]> git.ipfire.org Git - thirdparty/squid.git/log
thirdparty/squid.git
23 years agoCreate logs dir during install
robertc [Wed, 26 Sep 2001 05:56:22 +0000 (05:56 +0000)] 
Create logs dir during install

23 years agoMove CACHE_HTTP_PORT and CACHE_ICP_PORT from config.h.in to acconfig.h.
robertc [Wed, 26 Sep 2001 05:45:12 +0000 (05:45 +0000)] 
Move CACHE_HTTP_PORT and CACHE_ICP_PORT from config.h.in to acconfig.h.
Move config.h.in to config.h to reduce unneeded recompilations after a configure.
Regenerate autotools files.

23 years agoRemoved copyright section. Not set.
hno [Tue, 25 Sep 2001 11:53:39 +0000 (11:53 +0000)] 
Removed copyright section. Not set.

23 years agosquid_ldap_auth documentation update. Make the two modes of operation
hno [Tue, 25 Sep 2001 11:49:32 +0000 (11:49 +0000)] 
squid_ldap_auth documentation update. Make the two modes of operation
more obvious.

23 years agoBacked out some bad delay-pool changes (Revision 1.542)
hno [Mon, 24 Sep 2001 04:17:51 +0000 (04:17 +0000)] 
Backed out some bad delay-pool changes (Revision 1.542)

23 years agoAdd some initial COSS information for people.
adrian [Wed, 19 Sep 2001 19:23:30 +0000 (19:23 +0000)] 
Add some initial COSS information for people.

From: diskio branch

23 years agoBugfix: Vary failed to calculate the variance when going thru a parent.
hno [Wed, 19 Sep 2001 01:04:28 +0000 (01:04 +0000)] 
Bugfix: Vary failed to calculate the variance when going thru a parent.
I obviously never learn to use httpState->orig_request when accessing
client data. Not the first time I have been bitten by httpState->request
being a bogus request missing most of the vital information when using
a peer proxy (only contains the request line, and how to contact the peer).

23 years agoFix for a segfault in ftp.c when only creating directories
hno [Tue, 18 Sep 2001 19:55:00 +0000 (19:55 +0000)] 
Fix for a segfault in ftp.c when only creating directories
(PUT with a trailing slash and no filename, content-length == 0)

23 years agojust reformatting long lines in comments to be less than 80 chars
wessels [Tue, 18 Sep 2001 10:17:02 +0000 (10:17 +0000)] 
just reformatting long lines in comments to be less than 80 chars

23 years agomade struct _digest_nonce_h use 'hash_link hash' directlty instead
wessels [Tue, 18 Sep 2001 10:10:50 +0000 (10:10 +0000)] 
made struct _digest_nonce_h use 'hash_link hash' directlty instead
of key,next elements that look just like the hash_link.

23 years agoChange of email to hno@squid-cache.org
hno [Wed, 12 Sep 2001 15:45:53 +0000 (15:45 +0000)] 
Change of email to hno@squid-cache.org

23 years agoGuido: found another function declare mismatch between protos.h and StatHist.c
hno [Mon, 10 Sep 2001 01:57:10 +0000 (01:57 +0000)] 
Guido: found another function declare mismatch between protos.h and StatHist.c
statHistDump

23 years agoRemove MEM_STMEM_BUF from the scene, and statically define the stmem
adrian [Sat, 8 Sep 2001 00:02:44 +0000 (00:02 +0000)] 
Remove MEM_STMEM_BUF from the scene, and statically define the stmem
buffer into the mem_node struct. This should cut down on the number
of allocation requests made per HTTP request.

