Re: patches

Here is the status of all my current 2.3.DEVEL2 patches in reverse
cronological order:

_ squid-2.3.DEVEL2.FreeBSD_statfs.patch
_ squid-2.3.DEVEL2.ftp_HEAD.patch
_ squid-2.3.DEVEL2.ipcache_reload.patch
- squid-2.2.STABLE4.daemon-mode-stdio.patch
_ squid-2.2.STABLE4.cachemgr_obj_listings.patch
- squid-2.2.STABLE4.helper_statistics-2.patch
_ squid-2.2.STABLE4.helper_rotate_log.patch
_ squid-2.2.STABLE4.helper_filedescriptors.patch
_ squid-2.2.STABLE4.pidfile_shutdown.patch
_ squid-2.3.DEVEL2.rebuild_release.patch
_ squid-2.3.DEVEL2.cache_mem_startup.patch
_ squid-2.3.DEVEL2.auth_regex.patch
_ squid-2.3.DEVEL2.CONNECT_miss_access.patch
X squid-2.3.DEVEL2.ident_lookup_access_documentation.patch
_ squid-2.3.DEVEL2.no_cache_full_acl_check.patch
X squid-2.3.DEVEL2.ssl_write_log_warning.patch
? squid-2.3.DEVEL2.acl_domain_compare.patch
x squid-2.3.DEVEL2.ftp_data_close.patch
X squid-2.3.DEVEL2.peer_exchanges_login.patch
X squid-2.3.DEVEL2.authenticate_ip_ttl.patch
X squid-2.2.STABLE4.unexpected_304_reply.patch
X squid-2.2.STABLE4.cd_peer_hit_type.patch
X squid-2.2.STABLE4.snmp_port_0.patch
X squid-2.3.DEVEL2.enable_underscores.patch
_ squid-2.3.DEVEL2.cache_swap_log_name-2.patch
X squid-2.2.STABLE4.ftp_top_level_slashes.patch
X squid-2.3.DEVEL2.escape_control-2.patch
_ squid-2.2.STABLE4.aio_cond_segfault-2.patch
X squid-2.3.DEVEL2.aio_false_warnings.patch
_ squid-2.2.STABLE4.log_ip_on_direct.patch
X squid-2.2.STABLE3.ftp_timeout_assert.patch
_ squid-2.3.DEVEL2.stat_ufs_fs.patch
X squid-2.2.STABLE3.ftp_abort_assert.patch
_ squid-2.2.STABLE3.persistent_connection_timeout.patch
_ squid-2.2.STABLE2.delay_pools_initial_level_overflow.patch
X squid-2.2.STABLE4.ftp_password_urls.patch
x squid-2.3.DEVEL2.request_entities_and_updates.patch
_ squid-2.3.DEVEL2.allow_miss_peer_option.patch
X squid-2.2.STABLE2.to_large_objects-2.patch
_ squid-2.3.DEVEL2.Solaris_x86_dlmalloc_configure.patch
X squid-2.2.STABLE3.swapin_validate_metadata.patch
_ squid-2.2.STABLE2.unix_domain_ipc-2.patch
- squid-2.2.STABLE2.icp_timeout_select_rtt_parent-2.patch
- squid-2.3.DEVEL2.icp_timeout_selection.patch
X squid-2.2.STABLE4.myport_acl.patch
X squid-2.2.STABLE2.log_mime_hdrs_readable.patch
X squid-2.3.DEVEL2.proxy_auth_header_syntax.patch
X squid-2.3.DEVEL2.http_port_bind_addr.patch
_ squid-2.2.DEVEL3.netdb_and_never_direct.patch
_ squid-2.2.STABLE3.nonhierarchical_direct.patch
_ squid-2.3.DEVEL2.tcp_dead_detection_and_retry.patch
_ squid-2.2.DEVEL3.icp_dead_detection.patch
X squid-2.2.STABLE3.ftp_PORT_fallback.patch
X squid-2.2.STABLE4.isolate_splay.patch
X squid-2.2.DEVEL3.ftp_broken_downloads.patch
X squid-2.2.DEVEL3.ftp_put_directory.patch
X squid-2.2.DEVEL3.store_log_blank_content_type.patch

Legend:
X Committed to 2.3.
x Something equivalent is committed
_ Not in the current 2.3.DEVEL tree
- I don't think it should be committed at this time.
? Not sure.

This status is now recorded on my Squid page
<http://hem.passagen.se/hno/squid/> and I'll try to update it when Duane
commits things, provided I get a notification or discovers it in rsync.

What I think of the patches not (yet) committed (in cronological order):

* squid-2.2.DEVEL3.icp_dead_detection.patch
* squid-2.3.DEVEL2.tcp_dead_detection_and_retry.patch

These two I think should be committed. It is adjustments of the
dead-peer detection to make it run more smoothly without harsh edges.
Does not change the basic operations of peer selection.

