February 2011 Archives by thread
Starting: Tue Feb 1 00:11:10 CET 2011
Ending: Mon Feb 28 22:47:24 CET 2011
Messages: 115
- [master] d6210a9 A little bit more coverage of vcc_expr.c
Poul-Henning Kamp
- [master] c157e5f Polishing the object allocation code and persistence a bit
Poul-Henning Kamp
- [master] e8b3d50 Don't wait for a CLI connection if the varnishd process closed the debug pipe.
Poul-Henning Kamp
- [master] 8f848fb Set the alignment of things in the persistent silo on a more systematic footing.
Poul-Henning Kamp
- [master] 6af4255 Overhaul the allocation & open-segment house-keeping code
Poul-Henning Kamp
- [master] 921a351 need more room for temporary gzip space on stack
Tollef Fog Heen
- [master] 85e6bd2 Change output format of parameter dump to rst
Tollef Fog Heen
- [master] 60a22cb Fix typo in parameter description
Tollef Fog Heen
- [master] 9df11bb Update parameter descriptions from rst
Tollef Fog Heen
- [master] 96895eb Get rid of extra whitespace in param description
Tollef Fog Heen
- [master] 6ea1b58 Fix typo in param description
Tollef Fog Heen
- [master] a6be2c1 Update parameter docs from varnishd dump
Tollef Fog Heen
- [master] 06399ae Fix typos
Tollef Fog Heen
- [master] afb0a4e Fix compilation error
Tollef Fog Heen
- [master] d9e97e8 Updated specfile, so that is builds trunk again
Ingvar
- [master] 6ed9fc8 Updated with correct instructions on how to build a rpm package from git.
Ingvar
- [master] bf96632 Ooops, this is embarrasing: Move the star so the example actually makes sense...
Poul-Henning Kamp
- [master] e0d21c9 Try to pay attention to actual C-syntax
Poul-Henning Kamp
- [master] 856edab Give the stevedore the chance to tell which LRU list a given object should be on.
Poul-Henning Kamp
- [master] 1d95c79 Add a debug.persistent command so we can fondle persistent storage for testing and debugging.
Poul-Henning Kamp
- [master] b70e631 Move the next_{bot, top} to the smp_sc struct, we can only have one segment open at a time anyway.
Poul-Henning Kamp
- [master] b4916ea Change a fieldname that keeps confusing me.
Poul-Henning Kamp
- [master] 959a7b3 Release the esi_data storage when we release the object storage it refers to.
Poul-Henning Kamp
- [master] cf64e7d Duh! Don't leak all the transient storage we use for pass requests.
Poul-Henning Kamp
- [master] bc2f3f0 Get object/objcore unref events sent through to the stevedore (via objcore method) and refer resultant empty persistant segments.
Poul-Henning Kamp
- [master] 8edfa76 Various polishing while I wait for the long test run to finish:
Poul-Henning Kamp
- [master] f9a8797 Ohh, man...
Poul-Henning Kamp
- [master] 99c0a17 Hmm, I still don't understand the args to git commit it seems...
Poul-Henning Kamp
- [master] 13add5f Reduce the default thread_add_delay to 2ms
Tollef Fog Heen
- [master] 7a081be Clone this from storage_persistent.c
Poul-Henning Kamp
- [master] 20b88b1 Split .h material out from storage_persistent.c
Poul-Henning Kamp
- [master] 0cca0bf clone from storage_persistent.c
Poul-Henning Kamp
- [master] c4c9db0 Split management process functions into their own source file
Poul-Henning Kamp
- [master] 1b33b0f clone from storage_persistent.c
Poul-Henning Kamp
- [master] 55f1241 Move various subroutines to storage_persistent_subr.c, these are the ones that later may find usage in a separate silo-maintenance utility.
Poul-Henning Kamp
- [master] 63105cc Clone from storage_persistent.c
Poul-Henning Kamp
- [master] 052473d Split the silo-centric from the policy parts of persistent storage code.
Poul-Henning Kamp
- [master] 89556f3 Doc: client.identity came with 2.1.4, not 2.1.3
Kristian Lyngstøl
- [master] 9b8035d When my old math teacher made this kind of mistake and we caught it, he would always say "Just checking if you were paying attention".