23 years agooops, need to add .txt to all target
wessels [Fri, 7 Sep 2001 07:32:43 +0000 (07:32 +0000)] 
oops, need to add .txt to all target

23 years agomake .txt version also
wessels [Fri, 7 Sep 2001 07:30:32 +0000 (07:30 +0000)] 
make .txt version also

23 years agoBugzilla #229: storeAufsOpenDone filedescriptor leak
wessels [Fri, 7 Sep 2001 06:51:21 +0000 (06:51 +0000)] 
Bugzilla #229: storeAufsOpenDone filedescriptor leak
patch from Brian (hiryuu of envisiongames dot net)

23 years agoAdded missing lib/ path
hno [Fri, 7 Sep 2001 03:37:29 +0000 (03:37 +0000)] 
Added missing lib/ path

23 years agoSplit those large blocks of filenames into more manageable lists
hno [Fri, 7 Sep 2001 03:17:39 +0000 (03:17 +0000)] 
Split those large blocks of filenames into more manageable lists

23 years ago2.5 bugfix: header_access failed to remove headers if there was no
hno [Fri, 7 Sep 2001 01:51:56 +0000 (01:51 +0000)] 
2.5 bugfix: header_access failed to remove headers if there was no
replacement set by header_replace.

Caused by lack of braces around if/else blocks.

Reported with a patch by Andres Kroonmaa

23 years agoOption '-p ldapport' to specify LDAP server port.
hno [Wed, 5 Sep 2001 13:58:15 +0000 (13:58 +0000)] 
Option '-p ldapport' to specify LDAP server port.
Old meaning of -p (persistent) is moved to -P to match openldap tools options.

23 years agoDon't override configure DEFS
robertc [Tue, 4 Sep 2001 09:09:09 +0000 (09:09 +0000)] 
Don't override configure DEFS

23 years agoFix all DEFAULT_helper variables; change SQUID_VERSION to VERSION in dnsserver.c
robertc [Tue, 4 Sep 2001 07:02:49 +0000 (07:02 +0000)] 
Fix all DEFAULT_helper variables; change SQUID_VERSION to VERSION in dnsserver.c

23 years agoset DEFAULT_PINGER
robertc [Tue, 4 Sep 2001 05:38:33 +0000 (05:38 +0000)] 
set DEFAULT_PINGER

23 years agoDon't try chdir("none") for coredump_dir option.
wessels [Tue, 4 Sep 2001 05:01:45 +0000 (05:01 +0000)] 
Don't try chdir("none") for coredump_dir option.

23 years agoAdd Makefile as a dependency to repl_modules and auth_modules.
robertc [Tue, 4 Sep 2001 04:59:26 +0000 (04:59 +0000)] 
Add Makefile as a dependency to repl_modules and auth_modules.

23 years agoI think store_modules.c must depend on 'Makefile' as well. Its
wessels [Tue, 4 Sep 2001 04:46:09 +0000 (04:46 +0000)] 
I think store_modules.c must depend on 'Makefile' as well.  Its
not getting recreated after I re-run configure with new store modules
added.

23 years agoMajor cleanup to separate error handling from normal flow
hno [Tue, 4 Sep 2001 04:15:07 +0000 (04:15 +0000)] 
Major cleanup to separate error handling from normal flow

Added creation of cfgaux if it does not exists

23 years agoSQUID_PREFIX is no longer needed and Makefile.am doesn't define it.
wessels [Tue, 4 Sep 2001 03:28:57 +0000 (03:28 +0000)] 
SQUID_PREFIX is no longer needed and Makefile.am doesn't define it.

23 years agotest(1) complains about == operator.
wessels [Tue, 4 Sep 2001 02:54:26 +0000 (02:54 +0000)] 
test(1) complains about == operator.

*manually* patched configure because automake stuff seems really
broken and unwieldy right now.

23 years agoMerge from ntlm work. Fixes: reconfiguring when ntlm authentication in progress,...
robertc [Mon, 3 Sep 2001 16:33:02 +0000 (16:33 +0000)] 
Merge from ntlm work. Fixes: reconfiguring when ntlm authentication in progress, no longer resets all ntlm auth progress for user foo when user foo has an error on one connection

23 years agoRequire version 1.5 of automake, and have bootstrap.sh provide basic error feedback.
robertc [Sun, 2 Sep 2001 18:16:54 +0000 (18:16 +0000)] 
Require version 1.5 of automake, and have bootstrap.sh provide basic error feedback.

