[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: postfix: possible RC bug?



On Wed, 2006-12-06 at 11:49 +0100, Adrian von Bidder wrote:
> I'm just wondering if #397771 (SASL auth breaks with current postfix + cyrus 
> sasl from testing) shouldn't be RC.  As far as I understand, basically 
> every postfix+sasl set up will break on sarge->etch upgrade.
> 
> (latest bug activity: 25. November)
> 
> Sorry to be unable to help.

No, thanks for being alert! The bug is in fact going to break all
postfix+sasl sarge->etch upgrades.

RC or not, fortunately, the fix is actually pretty trivial, or at least
it's known, but something unrelated to the fix itself doesn't go right.
The problem right now is that portions of code added by 10tls.dpatch
which should get built in when the SASL version is >= 2.1.22, don't get
built in.

Here's the compiler directive that attempts to test that the version
requirement is fulfilled:

#if SASL_VERSION_MAJOR >= 2 && (SASL_VERSION_MAJOR > 2 ||
SASL_VERSION_MINOR >= 1) && (SASL_VERSION_MAJOR > 2 ||
(SASL_VERSION_MINOR > 1 || SASL_VERSION_STEP >= 22))
/* ... 2.1.22-specific code ... */
#endif

That's sprinkled around in xsasl_cyrus_client.c, xsasl_cyrus_common.h,
xsasl_cyrus_log.c and xsasl_cyrus_server.c, all residing in src/xsasl in
the postfix source tree.

However, building postfix 2.3.4-2 results in a binary without the SASL
>= 2.1.22 -specific code. If I remove the version check above from all
places, the result is a working postfix (ie. one that sets the primary
SASL configuration file path to /etc/postfix/sasl). I've tested this
right down to a successful SMTP AUTH session.

Is the version check broken? I've read the patch repeatedly, and have
verified that all checks are the same. I even copied each version check
from 10tls.dpatch into a simple test case piece of code which does
#include <sasl.h> at the beginning to get the version components like
the postfix code, and all the conditional code was compiled in. I fail
to see the mistake.

Is this due to a broken compiler?

Help would be appreciated.

Thanks,
-- 
Fabian Fagerholm <fabbe@paniq.net>

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: