.conf.conf you would be best off asking the debian specific list which you can find more information by looking at the README file ( zless /usr/share/doc/exim4-config/README.Debian.gz )

 -Andrew

Re: Exim 4.80 RC5 uploaded

Nigel Metheringham — 2012-05-25T11:26:02


Phil Pennock wrote:

Yay - those bikesheds won't paint themselves!


This feels sort of 90s with everyone suddenly using their own packaging
formats before going to the tar.gz standard.


My feeling is that neither of these are compelling in any way - yes they
are better than bz2 - just - but it would take 31 downloads of that data
in .lz format to save the size of one bz2 download.

Heres a google docs SS with the figures for the RC5 distribution set:-

https://docs.google.com/spreadsheet/pub?key=0Ah5Bo78-JMXqdFJKeHhnSzZvZFk5RUQ1LUlDVEhvaGc&output=html
or smaller URL
  http://goo.gl/4FqSF


I confirm lz as smaller than xz and both smaller than bz2 -


My feeling is lets keep out of this particular fight until a winner has
emerged.  bz2 was a clear winner over gz, but neither of the new ones
are compelling on size.  If they become compelling for other reasons
then we can go for it then...  For now I'd prefer to watch from the
sidelines and not support either.

But this is all a personal opinion...

Nigel.

Re: Exim 4.80 RC5 uploaded

Phil Pennock — 2012-05-25T09:29:37

I've had private mail on this topic too, making it the most contested
feature of the new Exim release.  That ... says something.

GnuTLS and some other projects now ship only .xz and .lz tarballs, with
the .lz being smaller.

I'm not going to add *two* new compression formats.  It's not even
certain that I'll add one.

When looking, I picked the one which gave smaller files for Exim's main
source tarball.  The only reason to use Yet Another compression format
is to have smaller resulting files.

If we add a new compression format, it will be the one which compresses
better.  Compatibility is not an issue, since gzip/bzip2 provide
compatibility.  (Licensing would be an issue if any contender were
antagonistic to the GPL, which Exim uses).

I note from Wikipedia that coreutils uses .gz and .xz and the Linux
kernel supports .xz, but when I look at Linus' tree, I see that lzma is
handled too.  The gentoo claim is about .lzma being replaced by .xz, but
.lzma is not the same as .lz.  My source for this claim is:

Re: [exim-dev] Exim 4.80 RC5 uploaded

Phil Pennock — 2012-05-25T09:12:08

By the way, depending upon how your group memberships and sasldb2 file
ownership are set up, you might check the Exim Specification for an
example "cyrusless_crammd5" which shows how to combine Exim's cram_md5
driver with the new dbmjz lookup type added in 4.80, to access the
password from sasldb2 directly.

-Phil

Re: [exim-dev] Exim 4.80 RC5 uploaded

Phil Pennock — 2012-05-25T09:09:25

Fixed, thanks, and sorry.

I use SASL extensively and had expected that if I broke this I'd have
noticed, but had forgotten that since I wrote the heimdal_gssapi auth
driver I didn't have any Cyrus SASL drivers left, and I haven't yet
figured out a decent test for this to go in the test suite.

What happens is that SASL, in some mechanisms, is able to negotiate a
"protection layer".  Actually using SASL protection layers is not very
widely supported by most software, because the world moved on and uses
SSL/TLS instead.

I adjusted the Cyrus SASL integration to tell Cyrus about external
protection from TLS and to check afterwards to see if SASL had
negotiated its own protection.  If SASL has, Exim declares a failure
because Exim won't do the wrap/unwrap needed.

Wolfgang's fix is correct and has been applied.

-Phil

Re: Quota Limits - Setting Greater Amounts

Brian Spraker — 2012-05-24T23:26:53


--- On Tue, 5/22/12, Phil Pennock <exim-users< at >spodhuis.org> wrote:


I use the Horde Webmail 4 system on the Linux server - and I've not seen a spot where you can set it to receive IMAP alerts or anything?

When the user is fully maxed at their quote and they try to send a message, they do get an error that they are over their quota at that point, however.

But, the idea is to set it so they are at least informed that they are getting close to their quota - before they hit it.  Because once they hit it, their mail is rejected - which means they are losing out on incoming e-mails.

I set the quota inclusive setting to 'true' for now - so hopefully that will help.

Brian S.

Re: [exim-dev] Exim 4.80 RC5 uploaded

Wolfgang Breyha — 2012-05-24T21:08:38

I think, I found the cause...
http://bugs.exim.org/show_bug.cgi?id=1254

Greetings, Wolfgang

Re: Tricky DNS servers [OT]

Jonas Eckerman — 2012-05-24T16:19:22



My guess is that the speed difference between Python and Perl depends 
more on the design of the server than the choice of language.

Perl can be pretty fast when you have persistent code (including forked 
provesses), wich you have in a server.

Regards
/Jonas

Re: [exim-dev] Exim 4.80 RC5 uploaded

Wolfgang Breyha — 2012-05-24T15:50:35

Wolfgang Breyha wrote, on 24.05.2012 16:11:

Sorry, not SASL_PLAIN. It's CRAM-MD5 actually, or cyrus-sasl at all if TLS/SSL
is in use. Source of src/auth/cyrus_sasl.c changed and Changelog mentions
that, but not that it may break existing configurations? Bug or feature? ;-)

Only a wild guess reading the source:
src/auths/cyrus_sasl.c:259ff
sets EXTERNAL_SFF to the value of tls_bits if TLS/SSL is in use.