23 years agoMerged in the squid-amake branch from sourceforge.
robertc [Fri, 31 Aug 2001 17:19:04 +0000 (17:19 +0000)] 
Merged in the squid-amake branch from sourceforge.

Due to the different policy regarding the autotool generated files, the Makefile.in files
for cvs.squid-cache.org have been updated in place, and the extra autotool support files
- such as depcomp - added. These should be filtered from sourceforge.

Also of particular note:
* new make targets dist-all - creates two distribution tarballs, one .tar.gz, one .tar.bz2.
  These do not include all the source files - things like the test suite are deliberately
  left out.
* distcheck - creates a .tar.gz dist tarball and confirms that it builds with the default
  options.
* the module files - lru/heap/ufs/basic etc are now libraries - liblru.a etc.
* The module files are build from the directory above, ie to build just liblru cd to
  src/repl, and type make liblru.a. The benefit to this is that parallelised makes get
  better efficiency, as they don't have to serialise each stub node.

23 years agoReset basic user scheme credentials to unattempted in the event of authentication...
robertc [Fri, 31 Aug 2001 05:00:36 +0000 (05:00 +0000)] 
Reset basic user scheme credentials to unattempted in the event of authentication failure

23 years agoBUGFIX: potential coredump condition on snmpwalk of empty tables.
hno [Thu, 30 Aug 2001 15:51:40 +0000 (15:51 +0000)] 
BUGFIX: potential coredump condition on snmpwalk of empty tables.
If the indexing function did not return a new OID, make sure the
parsing function is forgotten as well.

23 years agolatest rollup from ntlm.
robertc [Wed, 29 Aug 2001 20:57:33 +0000 (20:57 +0000)] 
latest rollup from ntlm.
Includes:
* bugfix for ntlm helpers getting stuck
* bugfix for "Direction" error
* NTLMSSP helper no longer hangs
* some minor authentication code tidyups
* ipcCreate returns PID.
* Stateful helpers show helper PID in stats page.
* Andrew Doran has updated all his email address's. Hi Andrew!

23 years agoRemoved some bad & unneeded SSL junk
hno [Mon, 27 Aug 2001 04:24:56 +0000 (04:24 +0000)] 
Removed some bad & unneeded SSL junk

23 years agoBUGFIX: Plug a large SSL memory leak. SSL connections was never freed.
hno [Mon, 27 Aug 2001 04:22:43 +0000 (04:22 +0000)] 
BUGFIX: Plug a large SSL memory leak. SSL connections was never freed.

Note: This is done in a quite crude manner, and there most likely is
better ways of doing this. Have investigated the use of a close handler
but it won't fit well with lingering closes.

23 years agoDon't log "null" SSL errors, as seen if the client aborts the connection
hno [Mon, 27 Aug 2001 04:17:23 +0000 (04:17 +0000)] 
Don't log "null" SSL errors, as seen if the client aborts the connection
before sending a SSL handshake

23 years agoRemoved unused function declaration
hno [Mon, 27 Aug 2001 04:02:19 +0000 (04:02 +0000)] 
Removed unused function declaration

23 years agoBUGFIX: Certain non-fatal errors could case a client connection to hang
hno [Fri, 24 Aug 2001 21:02:43 +0000 (21:02 +0000)] 
BUGFIX: Certain non-fatal errors could case a client connection to hang
until connection timeout. For example after the following error:
  clientReadRequest: FD 17: no data to process

23 years ago--with-aio option for including AIO libraries. Adds -lrt if found to have
hno [Fri, 24 Aug 2001 20:29:32 +0000 (20:29 +0000)] 
--with-aio option for including AIO libraries. Adds -lrt if found to have
aio_read.

--with-aio-threads renames to --with-aufs-threads to not cause confusion
between aio and aufs I/O.