Poul-Henning Kamp
- [master] ef2face The LRU is a property of the objcore which is on it, not of whatever storage we have put the object into.
Poul-Henning Kamp
- [master] d0db0a5 Get rid of the LRU list sentenniel.
Poul-Henning Kamp
- [master] f9848a6 Allocate the LRU list in the child process instead of the master process.
Poul-Henning Kamp
- [master] 195fa50 Split the expire lock into expire and per-lru list locks.
Poul-Henning Kamp
- [master] b6633c0 BINHEAP_NOIDX should come from the binheap implementation.
Poul-Henning Kamp
- [master] abf653c Argument polishing
Poul-Henning Kamp
- [master] 57fed77 Add a LRU_Free() function
Poul-Henning Kamp
- [master] 2b2f723 Now we do.
Poul-Henning Kamp
- [master] 1a30dde Give the persistent watcher-thread something to do
Poul-Henning Kamp
- [master] 7819754 rhel4 build fix, plus changed som notes in the comments
Ingvar
- [master] 7a77305 typofix
Ingvar
- [master] 3a8158a Ohh man, I should be too old to make this kind of mistake.
Poul-Henning Kamp
- [master] 3718b46 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] 36a303d Use uintptr_t to avoid a downgrade warning on 32 bit systems, which a smarter compiler than GCC could have seen through.
Poul-Henning Kamp
- [master] 4a2d55b Poll on HUP rather than OUT to detect worker process failure, this works better with Solaris
Poul-Henning Kamp
- [master] 81ce41b On 32 bit, set the worker thread stacksize to the lower of 64k and the minimum demanded by the system.
Poul-Henning Kamp
- [master] 4981c41 Push the output gzip buffer before we process an included object, to maintain proper ordering.
Poul-Henning Kamp
- [master] 499c3de Use malloc() for the vgz structures. Allocating it from the workspace runs into trouble when we reset the workspace as part of ESI:include processing.
Poul-Henning Kamp
- [master] 64bb593 Regression test for #861
Poul-Henning Kamp
- [master] 09e0e8e Continuation of #861 fix: The vef structure cannot be on the workspace either.
Poul-Henning Kamp
- [master] 9764cc4 Add libedit support to varnishadm
Tollef Fog Heen
- [master] d4441ab This is getting embarrasing: don't double free...
Poul-Henning Kamp
- [master] e974d51 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] 9aa8317 Zero the wrk->vgz_rx pointer once we have freed it.
Poul-Henning Kamp
- [master] ea15f75 If an esi-processed job starts with <esi:include without any preceeding verbatim bytes, we would not emit the gzip header until after the included file.
Poul-Henning Kamp
- [master] ca6934f Also print out the text description of the gunzip error.
Poul-Henning Kamp
- [master] dd54061 Test corner-cases with an object containing only an <esi:include> with/without leading and training verbatim data.
Poul-Henning Kamp
- [master] 6cc6035 (missed bit of this commit:) If an esi-processed job starts with <esi:include without any preceeding verbatim bytes, we would not emit the gzip header until after the included file.
Poul-Henning Kamp
- [master] 6fa90ee Take another stab at fixing #861 by clearing the correct bit position of the "last" bit in the included ESI object.
Poul-Henning Kamp
- [master] 1a8f105 Log useful gunzip state before we bail out
Poul-Henning Kamp
- [master] 2fface7 A short rant about why I will not add SSL support to Varnish.
Poul-Henning Kamp
- [master] 4c00fe8 Rename "pass" in vcl_fetch to "hit_for_pass" and respect a zerp TTL from VCL.
Poul-Henning Kamp
- [master] 84b7dda I had forgotten that arg spec strings are NUL terminated centrally so I added one NUL too many to enum spec strings, preventing any subsequent args from being handled correctly.
Poul-Henning Kamp
- [master] bb2765d Don't NUL terminate the enum spec in the enum spec writing function, the function-emitting function will do that.
Poul-Henning Kamp
- [master] b07804b Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] 10a2ab5 Rename max_esi_includes param to max_esi_depth and enforce it again.
