December 2014 Archives by thread

• Messages sorted by: [ subject ] [ author ] [ date ]
• More info on this list...

Starting: Mon Dec 1 08:55:05 UTC 2014
Ending: Tue Dec 30 03:17:27 UTC 2014
Messages: 392

• [master] 6fbf247 Restore the default SIGSEGV handler during pan_ic   Nils Goroll
• [master] fb25963 Check pointers before dereferencing - avoid SIGSEGV while panicking   Nils Goroll
• [3.0] b69fff3 Use better criteria for determining if child CLI connection is hosed.   Lasse Karstensen
• [master] cbb2f5d Initialize bo->vfc along with the rest of the busyobj so panics work. (slink's pointer test for bo->vfc is impotent because its bo->vfc[1])   Poul-Henning Kamp
• [master] 0484c08 Permit ABI variables to be empty.   Lasse Karstensen
• [master] 6058ced Revert to binary output by default when using -w   Federico G. Schwindt
• [master] 5f345db bump the varnishtest CLI timeout from 20 to 30 seconds   Nils Goroll
• [master] 7e3370e Increase VRT_MINOR_VERSION due to VRT_selecthttp   Lasse Karstensen
• [master] bb5fab8 White space OCD.   Lasse Karstensen
• [master] 100e377 Remove trailing white space.   Lasse Karstensen
• [master] 165b4ca Typo   Federico G. Schwindt
• [master] 6066a2f Improve Range handling   Federico G. Schwindt
• [master] 635ff46 Minor typo in varnishsizes' -P documentation   Tollef Fog Heen
• [master] 525885e Typo   Tollef Fog Heen
• [master] 54fdb1c Remove varnishreplay from the index   Federico G. Schwindt
• [master] 7f48458 And remove rst file as well   Federico G. Schwindt
• [master] 47e8292 Extend the type of argument and return value of V1L_Write   Martin Blix Grydeland
• [master] 74c3c3b Sync up with Debian downstream.   Lasse Karstensen
• [master] 09c93b4 Improve systemd service units   Federico G. Schwindt
• [master] d8b86e8 No systemd on our wheezy systems.   Lasse Karstensen
• [master] 7b33c92 Also remove autoreconf.   Lasse Karstensen
• [master] 3eff253 Replace the separate init/fini functions for the compiled VCL with a single "event" function which takes the event as parameter.   Poul-Henning Kamp
• [master] 263bb06 Expand inifin mechanism to also handle events   Poul-Henning Kamp
• [master] 938e872 Add STRING + REAL support   Federico G. Schwindt
• [master] 65b1900 Permit to compare 2 times using standard operators   Federico G. Schwindt
• [master] 3dc48ce Continuously update our total object size estimate when cond-fetching a still-being-streamed object.   Poul-Henning Kamp
• [master] f0c2d16 Fix indenting in pan_wrk part of panic message   Martin Blix Grydeland
• [master] e0433a7 Add a missing ',' in pan_req panic string   Martin Blix Grydeland
• [master] bfe5df8 If we cannot find nobody/nogroup, lookup current process uid/gid.   Poul-Henning Kamp
• [master] 60c98eb Add missing T mandated by ISO 8601   Federico G. Schwindt
• [master] b916400 Add std.time()   Federico G. Schwindt
• [master] de8cf10 Clarify comment   Federico G. Schwindt
• [master] d057bf9 Nils patch for retrying partial fetches, with minor changes by me.   Poul-Henning Kamp
• [master] 298ffb6 Relax an assertion for the IMS update candidate object   Martin Blix Grydeland
• [master] dba9daa Handle template object failures during condfetch   Martin Blix Grydeland
• [master] d68863d Varnishncsa escapes non-printable chars.   arianna-aondio
• [experimental-ims] 9a9a9cd typo   Lasse Karstensen
• [experimental-ims] db8b1d3 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [experimental-ims] a9d35b4 Clarify comment, add assert + comment   Poul-Henning Kamp
• [experimental-ims] d647764 Eliminate a single-use macro   Poul-Henning Kamp
• [experimental-ims] 9a63616 Cleanup all the easy bits of the VSL api in varnishd.   Poul-Henning Kamp
• [experimental-ims] a34c900 Eliminate some wrk->sp usage.   Poul-Henning Kamp
• [experimental-ims] 33a2242 Push struct worker out of STV_NewObject()   Poul-Henning Kamp
• [experimental-ims] ce3ca92 Fix a detail I didn't catch during code review:   Poul-Henning Kamp
• [experimental-ims] 91d192c Eliminate wrk->sp   Poul-Henning Kamp
• [experimental-ims] 6633013 Make RFC2616_Ttl() take a busyobj   Poul-Henning Kamp
• [experimental-ims] 3544388 More RFC2616 related argument cleanup   Poul-Henning Kamp
• [experimental-ims] 06b6c78 Detect client crashing during startup   Poul-Henning Kamp
• [experimental-ims] 1bda658 Hide the hash preallocation in cache_hash.c again, by tasking persistent with allocating its own objcores.   Poul-Henning Kamp
• [experimental-ims] 219aa02 Eliminate struct sess from HSH_Insert()   Poul-Henning Kamp
• [experimental-ims] e632739 Eliminate the ban-lurkers need for struct sess   Poul-Henning Kamp
• [experimental-ims] a1cdab1 Remove struct sess entirely from bgthreads   Poul-Henning Kamp
• [experimental-ims] ef74b85 Move the busyobj reference from worker to req   Poul-Henning Kamp
• [experimental-ims] e65be4b Belive it or not, I'm still not used to the guts of varnishd being exposed.   Poul-Henning Kamp
• [experimental-ims] 2608703 Give WRW a hint about which vsl to log to.   Poul-Henning Kamp
• [experimental-ims] 4ba2faf Eliminate WSL() usage.   Poul-Henning Kamp
• [experimental-ims] 42f4e95 Give the expiry thread its own VSL, and get rid of WSL() entirely.   Poul-Henning Kamp
• [experimental-ims] 8140011 Renovate the ban lurkers usage of VSL to cause less synchronous VSL activity.   Poul-Henning Kamp
• [experimental-ims] 41a719c Urg, forgot the outher loop of the ban-lurker:   Poul-Henning Kamp
• [experimental-ims] efb159e Put delivery G(un)zip VSL on req->vsl   Poul-Henning Kamp
• [experimental-ims] 89558df Allocate the req early and VSL the SessionOpen there.   Poul-Henning Kamp
• [experimental-ims] c53ab05 Get rid of the workthread VSL buffer.   Poul-Henning Kamp
• [experimental-ims] 27fd76a Don't use 8k stackbuffer, now that thread-stacks get smaller.   Poul-Henning Kamp
• [experimental-ims] a6de126 Use tweak_bytes() for wthread params.   Poul-Henning Kamp
• [experimental-ims] cdcaade Remove unused worker arguments to VDI/close/recycle   Poul-Henning Kamp
• [experimental-ims] ea9d288 silence flexelint   Poul-Henning Kamp
• [experimental-ims] edf72b4 Documentation fixes from Federico G. Schwindt <fgsch at lodoss.net>   Andreas Plesner Jacobsen
• [experimental-ims] 21beea8 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [experimental-ims] 4e529da Use american english   Lasse Karstensen
• [experimental-ims] a502bae Use american english   Lasse Karstensen
• [experimental-ims] cd0f595 Add more info about changes from 2.1 to 3.0. Thanks to xcir.net for inspiration.   Andreas Plesner Jacobsen
• [experimental-ims] 9f8a77d Fix syntax   Andreas Plesner Jacobsen
• [experimental-ims] bf2da4e forgot to delete, thank you scoof   Lasse Karstensen
• [experimental-ims] 3a00261 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [experimental-ims] db0f3b2 Grammar   Andreas Plesner Jacobsen
• [experimental-ims] e995b1a Add short example on how to get Websockets to work   Lasse Karstensen
• [experimental-ims] 064077f Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [experimental-ims] 36fca1e Add short example on how to get Websockets to work (#2)   Lasse Karstensen
• [experimental-ims] 195b601 Correct value for 2.1   Andreas Plesner Jacobsen
• [experimental-ims] c082835 Drop the body of hit-for-pass objects once we have delivered them to the original requester.   Poul-Henning Kamp
• [experimental-ims] 96e2a95 Avoid taking the saintmode lock if the list empty.   Poul-Henning Kamp
• [experimental-ims] 4d047a9 Accept-filters on FreeBSD has been broken for ages, fix them.   Poul-Henning Kamp
• [experimental-ims] 21e653c Use the hash digest as identification instead of the neutered objhead pointer, in order to not have dependency between trouble entry and objhead lifetime.   Poul-Henning Kamp