23 years agoRemoved unused variable "i".
hno [Fri, 24 Aug 2001 19:48:31 +0000 (19:48 +0000)] 
Removed unused variable "i".

23 years agorename cf_parser.c to cf_parser.h to reduce difference between HEAD and automake...
robertc [Thu, 23 Aug 2001 19:20:46 +0000 (19:20 +0000)] 
rename cf_parser.c to cf_parser.h to reduce difference between HEAD and automake branch - allowing for a code-change-free patch set

23 years agoPrevent segmentation fault when using emulate_httpd_log without ident(rfc931) enabled
robertc [Thu, 23 Aug 2001 05:26:05 +0000 (05:26 +0000)] 
Prevent segmentation fault when using emulate_httpd_log without ident(rfc931) enabled

23 years agoRename $(srcdir)/mime.conf to mime.conf.default in Makefile.in, and in CVS
robertc [Wed, 22 Aug 2001 15:25:41 +0000 (15:25 +0000)] 
Rename $(srcdir)/mime.conf to mime.conf.default in Makefile.in, and in CVS

23 years agoBugzilla #215: NULL pointer access for proxy requests in accel-only mode
wessels [Tue, 21 Aug 2001 11:54:13 +0000 (11:54 +0000)] 
Bugzilla #215: NULL pointer access for proxy requests in accel-only mode

This patch fixes the bug by checking for NULL in aclIsProxyAuth().
If access is denied due to receipt of a proxy request with
'httpd_accel_with_proxy off' then AclMatchedName is NULL.
clientAccessCheckDone() calls aclIsProxyAuth() with the NULL pointer,
so the check could go into either function.  It was cleaner to put
it in aclIsProxyAuth().

23 years agoResolve any possible conflicts with POSIX AIO
hno [Mon, 20 Aug 2001 12:55:31 +0000 (12:55 +0000)] 
Resolve any possible conflicts with POSIX AIO

* Use the prefix squidaio_ rather than aio_...
* Some minor cleanups

23 years agoredirector comments made vague reference to 'Release Notes' but this
wessels [Sun, 19 Aug 2001 22:33:17 +0000 (22:33 +0000)] 
redirector comments made vague reference to 'Release Notes' but this
should instruct the reader to look at the FAQ instead.

23 years agoremove reference to old 'cache_swap' option, and make this a warning
wessels [Sun, 19 Aug 2001 22:28:33 +0000 (22:28 +0000)] 
remove reference to old 'cache_swap' option, and make this a warning
rather than a fatal error.

23 years agoBugfix: squid.conf.default needs renaming when installing
robertc [Sun, 19 Aug 2001 15:34:36 +0000 (15:34 +0000)] 
Bugfix: squid.conf.default needs renaming when installing

23 years ago * src/Makefile.in: Change build dir squid.conf to squid.conf.default - it is more...
robertc [Sun, 19 Aug 2001 09:49:38 +0000 (09:49 +0000)] 
* src/Makefile.in: Change build dir squid.conf to squid.conf.default - it is more accurate.
* src/cf_gen.c: Ditto.

23 years agoChange the squid-users mailing list to the new address.
robertc [Sun, 19 Aug 2001 06:46:20 +0000 (06:46 +0000)] 
Change the squid-users mailing list to the new address.

24 years agoWIN32/Cygwin fixes by Guido
hno [Thu, 16 Aug 2001 06:16:15 +0000 (06:16 +0000)] 
WIN32/Cygwin fixes by Guido

- Better GetOSVersion function, more squid "aware", added support for
Windows XP and future Windows versions.
- Added trap in WIN32_Subsystem_Init() for unknown or unsupported (WIN32s)
Windows version.
- Fixed compile warning in comm.c
- Changed exit() WIN32 hook, now less intrusive

And some ident fixes

24 years agoTest commit (by Adrian)
robertc [Thu, 16 Aug 2001 05:47:05 +0000 (05:47 +0000)] 
Test commit (by Adrian)

