serassio [Fri, 3 Jun 2005 21:00:55 +0000 (21:00 +0000)]
Use PRIu64 ISO C99 Standard printf() macro for 64 bit unsigned int
instead of llu. This allow the correct formatting of 64 bit printf
on some systems like native Windows.
serassio [Fri, 3 Jun 2005 18:59:06 +0000 (18:59 +0000)]
First step of MinGw configure support.
Changes:
- MinGW environment identification
- Windows PSAPI.DLL support
- Support for Windows native port of OpenSSL
- Winsock support detection
- Identification of maximum FDs number
- Windows native resolver identification
serassio [Sun, 29 May 2005 02:01:06 +0000 (02:01 +0000)]
Bug #1305: Double content-length often harmless
There is quite many web servers out there with broken banner engines
forgetting to delete the original content-length after adding the
banner. Currently these are (rightfully) rejected by Squid.
Instead of rejecting we could select the biggest content-length header
found and remove the other. This should fix up these replies while not
allowing for attacks.
hno [Mon, 9 May 2005 23:11:57 +0000 (23:11 +0000)]
Bug #1229: Random data corruption / lost whitespace characters.
This patch both corrects the whitespace delete between replies and
deactivates it entirely as this breaks certain HTTP/0.9 replies.
The response splitting counter measures is extremely likely to trigger
on this anyway, killing the connection before it may be reused (still
need to be forward-ported from Squid-2.5).
hno [Fri, 6 May 2005 07:57:55 +0000 (07:57 +0000)]
Bug #1166: Configuration confusing when empty acls are encountered
Bug #1255: http_access line with unknown acls
This patch makes Squid very strict about access configuration errors.
Previously Squid ignored most errors, now it rejects the configuraiton
with an description of the error seen.
serassio [Thu, 5 May 2005 21:44:44 +0000 (21:44 +0000)]
Bug 1298: Multiple OIDs in SNMP GETNEXT not processed
Bug 1299: SNMPv2 requests should result in SNMPv2 responses
This patch fixes some minor issues in the SNMP Agent. The GETNEXT/BULK
requests both may contain more than one OID just like normal GET, but
Squid ignored all but the first.
Also the response version was always SNMP Version 1, even if the request
was a SNMP Version 2(c) request.
Bug #1223: Make the use of the %m error page to return auth info
messages
This patch extends the helper protocols for Basic and Digest to provide
some basic information in error responses, and makes use of the error
response already included in the NTLM helper protocol, making these
messages available as %m in error pages. Can be used if desired to
indicate why a login failed. The exact messages returned is helper
dependent.
Bug #1223: Make the use of the %m error page to return auth info
messages
This patch extends the helper protocols for Basic and Digest to provide
some basic information in error responses, and makes use of the error
response already included in the NTLM helper protocol, making these
messages available as %m in error pages. Can be used if desired to
indicate why a login failed. The exact messages returned is helper
dependent.
Bug #1094: Fix for CVE-1999-0710: cachemgr malicouse use
This patch adds access controls to the cachemgr.cgi script, preventing
it from being abused to reach other servers than allowed in a local
configuration file.
IPv6 preparation by Rafael Martinez Torres <rafael.martinez@novagnet.com>
- This patch does change literally the term IN_ADDR by in_addr in all the
files containing this term, except the file "inet_ntoa.c", the only one
not including "config.h" (either direct or undirectly).
- The correposding "defines" are in config.h . They are guarded by "INET6"
define, (#ifdef INET6), so branches others that squid3-ipv6 should not be
affected in their compilations.
- Enhance performance by zero-copy writes, enabled by making the mem
nodes reference counted.
- Implement ASYNC_CLOSE define, default to off.
- Remove unused aioFDWasClosed call
- Kill warning about event notification filedescriptors still
being open on shutdown.
Also optimizes slightly by initializing the squidaio layer early,
avoiding the need to check if initialized in each and every call.
hno [Tue, 29 Mar 2005 04:44:12 +0000 (04:44 +0000)]
Crude implementation of ESIParser registry, making ESIParser.cc
independent of the available parsers, indirectly solving the
libxml2/expat conflict and preparing for easy buildtime selection of the
available parsers.