• [experimental-ims] 09b77e3 backend cookies   Per Buer
• [experimental-ims] 86e85bb 3.0 syntax   Andreas Plesner Jacobsen
• [experimental-ims] 5c9d66c Really change to 3.0 syntax   Andreas Plesner Jacobsen
• [experimental-ims] 5cd69e0 Eliminate obj arg to FetchBody()   Poul-Henning Kamp
• [experimental-ims] b399a62 Add a set of VFP method functions which can do various sanity asserts for us in a single place.   Poul-Henning Kamp
• [experimental-ims] f9e413c Don't assert if we fail to get storage in VFP_Begin()   Poul-Henning Kamp
• [experimental-ims] 501af0b Add a bo->state that tells us what's going on. Improve error reporting at the same time.   Poul-Henning Kamp
• [experimental-ims] 071d905 Eliminate the returnvalue from FetchBody(), rely on bo->state instead   Poul-Henning Kamp
• [experimental-ims] 8f2d2e7 Make FetchBody() pool-task compatible.   Poul-Henning Kamp
• [experimental-ims] af587ef Grammar in varnishncsa man page   Tollef Fog Heen
• [experimental-ims] 4fe672c Add section on device detection   Lasse Karstensen
• [experimental-ims] fe828b4 Fixing a typo in the documentation.   Tollef Fog Heen
• [experimental-ims] 72cb9f6 typos, change example3 to allow for easier purging   Lasse Karstensen
• [experimental-ims] 4b6e84e Merge branch 'master' of /home/lkarsten/work/varnish-cache-fork   Lasse Karstensen
• [experimental-ims] e725920 Remove old Date: header before adding our new one.   Poul-Henning Kamp
• [experimental-ims] 59c6ae4 Join struct vbo and struct busyobj   Poul-Henning Kamp
• [experimental-ims] d1d7134 Always try to set another thread on the body-fetch job.   Poul-Henning Kamp
• [experimental-ims] c5e141b Correct function name   Andreas Plesner Jacobsen
• [experimental-ims] f740e80 Small doc fixes   Andreas Plesner Jacobsen
• [experimental-ims] b8d4eb8 Hand FetchBody() its own busyobj->refcount, and have it release it when done.   Poul-Henning Kamp
• [experimental-ims] b78280f Decrement the n_waitinglist counter when freeing waitinglists.   Martin Blix Grydeland
• [experimental-ims] 40d62bd Have FetchBody reset bo->vfp once it's done with it.   Poul-Henning Kamp
• [experimental-ims] 7b25530 These testcases depend on us not streaming, so they get a 503 for the failure to fetch a body. Make them explicitly disable streaming.   Poul-Henning Kamp
• [experimental-ims] d96e763 Include device class when hashing   Lasse Karstensen
• [experimental-ims] 9831187 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [experimental-ims] 0326c20 Quick polish over HSH_Lookup():   Poul-Henning Kamp
• [experimental-ims] aea8559 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Poul-Henning Kamp
• [experimental-ims] e556146 More deref-reduction:   Poul-Henning Kamp
• [experimental-ims] c92ad59 Utilize the oc->priv2 variable to store the pointer to the allocating stevedore for the default case (malloc and file)   Martin Blix Grydeland
• [experimental-ims] 88298f1 More sp -> req argument weakening   Poul-Henning Kamp
• [experimental-ims] cf81441 Tag oc's with COMPLETE or FAILURE when we fetch the body.   Poul-Henning Kamp
• [experimental-ims] 0041032 Let the body-fetching thread Unbusy the object.   Poul-Henning Kamp
• [experimental-ims] 2b21621 Also move dropping failed objects to FetchBody()   Poul-Henning Kamp
• [experimental-ims] d74bc62 Add a mutex timing flag to diag_bitmap   Martin Blix Grydeland
• [experimental-ims] 6393f26 Fix a misfeature in the hasher-API: Pass the digest in, so we don't have to touch the (potential) new objhead, unless we intend to insert it.   Poul-Henning Kamp
• [experimental-ims] 4bbbc69 In theory diag_bitmap could change between (non-)assignment to t0 and use of t0. Always initialize t0 to something.   Poul-Henning Kamp
• [experimental-ims] dbad89e Now that saint-mode uses the digest, we can remove the temporary objcore assignment.   Poul-Henning Kamp