24 years agoImplement 'sync'
adrian [Mon, 13 Aug 2001 04:23:22 +0000 (04:23 +0000)] 
Implement 'sync'

24 years ago* return the load based on the aio queue depth
adrian [Mon, 13 Aug 2001 04:18:35 +0000 (04:18 +0000)] 
* return the load based on the aio queue depth
* add a statistic line showing the number of pending disk operations

24 years ago* Add support for the async write op
adrian [Sun, 12 Aug 2001 21:29:58 +0000 (21:29 +0000)] 
* Add support for the async write op
* Make sure we call freefunc() on the buffer in the write op case

24 years agoCommit the basic code to implement async io via posix aio.
adrian [Sun, 12 Aug 2001 20:58:24 +0000 (20:58 +0000)] 
Commit the basic code to implement async io via posix aio.
only read is done at the moment - write and sync haven't been
implemented yet.

24 years ago* oops - move the aiocb array to the queue, rather than the queue entry.
adrian [Sun, 12 Aug 2001 20:02:01 +0000 (20:02 +0000)] 
* oops - move the aiocb array to the queue, rather than the queue entry.
* Write some comments outlining what the code is going to ahve to do.

24 years agoWrite a storeCossDirCallback() :)
adrian [Sun, 12 Aug 2001 16:25:00 +0000 (16:25 +0000)] 
Write a storeCossDirCallback() :)

24 years agoAdd a (very evil) async IO framework. It doesn't do much - just get
adrian [Sun, 12 Aug 2001 16:20:41 +0000 (16:20 +0000)] 
Add a (very evil) async IO framework. It doesn't do much - just get
the calls in there. The next (few) commits will be the async code.

Yes, yes, I know it should be "Generic", but I'm trying to get COSS
to "user testing" state, so this is the best 'comprimise'.

For now, COSS will use the libaio routines (or the sync stuff in
disk.c - the calling API is the same so wrapping stuff in an #ifdef
will be quite easy) which are available on Solaris, FreeBSD
and possibly Linux.

24 years agoOops, make sure we actually *set* the selected flag!
adrian [Sun, 12 Aug 2001 01:09:21 +0000 (01:09 +0000)] 
Oops, make sure we actually *set* the selected flag!

24 years agoThe four-liner below alters fatal() so that fatal errors cause squid
adrian [Sun, 12 Aug 2001 00:27:20 +0000 (00:27 +0000)] 
The four-liner below alters fatal() so that fatal errors cause squid
to dump core rather than exit more-or-less-cleanly (of course, the actual
outcome depends on the user's ulimit settings).

Submitted by: Chemolli Francesco <ChemolliF@GruppoCredit.it>

24 years agoSelection fixes - don't try to use a store dir that most definitely
adrian [Sat, 11 Aug 2001 22:34:37 +0000 (22:34 +0000)] 
Selection fixes - don't try to use a store dir that most definitely
can't take the object. COSS, for example.

24 years agoFatal if we can't open a coss store file at init time, rather than
adrian [Fri, 10 Aug 2001 09:20:37 +0000 (09:20 +0000)] 
Fatal if we can't open a coss store file at init time, rather than
first IO time.

24 years agoCOSS fixes from diskio.
adrian [Fri, 10 Aug 2001 03:41:53 +0000 (03:41 +0000)] 
COSS fixes from diskio.
This code is now stable, but <2gb and sync.
Next, >2gb.

24 years agoMismatching parameters to a debug() call
adrian [Thu, 9 Aug 2001 17:00:10 +0000 (17:00 +0000)] 
Mismatching parameters to a debug() call
Reported by: Robin Garner <robin.garner@iname.com>
Submitted by: rbcollins

24 years ago* Commit fix to authenticateAuthUserRequestSetIp()
adrian [Tue, 7 Aug 2001 18:35:26 +0000 (18:35 +0000)] 
* Commit fix to authenticateAuthUserRequestSetIp()
* Tidy up comment (forced by the indent check :)

Noticed by: Robin Garner <Robin.Garner@crsrehab.gov.au>
Submitted by: rbcollins

24 years agoWhy was "radix.h" included twice?
wessels [Fri, 3 Aug 2001 22:55:21 +0000 (22:55 +0000)] 
Why was "radix.h" included twice?