And in
:410ff
   if (negotiated_ssf > 0)
is checked... does cyrus-sasl return tls_bits here and exim fails selflarting
itself?

Greetings, Wolfgang

Re: Exim 4.80 RC5 uploaded

Nigel Metheringham — 2012-05-24T15:47:32

I've put that change into the git tree:-

http://git.exim.org/exim.git/commitdiff/a3c1395faebdb088bcef9cdb55bb42a791433ccd

Re: Exim 4.80 RC5 uploaded

Todd Lyons — 2012-05-24T15:08:23

Confirmed for CentOS 5.x that I no longer need the extra CFLAGS.
Built and deployed RC5 on one of my production servers.  RC4 ran for a
full day with no issues to report. I will yell if there are any issues
that pop up with RC5.

To the other poster, I don't use SASL_PLAIN so I am unable to offer
any confirmation of the issue.

...Todd

Re: Tricky DNS servers [OT]

Marc Perkel — 2012-05-24T15:04:56


On 5/20/2012 9:57 AM, Christof Meerwald wrote:

Oh really! Definitely need to look at that.

Thanks

Re: Tricky DNS servers [OT]

Marc Perkel — 2012-05-24T15:02:03


On 5/20/2012 8:25 AM, Jonas Eckerman wrote:

Thanks - this looks useful. Would python be faster?

Re: Exim 4.80 RC5 uploaded

A J Thew — 2012-05-24T14:20:37

Great. Many thanks

--
Alan Thew

Re: [exim-dev] Exim 4.80 RC5 uploaded

Wolfgang Breyha — 2012-05-24T14:11:49

Hi!

Phil Pennock wrote, on 24.05.2012 07:32:

I tried to activate RC5 and fail badly with
if I try to use SASL_PLAIN auth? The exactly same config works with 4.77.

Any ideas what goes wrong?

Greetings, Wolfgang

Re: Exim 4.80 RC5 uploaded

Dennis Davis — 2012-05-24T09:48:00


...


Just tried this.  Seen this before with gcc 2.95.2.  It *really*
likes to have the declarations before any of the executable
statements.  So apply this simple patch:

*** pdkim.c.orig        Thu May 24 04:43:20 2012
--- pdkim.c     Thu May 24 10:41:22 2012
***************
*** 1383,1392 ****
        char *b = strdup(sig->headernames);
        char *p = b;
        char *q = NULL;
        if (b == NULL) return PDKIM_ERR_OOM;
  
        /* clear tags */
-       pdkim_stringlist *hdrs = ctx->headers;
        while (hdrs != NULL) {
          hdrs->tag = 0;
          hdrs = hdrs->next;
--- 1383,1392 ----
        char *b = strdup(sig->headernames);
        char *p = b;
        char *q = NULL;
+       pdkim_stringlist *hdrs = ctx->headers;
        if (b == NULL) return PDKIM_ERR_OOM;
  
        /* clear tags */
        while (hdrs != NULL) {
          hdrs->tag = 0;
          hdrs = hdrs->next;

Re: Exim 4.80 RC5 uploaded

A J Thew — 2012-05-24T08:59:27

Hi has anyone tried with an old gcc (2.95)? On Solaris 2.9/gcc 2.95.2

I get:

gcc pdkim.c
pdkim.c: In function `pdkim_feed_finish':
pdkim.c:1389: parse error before `*'
pdkim.c:1390: `hdrs' undeclared (first use in this function)
pdkim.c:1390: (Each undeclared identifier is reported only once
pdkim.c:1390: for each function it appears in.)
gmake[2]: *** [pdkim.o] Error 1

Thanks

--
Alan Thew

Re: Exim 4.80 RC5 uploaded

A J Thew — 2012-05-24T08:57:04

Also OK with previous problem RHEL 5 64 bit system

--
Alan Thew

Re: Exim 4.80 RC5 uploaded

René Berber — 2012-05-24T06:16:26

It works with no need for extra parameters to gcc where before it needed 
the "-fgnu89-inline -std=gnu99".


.lz is not being used since .xz came up, and is the more common new format.

Exim 4.80 RC5 uploaded

Phil Pennock — 2012-05-24T05:32:45

I have uploaded Exim 4.80 RC5 to:
        ftp://ftp.exim.org/pub/exim/exim4/test/

There is an attempt to resolve the compiler issues on some versions of
Linux, a fix for OpenBSD's resolver library not having an expected
typedef, and fixes to work on older OpenSSL libraries which lack SNI
support.  Everything else changed is documentation and test suite, so
I'm going to permit myself a little optimism that this will be the last
RC and that the final release can be cut late in the weekend, to await
people on Monday morning.

As an experiment, this time there are three versions of each file;
bzip2, gzip and lzip.  If you like lzip and want it to stay, please let
us know.  The main distribution is only 8% smaller than bzip2 (28%
smaller than gzip), but perhaps this will really add up.  I'm inclined
to put out the 4.80 release with lzip and see what the reaction is.

I'm also thinking that 4.80 might be the last release with SHA1
checksums in the release mails and that afterwards we'll stick to just
SHA256.  If