• [experimental-ims] ca3099c Lessen the load on oh->mtx a bit.   Poul-Henning Kamp
• [experimental-ims] bbc0bc4 Start preparing hash-lookup for streaming objects   Poul-Henning Kamp
• [experimental-ims] d734511 Update comment to match reality   Poul-Henning Kamp
• [experimental-ims] 20d561d Redo the oc->NOTYET stuff after pondering.   Poul-Henning Kamp
• [experimental-ims] bd0246a Edge closer to the new worldorder   Poul-Henning Kamp
• [experimental-ims] c40a7d0 Unbusy the objectcore when we start fetching the body.   Poul-Henning Kamp
• [experimental-ims] ed645d2 Remove busyobj from objcore under oh-mtx   Poul-Henning Kamp
• [experimental-ims] 4edc10c The busyobj refcount must be protected by the oh->mtx just like the oc->busyobj pointer.   Poul-Henning Kamp
• [experimental-ims] 4d33fab Introduce a couple of local busyobj variables for clarity   Poul-Henning Kamp
• [experimental-ims] ab58597 Add a separate table for the http VSL headers   Poul-Henning Kamp
• [experimental-ims] e4996ef Link the HTTP VSLs to the index in the http structure   Poul-Henning Kamp
• [experimental-ims] 61367de Simplify the CPP magic involved in HTTP headers, now that we can rely on everybody agreeing about the order of these fields.   Poul-Henning Kamp
• [experimental-ims] f177a35 Rework the way we log HTTP headers: Rather than the Rx/Tx/Obj, tag them with Req/Resp/Bereq/Beresp/Obj instead.   Poul-Henning Kamp
• [experimental-ims] 745c828 Add two fields to the VSL tag definition, for documentation use.   Poul-Henning Kamp
• [experimental-ims] 9fc9e7d Add short descriptions   Andreas Plesner Jacobsen
• [experimental-ims] 5a1b770 Adjust varnishncsa and varnishreplay for the new tag names   Tollef Fog Heen
• [experimental-ims] e37585d Make sure we distribute vsl_tags_http.h too   Tollef Fog Heen
• [experimental-ims] c5a1c75 With streaming, we cannot allow a storage chunk to move once we have started to fill it. Add a param to the STV_trim() method to indicate acceptability of move.   Poul-Henning Kamp
• [experimental-ims] 3a4edf1 Centralize the updating of the fetch_obj->len field.   Poul-Henning Kamp
• [experimental-ims] 70468df add a label for refs   Per Buer
• [experimental-ims] 56cfc90 alternative, simpler cookie stripper by James Light   Per Buer
• [experimental-ims] f93c6fa kill warnings. fix a ref   Per Buer
• [experimental-ims] fc8678c Remove unused tag   Andreas Plesner Jacobsen
• [experimental-ims] 9cd5b47 escape \0   Per Buer
• [experimental-ims] 8b8adc6 Fix typo in HTTP header name   Tollef Fog Heen
• [experimental-ims] 3cf4384 Adapt to FreeBSD 10 #include changes   Poul-Henning Kamp
• [experimental-ims] 9f9429c move FreeBSD hacks to the (out of varnish-tree) freebsd configs   Poul-Henning Kamp
• [experimental-ims] e6d7f46 Fix for #1109   Poul-Henning Kamp
• [experimental-ims] 9dc28b2 Fix libedit detection on *BSD OS's   Tollef Fog Heen
• [experimental-ims] 7e51d6f Stopgap fix to get FreeBSD 10-current compiling again.   Poul-Henning Kamp
• [experimental-ims] 480c0a5 Go over the return(restart) code.   Poul-Henning Kamp
• [experimental-ims] 10f1de6 Fix #1126 by properly setting the mask token to the IP number token.   Poul-Henning Kamp
• [experimental-ims] 35475f8 Missing errorchecks incompilation of regsub()   Poul-Henning Kamp
• [experimental-ims] 8bc9a68 Better fix for #1126   Poul-Henning Kamp
• [experimental-ims] 7175bb3 Remove the -w argument, -p should be used instead.   Poul-Henning Kamp
• [experimental-ims] 9a1b232 remove the -w option from man varnishd   Per Buer
• [experimental-ims] 8931bc0 :: need to be on a separate line in RST.   Per Buer
• [experimental-ims] 3a82559 Fix escaping in regsub docs   Tollef Fog Heen
• [experimental-ims] 2612267 Remove unused #include   Poul-Henning Kamp
• [experimental-ims] 0679d60 Be much more cautious about how much workspace we have to build predictive vary string.   Poul-Henning Kamp
