hno [Tue, 4 Mar 2003 05:14:13 +0000 (05:14 +0000)]
Get rid of that ugly depencendy on COMM_ERR_CLOSING for freeing
the FTP control channel buffer on aborts while sending commands.. the
data was even kept in the ftpState already, making the extra buffer
totally unneeded.
hno [Mon, 3 Mar 2003 06:13:48 +0000 (06:13 +0000)]
Cleanup of errno management to rely on the errno to be sent by
the comm code rather than assuming it is kept intact in errno. Due
to the comm callback queue there may be a significant delay and relying
on errno is quite likely to get it wrong
hno [Sun, 2 Mar 2003 16:57:44 +0000 (16:57 +0000)]
Removed the nowdays obsolete test-ellipses.pl source verification program.
We are today using the compiler directly for the same task (identifying
printf format mismatches in custom functions)
hno [Sun, 2 Mar 2003 00:31:17 +0000 (00:31 +0000)]
date: 2003/02/11 05:22:12; author: wessels; state: Exp; lines: +1 -1
make the ...cacheDnsNumberServers OID return the number of name
servers, rather than 0, for the internal DNS case. For the external
DNS case, this OID returns the number of (running) dnsserver helper
processes.
robertc [Sun, 23 Feb 2003 07:08:03 +0000 (07:08 +0000)]
Summary: More Windows Merges.
Keywords:
Hi,
This is the 5th of some splitted native Windows patches grouped by
functionality.
Native Windows port enhancements:
- Added native Windows and Cygwin support to pinger.cc
- Introduced the use of IPPROTO_TCP and IPPROTO_UDP defines instead of '0'
on comm_open, needed by Winsocket. See this old squid-dev thread about:
http://www.squid-cache.org/mail-archive/squid-dev/200108/0162.html
hno [Sat, 22 Feb 2003 21:59:32 +0000 (21:59 +0000)]
Cleanup of the relation between accelerated request and transparently
intercepted request. The two are now handled separately from each other.
This fixes two issues:
a) Transparently intercepted requests is no longer under the restrictions
of accelerated requests in peering relations etc..
b) No risk of confusion in authentication. Authentication is now allowed
for accelerated requests but not transparently intercepted requests.
hno [Sat, 22 Feb 2003 03:52:51 +0000 (03:52 +0000)]
Don't remove the icons on "make distclean". The icons are included in
the distribution and should not be removed even if they can be regenerated
from icons.shar
If an invalid HTTP identifier is received then access.log shows
"error: invalid HTTP-ident" in the URL column, making problems for
log parsers to read the line correctly.
acl checking on failed-and-needing-async-lookups with the ! operator, leads to multiple lookups being needed. So, we stop processing the current checklist when a async state is detected.
robertc [Sun, 16 Feb 2003 09:23:18 +0000 (09:23 +0000)]
Summary: Merge ACL codebase in.
Patches applied:
* robertc@squid-cache.org--squid/squid--acl--3.0--patch-15
Merge from HEAD.
* robertc@squid-cache.org--squid/squid--acl--3.0--patch-14
More acl refactoring.
* Improve checkAccesslist debug clarity.
* Check for dangling async calls in markFinished() and asyncInProgress().
* Reuse DestinationIPLookup for async ASN lookups.
* Introduce IdentLookup ACLChecklist state singleton.
* Make ACLIdent self register.
* Parameterise ACLIdent by the type of match and parse logic.
* Bugfix for dual-registered ACL types - typeString has to be dynamic.
* Introduce ACLDestinationDomain for DST_DOMAIN and DST_DOMAIN_REGEX.
* Introduce ACLSourceDomain for SRC_DOMAIN and SRC_DOMAIN_REGEX.
* rename acl_cert_data to ACLCertificateData.
hno [Sun, 16 Feb 2003 01:24:17 +0000 (01:24 +0000)]
Incremental merge from rproxy
remove the "hack" to have ICP disabled in accelerator only setups even if
icp_port is configured. This should be done by setting "icp_port 0", not
by the fact that Squid is configured as a reverse proxy.
Note: later merges from rproxy will remove the httpd_accel_* directives,
making the "accelerator vs proxy" boundary much thinner.
hno [Sun, 16 Feb 2003 01:13:35 +0000 (01:13 +0000)]
Incremental merge from rproxy
remove the "hack" to have ICP disabled in accelerator only setups even if
icp_port is configured. This should be done by setting "icp_port 0", not
by the fact that Squid is configured as a reverse proxy.
Note: later merges from rproxy will remove the httpd_accel_* directives,
making the "accelerator vs proxy" boundary much thinner.
hno [Sat, 15 Feb 2003 07:15:51 +0000 (07:15 +0000)]
First batch of rproxy merges
New cache_peer options for reverse proxy setups
originserver
name=XXX
forceddomain=XXX
Cleanup of authentication forwarding, and added gatewaying proxy->reverseproxy
when the same Squid is acting as both proxy and reverseproxy with authentication
wessels [Fri, 14 Feb 2003 04:22:28 +0000 (04:22 +0000)]
Now incrementing statCounter.syscalls.disk.unlinks for external unlinkd
helper process to be more consistent with other places that we increment
the counter.
wessels [Fri, 14 Feb 2003 04:04:51 +0000 (04:04 +0000)]
store_check_cachable_hist.no.release_request was always zero because
storeReleaseRequest() always clears ENTRY_CACHABLE. Such objects
wer always counted as not_entry_cachable instead.
Removed references to store_check_cachable_hist.no.release_request
Removed superfluous EBIT_CLR ENTRY_CACHABLE calls before/after
storeReleaseRequest and left comments in their place.
wessels [Fri, 14 Feb 2003 03:52:42 +0000 (03:52 +0000)]
in 'client_list' cachemgr output, remove the percentage from the FETCHES
line. It is no longer menaingful because fetches is incremented for
non-ICP based forwarding decisions. Also moved FETCHES to the
top of the output with other "TCP-ish" stats and out of the "UDP-ish"
stats section.
hno [Tue, 11 Feb 2003 21:34:30 +0000 (21:34 +0000)]
Cleanup of winbind header dependencies to correct building with
current Samba releases and make migration easier in future (and also
backward to older Samba versions if needed)
1. Centralize the headers to include/samba/nsswitch
2. Upgraded the headers to Samba-2.2.7a, without any changes
3. Added a --with-samba-source=... configure option to manually specify
other headers to use (top level Samba source directory)
comm_write returned the number of bytes written in the last segment if the
write got split in multiple segments, not the total written size.
I also removed some bogous special case for write() == 0 which was
copy-pasted from comm_read. A write() == 0 only occurs if 0 bytes
was requested to be written, and is no different from a normal write
projects / thirdparty / squid.git / log
