kostas [Thu, 26 Feb 1998 16:01:10 +0000 (16:01 +0000)]
Reorganized cachePerf.cacheProtoStats to cacheProtoAggregateStats and
cacheMedianSvcTable, which is a table who's row represents the minutes
of history used for medians.
Fixed cacheTimeout bug which returned cacheLogLevel
Fixed netdb, ipcache and fqdn returning wrong "timeticks"
Fixed cacheUptime to return SMI_TIMETICS.
Several memory allocation fixes (used dmalloc-3.3 to find them).
rousskov [Thu, 26 Feb 1998 06:56:51 +0000 (06:56 +0000)]
- Fixed(?) time-tick bug in stat.c. StatCounters where copied in a wrong way
resulting in using the same set of pointers in two objects.
- Added Squid "signature" to all ERR_ pages. The signature is hard-
-coded and is added on-the-fly. The signature may use %-escapes.
Added interface to add more hard-coded responses if needed (see
errorpage.c::error_hard_text).
- Both default and configured directories are searched for ERR_ pages
now. Configured directory is, of course, searched first. This allows
you to customize a subset of ERR_ pages (in a separate directory)
without danger of getting other copies out of sync.
wessels [Wed, 25 Feb 1998 04:17:01 +0000 (04:17 +0000)]
- Added more bytes in/out and service time tracking. There is one pair
of byte counters for the server-side (http,ftp,gopher,wais,ssl,pass).
- Also separately track cache hits, misses, and not-modified replies.
- Try to track ICP query service time, but this may not be implemented
right.
- Removed '60min' average, now we have 5 minutes max
wessels [Tue, 24 Feb 1998 23:39:22 +0000 (23:39 +0000)]
From: "Michael O'Reilly" <michael@metal.iinet.net.au>
so I've added the patch below, which checks the make sure the write
request really does match the data passed back from the async IO
call.
note that I haven't actually rebooted with this patch yet, so
'provisional' is an understatement.
rousskov [Mon, 23 Feb 1998 20:03:01 +0000 (20:03 +0000)]
- Cache Manager got new Web interface (cachemgr.cgi). New .cgi script forwards
basic authentication from browser to squid. Cachemgr.cgi now recognizes
"action protection" types described below.
- Added better recognition of available protection for actions in Cache
Manager. Actions are classified as "public" (no password needed),
"protected" (must specify a valid password), "disabled" (those with a
"disable" password in squid.conf), and "hidden" (actions that require a
password, but do not have corresponding cachemgr_passwd entry in squid.conf).
rousskov [Sun, 22 Feb 1998 01:46:28 +0000 (01:46 +0000)]
Removed USE_ALEX_CODE configuration warning.
Added simple context-based debugging to debug.c. Currently, the context is
defined as a constant string. Context reporting is triggered by debug() calls.
Context debugging routines print minimal amount of information sufficient to
describe current context. The interface will be enhanced in the future.
rousskov [Sat, 21 Feb 1998 11:22:04 +0000 (11:22 +0000)]
Started to describe alex_2 changes. The follwoing entries were added to
ChangeLog. These are some of the low level changes coming from thealex_2 branch.
- Replaced _http_reply with HttpReply. HttpReply is a stand-alone object
that is responsible for parsing, swapping, and comm_writing of HTTP
replies. Moved these functions from various modules into HttpReply
module.
- Added HttpStatusLine, HttpHeader, HttpBody.
- All HTTP headers are now parsed and stored in a "compiled" form in the
HttpHeader object. This allows for a great flexibility in header
processing and builds basis for support of yet unsupported HTTP headers.
- Added Packer, a memory/store redirector with a printf interface. Packer
allows to comm_write() or swap() an object using a single routine.
- Added MemBuf, a auto-growing memory buffer with printf capabilities.
MemBuf replaces most of old local buffers for compiling text messages.
- Added MemPool that maintains a pre-allocated pool of opaque objects.
Used to eliminate memory thrashing when allocating small objects (e.g.
field-names and field-value in http headers).
wessels [Sat, 21 Feb 1998 04:04:50 +0000 (04:04 +0000)]
little rewriting needed for disk write callbacks. The callback
may close the FD, so it must be the last thing we do in
that function. Fixed cbdata leak.