• [experimental-ims] f992125 Reset output buffer on VGZ_WrwFlush()   Martin Blix Grydeland
• [experimental-ims] 5625a9a Expose resp.body for inspection and testing in varnishtest   Martin Blix Grydeland
• [experimental-ims] 058fa0f stats are no longer in CLI   Andreas Plesner Jacobsen
• [experimental-ims] 0f7d9a3 Cave in to the microoptimization demands and remove a pointless unlock/lock sequence when using critbit, at the expense of a slightly more ornate API for hashers.   Poul-Henning Kamp
• [experimental-ims] b8e04e6 Don't have mempool->free police the max parameter, leave that to the guard. This makes 'max' parameter mis-sizing much less problematic.   Poul-Henning Kamp
• [experimental-ims] 4601ddc Fix some stats-counter issues scoof pointed out yesterday:   Poul-Henning Kamp
• [experimental-ims] 5961b65 Fix a flexelint nit   Poul-Henning Kamp
• [experimental-ims] 353073c Make VSC->hcb_nolock a worker local counter for consistency, locking it would seriously miss the point.   Poul-Henning Kamp
• [experimental-ims] 017ba98 Add an explicit macro_undef() function so we don't pass a NULL argument to a printflike function.   Poul-Henning Kamp
• [experimental-ims] cab6375 Extra asserts to try to catch a weird panic on tinderbox::osx   Poul-Henning Kamp
• [experimental-ims] 8bf6c6d Also reflect the VCC exit code through if -C is specified.   Poul-Henning Kamp
• [experimental-ims] ad9356a Clean up docs about esi:remove and <!--esi   Andreas Plesner Jacobsen
• [experimental-ims] 562c67d Make it possible for Open/Net/DragonFly-BSD to find readline.h   Poul-Henning Kamp
• [experimental-ims] a8ac09b Add missing + in docs   Tollef Fog Heen
• [experimental-ims] 9c449d0 Make things compile on OpenBSD   Poul-Henning Kamp
• [experimental-ims] 563b21e Missed these two in yesterdays OpenBSD compat commit.   Poul-Henning Kamp
• [experimental-ims] f564428 Be consistent about what environment we test the compiled VCL in.   Poul-Henning Kamp
• [experimental-ims] 6d5c733 Variable rename for easier searching.   Poul-Henning Kamp
• [experimental-ims] 2361947 Another bit of the OpenBSD commit I missed.   Poul-Henning Kamp
• [experimental-ims] 0b00877 Add a -r ("read-only") argument which can protect parameters from subsequent changes.   Poul-Henning Kamp
• [experimental-ims] 2b623ea Add a parameter to disable inline-C code in VCL   Poul-Henning Kamp
• [experimental-ims] fb6f3c5 Add parameter vcc_unsafe_path which allows '/' in include "..." and 'import ... from ...'. Default is on (= no change)   Poul-Henning Kamp
• [experimental-ims] 1d588b3 Silence an annoying message when running varnishd -C   Poul-Henning Kamp
• [experimental-ims] ee1f86d Writing an article about the utility of asserts, I noticed some inconsistent whitespace with asserts.   Poul-Henning Kamp
• [experimental-ims] f222778 Eliminate the prepfetch state   Poul-Henning Kamp
• [experimental-ims] 19a9743 -r option for read only parameters   Per Buer
• [experimental-ims] 170d084 explain the reason behind -r   Per Buer
• [experimental-ims] 816c1c3 Enable PCRE JIT-compiled regular expressions   Tollef Fog Heen
• [experimental-ims] 9190dc3 Add 16k more thread stack to cater for PCRE-JIT.   Poul-Henning Kamp
• [experimental-ims] 1eac7c2 Add code coverage of the RST dumping code   Poul-Henning Kamp
• [experimental-ims] 227a394 Add a dummy pipe to use for poll(2)'ing the death of varnishd, while we wait for it to establish CLI connection.   Poul-Henning Kamp
• [experimental-ims] 8d2ce2c Revert "Add a dummy pipe to use for poll(2)'ing the death of varnishd, "   Poul-Henning Kamp
• [experimental-ims] cb13f32 Convert a lot of testcases which don't care how the response was transferred to examine resp.bodylen instead of resp.http.content-length.   Poul-Henning Kamp