24 years agoGuido Serassio discovered missing pointer in aclNBCheck declaration.
wessels [Fri, 3 Aug 2001 22:54:51 +0000 (22:54 +0000)] 
Guido Serassio discovered missing pointer in aclNBCheck declaration.

24 years agoLatest NTLM snapshot code, courtesy of kinkie and robert.
adrian [Fri, 3 Aug 2001 21:12:57 +0000 (21:12 +0000)] 
Latest NTLM snapshot code, courtesy of kinkie and robert.

<quote>
It's been a while coming... but Kinkie and I now have
probably-production ready NTLM code. It's gone past 1.5 million requests
without failing. The attached patch is a rollup of many minor fixes and
some enhancements:

* Digest: fix crash on shutdown
* ACL: New type max_user_ip to replace authenticate_ip_is_strict.
* ACL: Authenication refactored, to separate authentication and
  authorisation as well as allow 'lazy auth' where authentication is
  triggered by the presence of an authentication using ACL, not
  necessarily "proxy_auth". I.e. using max_user_ip will trigger
  authentication even if no proxy_auth acl's are defined.
* Authentication: API for schemes extended to allow handling duplicate
  authentication on a authenticated _connection_ (yes NTLM again).
* Authentication: Authentication API extended to allow multiple IP's per
  auth_user, and to allow authentication as an orthogonal process to
  authorisation.
* NTLM: many minor races fixed. Should prevent the "Direction before
  ..." errors and the "Duplicate auth" + login popup bug.
* Delay pools: Support authentication (Non-NTLM)in delay pools acl's.
  Beginnings of NTLM support for same.
* Helpers: Stateful helper bugfixes to prevent negative deferred helper
  counts (which prevented ntlm challenge refreshing).
* configure: detect two more headers for NTLM compilation on some
  platforms.

</quote>

24 years agoMinor arguments mismatch in file_write declaration found by Andrey Shorin.
hno [Sun, 29 Jul 2001 04:43:24 +0000 (04:43 +0000)] 
Minor arguments mismatch in file_write declaration found by Andrey Shorin.

24 years agoACL driven reply_body_max_size by Robert Collins.
hno [Sat, 28 Jul 2001 15:21:31 +0000 (15:21 +0000)] 
ACL driven reply_body_max_size by Robert Collins.

Bugzilla #75

24 years agoLimit memory used for select() statistics. It was allocating space for
hno [Tue, 17 Jul 2001 16:35:46 +0000 (16:35 +0000)] 
Limit memory used for select() statistics. It was allocating space for
up to SQUID_MAXFD active filedescriptors per select(), but we are
only interested in a much smaller range.

This statistics could use quite a bit of memory when SQUID_MAXFD is large
(think I measured something like 8MB for 16K filedescriptors)

24 years agoMake sure flags is set properly when calling httpBuildRequestHeader
hno [Tue, 17 Jul 2001 16:33:28 +0000 (16:33 +0000)] 
Make sure flags is set properly when calling httpBuildRequestHeader
(specifically the proxying flag)

24 years agoRobert Collins:
hno [Tue, 17 Jul 2001 15:50:38 +0000 (15:50 +0000)] 
Robert Collins:

Bugzilla #198: authenticateAuthUserRequestLock() SEGV bugfix.
[got lost somewhere...]

24 years agoBugzilla #201: useragentLogClose() may call logfileClose(NULL)
wessels [Tue, 17 Jul 2001 03:35:05 +0000 (03:35 +0000)] 
Bugzilla #201: useragentLogClose() may call logfileClose(NULL)

24 years agoFix a 0-byte read condition while reading request bodies. We need to
hno [Mon, 16 Jul 2001 18:44:05 +0000 (18:44 +0000)] 
Fix a 0-byte read condition while reading request bodies. We need to
stop reading when the buffer is full, and continue again when data
have been used. Thanks to Vladimir B. Savkin for finding this issue.

