https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6511

Jeremy Chadwick <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[email protected],
                   |                            |[email protected]

--- Comment #2 from Jeremy Chadwick <[email protected]> 2010-11-06 14:22:36 UTC 
---
Regarding the mirror being broken: I've added [email protected] to the CC
list in hopes that this can be investigated.

Regarding the exit code 0 situation -- I've reviewed the sa-update code, and I
see exactly how this situation is occurring.  It's actually by design. 
sa-update gets a list of mirrors (from MIRRORED.BY or via the channel), picks a
random mirror, and attempts to fetch the applicable .tar.gz from the mirror 3
times.  If all 3 HTTP GET attempts fail, the code spits out a warning (the
error message shown above) and moves on to the next mirror in the hash.

The relevant code bits are between lines 617 and 649.  Also be sure to look at
the http_get() function.

So the logic of the code seems to be "as long as one of the mirrors worked/has
the content we need, exit code 0 is valid".  This makes sense, but the man page
could use some added clarification.  My recommendation would be to improve the
sa-update man page from:

       An exit code of 0 means an update was available, and was downloaded and
       installed successfully if --checkonly was not specified.

...to:

       An exit code of 0 means an update was available from at least one
       mirror and was downloaded and installed successfully if --checkonly
       was not specified.  A warning will be output ("http: GET request
       failed") if, after 3 repeated fetch attempts, a mirror lacks the
       update or returns a non-200 HTTP status code.

Regarding Daryl and the daryl.dostech.ca mirror -- it's not personal, it's just
that I can't find any mention of this mirror in the SA documentation or on the
SA Wiki.  These sorts of things should be more public.  It also doesn't reflect
well on the Apache Foundation when you have a mirror which appears to be broken
and maintained by someone who's self-proclaimed "lazy" (which in turn makes the
visitor wonder, in this circumstance, just how valid that claim is), even if he
is a chairman.

I'm sorry if it sounds like I'm overreacting, but given the situation and the
circumstances, my first inclination was to assume someone had circumvented one
of the SA mirrors and induced an HTTP redirect to some "mystery" box that had
been compromised -- or that the system administrator isn't paying attention to
failures (or the failures are unbeknownst to him).  I hope you understand my
POV, as I do understand and accept yours.

-- 
Configure bugmail: 
https://issues.apache.org/SpamAssassin/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

Reply via email to