• [experimental-ims] b6c079b In the current scheme of things, this can never be the right thing to do.   Poul-Henning Kamp
• [experimental-ims] 49a42aa These three tests depend on not running with streaming   Poul-Henning Kamp
• [experimental-ims] 8a22dbf Account for all attempted fetches.   Poul-Henning Kamp
• [experimental-ims] d8094d2 Set streaming by default, to flush out problems. (We don't actually stream yet, we just pretend for some of the way)   Poul-Henning Kamp
• [experimental-ims] 31ffe7c Make cnt_prepresp run on incomplete busyobj   Poul-Henning Kamp
• [experimental-ims] a0e3078 Go over prepresp and remove a suboptimal optimization for len=0 objects.   Poul-Henning Kamp
• [experimental-ims] a9e636e Make the installed headers be nobase.   Martin Blix Grydeland
• [experimental-ims] 0351ac9 Move lib/libvmod_std/vmod.py to lib/libvcl/vmodtool.py   Martin Blix Grydeland
• [experimental-ims] d12b924 Install vmodtool.py in pkgdatadir   Martin Blix Grydeland
• [experimental-ims] 4d14c2f Install headers needed for vmod compilation in datadir/varnish/include, and export these locations in the pkg_config file.   Martin Blix Grydeland
• [experimental-ims] 7bf7038 Add a varnishapi-uninstalled.pc file, for use by out-of-tree vmod building against an uninstalled Varnish build tree.   Martin Blix Grydeland
• [experimental-ims] 40ee205 These are not debug commands.   Poul-Henning Kamp
• [experimental-ims] 4020b13 Req.hash_always_miss now implies req.hash_ignore_busy.   Dag Haavi Finstad
• [experimental-ims] 3814b12 Use admin health to control test-state   Poul-Henning Kamp
• [experimental-ims] 0352ea2 Make the random/hash/client director use a stable hash.   Poul-Henning Kamp
• [experimental-ims] 9d0e486 Don't consider a vcc expr to be constant after vcc_expr_tostring.   Martin Blix Grydeland
• [experimental-ims] 5b3e41b Fix vmod_log (VRT_StringList returns end of string, not beginning)   Martin Blix Grydeland
• [experimental-ims] 9efa2b0 Introduce the obj_readonly parameter which disables obj.hits and obj.lastuse in order to not dirty VM pages containing cached objects.   Poul-Henning Kamp
• [experimental-ims] 390fa49 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Poul-Henning Kamp
• [experimental-ims] f289def Make all objects have an objcore, even for the vcl_recv->pass case.   Poul-Henning Kamp
• [experimental-ims] 3f56d4e Fix ttl when backend fetches are salvaged into transient storage.   Poul-Henning Kamp
• [experimental-ims] 304683c Adopt Martins fixed for #1138, which is mostly an artifact of me being interrupted half-way through committing a bunch of stuff.   Poul-Henning Kamp
• [experimental-ims] e0c87d4 Move the HTTP-conditional check up before we call vcl_deliver{}, there is plenty of opportunity to prevent conditional delivery in all the other vcl_*{}'s and there is a value in exposing the actual response.   Poul-Henning Kamp
• [experimental-ims] 44c0c6e Collect all aspect of resp.* creation together.   Poul-Henning Kamp
• [experimental-ims] 7755ce5 Duration needs a unit   Andreas Plesner Jacobsen
• [experimental-ims] cf811a3 Give struct req a back pointer to struct sess   Poul-Henning Kamp
• [experimental-ims] 42cf6a9 Downgrade the VCL_*_method() argument from sess to req   Poul-Henning Kamp
• [experimental-ims] 9243e4a Stop VCL's health-polling of backend already when we discard the VCL.   Poul-Henning Kamp
• [experimental-ims] 2342c8b req.url, not req.http.url   Tollef Fog Heen
• [experimental-ims] 6945ed5 Add missing "}".   Tollef Fog Heen
• [experimental-ims] ecb8b09 Wrong description for regsuball.   Tollef Fog Heen
• [experimental-ims] 344a709 std.fileread() should not blindly return whatever file it returned last without checking if the filename changed.   Poul-Henning Kamp
• [experimental-ims] 5011af2 Add some asserts to narrow #1147 and #1148 down   Poul-Henning Kamp