Also some cleanups in httpRequestFree()
* Permanently removed some old suspicious code Duane had found and commented out.
* Added a new comment at the same place, asking on the validity of the few next
  lines. Looks like we are looking a bit too deap into the store entry
  there..

24 years agoRobert Collins:
hno [Fri, 13 Jul 2001 04:30:25 +0000 (04:30 +0000)] 
Robert Collins:

Correct proxy_auth to generate a new challenge when the user was denied
due to a proxy_auth acl.

24 years agoBugzilla #197: refreshIsCachable() incorrectly checks entry->mem_obj->reply
wessels [Fri, 13 Jul 2001 01:40:32 +0000 (01:40 +0000)] 
Bugzilla #197: refreshIsCachable() incorrectly checks entry->mem_obj->reply

The entry->mem_obj->reply is backwards.  It may cause NULL pointer
access.

24 years agoMinor cosmetic fixes
hno [Thu, 12 Jul 2001 18:59:31 +0000 (18:59 +0000)] 
Minor cosmetic fixes

24 years agoBugzilla #195: ERR_ZERO_SIZE_OBJECT page shows [no URL]
wessels [Thu, 12 Jul 2001 04:57:54 +0000 (04:57 +0000)] 
Bugzilla #195: ERR_ZERO_SIZE_OBJECT page shows [no URL]

This patch links the request structure to the ErrorState structs
so errorpage.c can display a URL in these cases.

24 years agoBugzilla #194
hno [Thu, 12 Jul 2001 04:29:50 +0000 (04:29 +0000)] 
Bugzilla #194

Compilation fails on index() on some non-BSD plaforms

should use strchr() instead

24 years agoBugzilla #184: storeDiskdShmGet() assertion
wessels [Tue, 10 Jul 2001 21:35:45 +0000 (21:35 +0000)] 
Bugzilla #184: storeDiskdShmGet() assertion

This patch changes the mechanism for keeping track of unused shared
memory chunks.   Instead of using a linked list, now we'll use a
bitmap.  It should eliminate the problems I'm seeing on Alphas
where the linklist->ptr value has bit #33 mysteriously set.  In
the new scheme, the shm chunk pointer is always recalculated based
on the index of the bitmap.

24 years agoSwapDir->suggest is unused. It has been moved to SwapDir->fsdata->suggest.
wessels [Tue, 10 Jul 2001 01:12:21 +0000 (01:12 +0000)] 
SwapDir->suggest is unused.  It has been moved to SwapDir->fsdata->suggest.

24 years agoAdded back some rudimentary removal policy statistics (only implemented
hno [Wed, 4 Jul 2001 06:12:05 +0000 (06:12 +0000)] 
Added back some rudimentary removal policy statistics (only implemented
for the "lru" policy as of yet)

  policy.Stat(policy, sentry)

  appends policy statistics to sentry

24 years agoBugzilla #177: LinuxPPC 2000 segfault bug
wessels [Sat, 30 Jun 2001 03:16:42 +0000 (03:16 +0000)] 
Bugzilla #177: LinuxPPC 2000 segfault bug

The previous patch wasn't sufficient to fix this bug.  We have to
call va_start() for each time the args will be used in some "v"
(vprintf, etc) function later on.  This patch has been
tested on Joe's powerpc-unknown-linux-gnu system.

24 years agoBugzilla #187: bugs in lib/base64.c
wessels [Sat, 30 Jun 2001 02:20:01 +0000 (02:20 +0000)] 
Bugzilla #187: bugs in lib/base64.c

Possible buffer overrun and sign/unsigned conversion bugs.

24 years agoHack to allow safe_free to be called on const pointers.
hno [Fri, 29 Jun 2001 20:48:06 +0000 (20:48 +0000)] 
Hack to allow safe_free to be called on const pointers.