* squid-2.2.STABLE3.nonhierarchical_direct.patch

While I agree in what you said about the current state of peer selection
being a big mess and needing documentation, but this patch changes
non-ICP peering Squid to behave more like people expect it to.

* squid-2.2.DEVEL3.netdb_and_never_direct.patch

This is a pure bug fix I think. never_direct is a rather strong
qualifier, and I don't think a diffuse qualifier like
minimum_direct_hops or netdb exchanges should override this.

* squid-2.3.DEVEL2.icp_timeout_selection.patch
* squid-2.2.STABLE2.icp_timeout_select_rtt_parent-2.patch

These two have been discussed before and are more of a point fix. Should
probably not be committed at this time.

* squid-2.2.STABLE2.unix_domain_ipc-2.patch

Is not really needed. Changes squid to use UNIX domain sockets for it's
helper processes. The benefit is that the communication is more secure,
the drawback is that UNIX domain sockets may experience problems or
wierd behaviour on some platforms (but so does the loopback IP
interface..)

* squid-2.3.DEVEL2.Solaris_x86_dlmalloc_configure.patch

Don't remember much of this patch. It was some guy having problem to
compile on Solaris X86 with dlmalloc.

* squid-2.3.DEVEL2.allow_miss_peer_option.patch

New feature to boost hit ratio in tightly cache clusters somewhat,
especially useful in combination with proxy-only. The intent is to allow
refreshes of objects in siblings.

* squid-2.2.STABLE2.delay_pools_initial_level_overflow.patch

A small equivalent rearrangement of the initialization math to allow a
larger initial pool value.

* squid-2.2.STABLE3.persistent_connection_timeout.patch

Fix for a mismatch between squid.conf and the code.

* squid-2.3.DEVEL2.stat_ufs_fs.patch

Adjustment of cachemgr filesystem statistics to show usable size (not
raw size) and broadened platform support.

* squid-2.2.STABLE4.log_ip_on_direct.patch

Requested by a number of people for accounting reasons.

* squid-2.2.STABLE4.aio_cond_segfault-2.patch

Forces Squid to make proper use of mutex calls on Linux. The default
mode isn't entirely legal and segfaults on Linux SMP machines due to
races in the threads library.

* squid-2.3.DEVEL2.cache_swap_log_name-2.patch

cache_swap_log_name isn't very useful without this feature. Should
probably be the default.

* squid-2.3.DEVEL2.no_cache_full_acl_check.patch

Don't need to say much about this one..

* squid-2.3.DEVEL2.CONNECT_miss_access.patch

Pure bug fix. CONNECT currently bypasses miss_access.

* squid-2.3.DEVEL2.auth_regex.patch

New ACL types for pattern matching user names (ident_regex and
proxy_auth_regex).

* squid-2.3.DEVEL2.cache_mem_startup.patch

cache_mem currently grows uncontrolled while Squid rebulds it cache.

* squid-2.3.DEVEL2.rebuild_release.patch

Unnededly many storeentries are queued for "late release", even such
entries not related to on-disk cache, and Squid leaks store entries on
"cancelled" objects.

* squid-2.2.STABLE4.pidfile_shutdown.patch

Shouldn't unlink the pid file until shutdown is complete.

* squid-2.2.STABLE4.helper_filedescriptors.patch

lsof showed a lot of filedescriptors being open in helper processes.
Also fixes a minor error when shutting down helpers (didn't allow the
helper to finish is't processing)

* squid-2.2.STABLE4.helper_rotate_log.patch

Makes sure cache.log is rotated properly.

* squid-2.2.STABLE4.helper_statistics-2.patch

Improved helper output in cachemgr. Some may consider it a security
hasard for proxy_auth helpers as it shows the plaintext
username:password while a password is being verified.

* squid-2.2.STABLE4.cachemgr_obj_listings.patch

The cachemgr object listings isn't very practical to use as it is due to
slow speed. Also some new selections where the most important is listing
of objects using a large amount of cache_mem.

* squid-2.2.STABLE4.daemon-mode-stdio.patch

I don't think we should ever run without stdio filedescriptors. This
patch changes it to simply not close those when runnin in daemon mode.
Should probably close stdin and reopen it as /dev/null. My startup
scripts makes the needed redirections (</dev/null >/dev/console 2>&1).
Should not be committed in it's present form.

* squid-2.3.DEVEL2.ipcache_reload.patch

Bug fix & new feature. Reload did not always expunge invalid ipcache
entries, and extends PURGE to also purge ipcache.

* squid-2.3.DEVEL2.ftp_HEAD.patch

HEAD ftp://... without fetching the whole object.

* squid-2.3.DEVEL2.FreeBSD_statfs.patch

FreeBSD 3.3 addition to squid-2.3.DEVEL2.stat_ufs_fs.patch

/Henrik
Received on Sat Oct 23 1999 - 14:20:30 MDT