Matus UHLAR - fantomas wrote:
Matus UHLAR - fantomas wrote:
I was searching for the logic of setting visible_hostname and
unique_hostname. I found out that value of unique_hostname is set by calling
getMyHostname() function, which returns value of visible_hostname, if it's
set. However, I would prefer not to do this - to use autodected hostname,
and only change visible_hostname in configuration file.
I would like to ask, could the *hostname logic be changed, so people could
set visible_hostname and leave unique_hostname to rely on the internal
logic? Should I fill bugreport for this?
On 24.04.09 12:15, Amos Jeffries wrote:
Good point. It does appear as if the unique vs visible dependency is
reversed.
Yes, make its a enhancement bug if you like. I'll be looking into it at
some point for 3.2 unless someone gets there first.
Submitted:
http://www.squid-cache.org/bugs/show_bug.cgi?id=2654
Yes, the dependency seeme reversed. Changing that should not be that hard
imho.
FWIW the whole process of this auto-detection needs to be updated. It's
too complex and failing on to many OS these days to keep as-is. If
anyone wants a small job to fix it get in touch with me for the specs.
Does it fail? According to the code it should return fqdn of first listening
socket, which is imho a correct solution of the problem...
At least on Debian and probably elsewhere its returning nothing, even
when hostname of the system is correctly set.
I term that failure, though its probably failure due to the timing of
the lookup (config parse time rather than after socket openings). The
various auto-detect lookups need to be checked for usefulness and
possibly re-ordered.
Amos
--
Please be using
Current Stable Squid 2.7.STABLE6 or 3.0.STABLE14
Current Beta Squid 3.1.0.7