24 years agoRemoved the need for the failing typecast..
hno [Fri, 29 Jun 2001 20:37:51 +0000 (20:37 +0000)] 
Removed the need for the failing typecast..
safe_free cannot be called on const pointers.

24 years agoRemoved a unneeded typecast that may confuse some compilers..
hno [Fri, 29 Jun 2001 20:31:34 +0000 (20:31 +0000)] 
Removed a unneeded typecast that may confuse some compilers..

24 years agoBugzilla #114
wessels [Fri, 29 Jun 2001 02:16:30 +0000 (02:16 +0000)] 
Bugzilla #114

The patch for 2.4 is simple, but HEAD has changed a lot and I don't
know if the patch is necessary there, or whats the best way to do it.
As a cop-out, I added some annoying debug message asking users to
report whether or not it works correctly.

24 years agoBugzilla #185: storeDiskdDirParseQ[12]() use wrong number of args in debug/printf
wessels [Thu, 28 Jun 2001 11:18:25 +0000 (11:18 +0000)] 
Bugzilla #185: storeDiskdDirParseQ[12]() use wrong number of args in debug/printf

24 years agoRemoving the MAGIC1/MAGIC2 #defines because they are not used.
wessels [Thu, 28 Jun 2001 10:41:48 +0000 (10:41 +0000)] 
Removing the MAGIC1/MAGIC2 #defines because they are not used.

24 years agoWarn the admin if client_db is disabled and there is a 'maxconn' ACL
wessels [Wed, 27 Jun 2001 23:46:42 +0000 (23:46 +0000)] 
Warn the admin if client_db is disabled and there is a 'maxconn' ACL
configured.

24 years agoDocumenting my concerns about deferring swapping out when cache_dir has
wessels [Wed, 27 Jun 2001 06:14:11 +0000 (06:14 +0000)] 
Documenting my concerns about deferring swapping out when cache_dir has
'max-size' options set.

24 years agoBugzilla #177
wessels [Wed, 27 Jun 2001 03:07:09 +0000 (03:07 +0000)] 
Bugzilla #177

Joe Laffey points out that on LinuxPPC2000 (and probably QNX) calling
something like vsnprintf() or similar leaves the "args" parameter
unusable.  Calling it again with the same "args" results in a coredump.
We have code to get around this in QNX by making a copy
of the args and calling va_start() twice.

This patch tries to address the problem by splitting the _db_print()
function into smaller parts, and passing the args as parameters to
the separate _db_print_file, _db_print_stderr, and _db_print_syslog
functions.

24 years agoThe recent patch for Bugzilla #172 doesn't compile on HEAD, so I've
wessels [Wed, 27 Jun 2001 03:02:05 +0000 (03:02 +0000)] 
The recent patch for Bugzilla #172 doesn't compile on HEAD, so I've
disabled it.  Looks like we can't access the RFC931 string in
peerAllowedToUse() until we make some changes to data structures.

24 years agoBugzilla #172: Ident Based ACLs fail when applied to cache_peer_access
wessels [Tue, 26 Jun 2001 23:17:53 +0000 (23:17 +0000)] 
Bugzilla #172: Ident Based ACLs fail when applied to cache_peer_access

The aclCheck_t structure passed for cache_peer_access didn't copy
the user_ident string from the request_t structure.  Further evidence
that the multiple locations for ident/proxyauth username strings needs
to be fixed and probably rewritten.

24 years agoBugzilla #165: "Store Mem Buffer" leaks badly
wessels [Tue, 26 Jun 2001 22:44:24 +0000 (22:44 +0000)] 
Bugzilla #165: "Store Mem Buffer" leaks badly

This fixes some excessive memory usage problems (for uncachable
responses only?) when 'maximum_object_size_in_memory' is not set
to a small value.  It was probably a coding bug from long ago
that didn't occur much because 'maximum_object_size_in_memory'
used to be hard-coded to a small value (8KB).