Poul-Henning Kamp
- [master] 7c14e6a Enable range support by default
Poul-Henning Kamp
- [master] a1f4fe2 Enable logging of hash string components by default.
Poul-Henning Kamp
- [master] 38f99fa Introduce a new category of boiler-plate notice for parameters: things you should not muck about with unless I tell you to.
Poul-Henning Kamp
- [master] 81a0fec At VUG3 this parameter was discussed and the consensus was that 0.4 seconds was a bit too aggresive for a default value, and after some talk forth and back 0.7 seconds was deemed more like it.
Poul-Henning Kamp
- [master] 3602936 Remove the backend connection memory structure cache, it predates our import of JEmalloc, has been turned of by default and does not improve performance when turned on.
Poul-Henning Kamp
- [master] 549c706 Add back a couple of AN()'s that I was a bit too fast in commenting out.
Poul-Henning Kamp
- [master] 089a335 Assert that dlclose()'ing the vmod succeeds.
Poul-Henning Kamp
- [master] 4600acf Add a bunch of (?i) to get the message that the Host: header is case insensitive
Per Andreas Buer
- [master] b6ec427 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Per Andreas Buer
- [master] 90ff618 Increase the default http RX timeout for server/client to 5 seconds.
Poul-Henning Kamp
- [master] faf767b Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] 29f5d0f Increase the maximum test duration to a full minute since people seem to run into the 30 second timeout
Poul-Henning Kamp
- [master] 830f983 Fix up spelling
Tollef Fog Heen
- [master] ef058a0 Fix up links and update for git.
Tollef Fog Heen
- [master] cf78444 Rename svn_version.c to vcs_version.c, simplify generation
Tollef Fog Heen
- [master] 6edf9c3 Overhaul HTTP request/response size limits:
Poul-Henning Kamp
- [master] 841596c Split the worker thread workspace size out into its own parameter "thread_pool_workspace"
Poul-Henning Kamp
- [master] fe5b71e Move all libs but libvarnishapi to a private directory, drop soname number
Tollef Fog Heen
- [master] f35a9b2 Drop version from libvmod_std
Tollef Fog Heen
- [master] 7b9df55 Merge branch 'privlibdir'
Tollef Fog Heen
- [master] e5e43bd Make sure to initialize the the {do|is}_gzip variables to zero before we start setting them otherwise.
Poul-Henning Kamp
- [master] 1c7c400 Also reduce worker thread workspace size for 32bit systems now that it has its own parameter.
Poul-Henning Kamp
- [master] 0956e00 white space cleanup
Poul-Henning Kamp
- [master] 543e3df More parameter tweaks based on VUG3 discussions:
Poul-Henning Kamp
- [master] 44f47f0 Avoid double-free of DNS directors
Tollef Fog Heen
- [master] 49308a9 Polish handling of timeval and timespec a bit.
Poul-Henning Kamp
- [master] ff1165c Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] 466d963 Also set grace = 0 when an object is banned.
Poul-Henning Kamp
- [2.1] varnish-2.1.5-1-g2650bec
Tollef Fog Heen
- [2.1] varnish-2.1.5-2-g465fe1f
Bjørn Ruberg
- [master] 4f9d45b remove undesired semicolons
Poul-Henning Kamp
- [master] f8b5e53 Implement the "purge" action, now that we have renamed "the facility previously known as purge" to ban.
Poul-Henning Kamp
- [master] 15a2f79 Add testcase for purge in vcl hit & miss
Poul-Henning Kamp
- [master] 430d24d Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] b590d0e Change description to note that this test is outdated.
Poul-Henning Kamp
- [master] e2287d2 Look for kqueue on NetBSD
Tollef Fog Heen
- [master] 682f287 Uncouple the gz* files from the build, we will never need them.
Poul-Henning Kamp
- [master] 89289cf Remove the gz* files.
Poul-Henning Kamp
- [master] b4f1774 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache
Poul-Henning Kamp
- [master] c85b3ea If we have a CLANG compiler, use it in the absense of any preferences.
Poul-Henning Kamp
Last message date:
Mon Feb 28 22:47:24 CET 2011
Archived on: Mon Feb 28 22:47:26 CET 2011
This archive was generated by
Pipermail 0.09 (Mailman edition).