Notable features:
* I'm building _all_ of the error page languages now
* I'm building a handful of the helper modules (the ones that don't
require ldap or samba to be installed)
* underscores in hostnames are enabled - every few weeks we squid people
get emails asking why someone can't get to some_user.geocities.com .
Bad geocities. :)
From the ChangeLog:
Changes to Squid-2.4.STABLE3 (Nov 28, 2001):
- Fixed bug #255: core dump on SSL/CONNECT if access denied by
miss_access
- Fixed bug #246: corrupt on-disk meta information preventing
rebuilds of lost swap.state files
- Fixed bug #243: squid_ldap_auth now supports spaces in passwords
- Fixed a coredump when creating FTP directories
- Fixed a compile time problem with statHistDump prototype
mistmatch, reported by some compilers
- Fixed a potential coredump situation on snmpwalk in certain
configurations
- Fixed bug #229: filedescriptor leakage in the "aufs" cache_dir
store implementation
- Serbian error message translations
This now means that 'stop' as an argument works correctly, and doesn't
end up having squid restarted.
I've added some sleep code to wait for squid to complete its shutdown,
but I have commented it out for the time being as I don't know whether
its good practice to have your machine sleep during shutdown.
I'll research it some more. :)
I ran off and updated the patch files on the squid website for 2.4.stable1,
and then updated this port to use them.
Note that diskd is still broken here - I haven't actually
committed a fix to squid yet .. :-P
Update the squid-2.4 port to actually _be_ squid-2.4 . Phew.
This port builds, installs, packages, pkg_delete's cleanly.
I'm going to run it through some more linting and tidying up before
I'm completely done with it.
Differences from squid22/squid23 :
* install-pinger isn't built. I'll tackle this later, possibly by creating
a squid user/group. I don't like having suid binaries installed,
even more so when 99% of the users of this port won't even enable
ICMP pinging.
* I've enabled the lru and heap replacement policies. LRU is used by
default, the beauty here is that the user can choose one or the other
without needing a recompile.
* I've enabled ufs (sync), diskd (async) and null (no caching, only proxying).
This again lets users choose what they want without needing a recompile.
The default is still a 100mb cache in /usr/local/squid/cache/ running
ufs. I would change it to diskd but if the user hasn't tweaked their
sysV shm/msg parameters sufficiently they'll just be puzzled when squid
gives mysterious sysV errors (and if they load it up enough to have UFS
become an issue, they'd be better off reading the squid FAQ anyhow..)
options `start' and `stop' now (unless I have forgotten any). This allows
us to call the scripts from /etc/rc.shutdown with the correct option.
The (42 or so) ports that already DTRT before are unchanged.
Synopsis:
"If you fill up your squid cache, CPU will go to 100% but the cache will not be
cleaned up."
PR: 18920
Submitted by: Mike Harding <mvh@ix.netcom.com>
Obtained from: http://www.squid-cache.org/Versions/v2/2.3/bugs/
- the distributed patches got regenerated to have the correct paths
- disable optimization because of gcc bugs (as recommended by squid folks)
- list a few new --enable/--disable options
Squid 2.0 is the-version-formerly-known-as-1.2.
v2.0 has NOVM-like functionality internally, so there's no seperate
NOVM version. v1.1.* is no longer officially supported.