• [experimental-ims] a9e5d35 Increase the id field in VSM to 128 bytes to make space for 64 backend VCL name + IPv4, IPv6 and portnumber.   Poul-Henning Kamp
• [experimental-ims] e75cd2e Typo   Tollef Fog Heen
• [experimental-ims] 6094581 Add more asserts to nail #1147   Poul-Henning Kamp
• [experimental-ims] 94fe4da Try to create a file in our workdir early on and give a suggestion if it fails.   Poul-Henning Kamp
• [experimental-ims] b7175b3 Priv-sep vcc and cc also.   Poul-Henning Kamp
• [experimental-ims] 45a408b update -r description   Poul-Henning Kamp
• [experimental-ims] caac740 Move disable static libraries, we do not use it   Tollef Fog Heen
• [experimental-ims] dc5b18b Improve "waiter" parameter description as discussed in param-audit.   Poul-Henning Kamp
• [experimental-ims] 65fe9ad Don't modify req after we freed it.   Poul-Henning Kamp
• [experimental-ims] 380a5b2 Remove the gzip_window param, it is useless.   Poul-Henning Kamp
• [experimental-ims] 14b7e6e Move t_req from sp to req, and give sp a t_rx for the handover from acceptor to center.   Poul-Henning Kamp
• [experimental-ims] 3cea894 Reduce RFC2616_Do_Cond() arg from sp to req   Poul-Henning Kamp
• [experimental-ims] 967af89 Pass struct req into VCL methods   Poul-Henning Kamp
• [experimental-ims] f80c6e3 Remember to pass req also when we call user defined functions   Poul-Henning Kamp
• [experimental-ims] 7d90d62 Move VRT_count() from sess to req   Poul-Henning Kamp
• [experimental-ims] 931ef7d Emit __match_proto__() for compiled functions   Poul-Henning Kamp
• [experimental-ims] 30ecc41 Downgrade all the VRT-var functions from sp to req   Poul-Henning Kamp
• [experimental-ims] b4d76fe Move more of the VRT interface from 'sp' to 'req'   Poul-Henning Kamp
• [experimental-ims] 56ce6e3 More VRT sp->req work   Poul-Henning Kamp
• [experimental-ims] 48345b6 Use older libtool invocation to please EL5 users. Welcome to 2005.   Tollef Fog Heen
• [experimental-ims] acab05e More VRT sp->req movements   Poul-Henning Kamp
• [experimental-ims] a15dd15 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Poul-Henning Kamp
• [experimental-ims] a20b6af Add a new VMOD.debug and move the stuff varnishtest uses to get out into the obscure corners over there, making vmod.std a better copy& paste example and removing things like "panic" from the VCC.   Poul-Henning Kamp
• [experimental-ims] e9e4aa4 Fix for an off-by-one issue in do_once_cb().   Dag Haavi Finstad
• [experimental-ims] b16e50d Add libvmod_debug to DIST_SUBDIRS too   Tollef Fog Heen
• [experimental-ims] e113b7f Don't install the debug vmod, it's just there for test purposes   Tollef Fog Heen
• [experimental-ims] 14b48c8 More VRT sess->req changes   Poul-Henning Kamp
• [experimental-ims] b71cb6a All the _r_ object accessor functions should by definition be able to take a "const struct req *" so enforce this.   Poul-Henning Kamp
• [experimental-ims] c74b02a Add -rpath to libvmod_debug_la_LDFLAGS, which forces creation of a .so   Tollef Fog Heen
• [experimental-ims] 4264387 reference VCL for naming storage   Per Buer
• [experimental-ims] 820aa11 document a couple of tags   Per Buer
• [experimental-ims] c0cf227 Another big wash of s/sp/req/ changes   Poul-Henning Kamp
• [experimental-ims] 7ac1238 More s/sp/req/ work   Poul-Henning Kamp
• [experimental-ims] 611815f Tag threads separately with request and session.   Poul-Henning Kamp
• [experimental-ims] 1c14a9d Make the waitinglist hold req instead of sp   Poul-Henning Kamp
• [experimental-ims] 47c7bc1 Rename STEP to SESS_STEP   Poul-Henning Kamp
• [experimental-ims] 99d6751 Make HTC_Init() part of getting a request   Poul-Henning Kamp
• [experimental-ims] 04fa5d7 Use automake's support for parallel tests   Tollef Fog Heen
• [experimental-ims] 4f15e16 Move cnt_first{} up above cnt_wait{}, this is a more logical order and will make subsequent diffs clearer.   Poul-Henning Kamp
• [experimental-ims] cbe58e7 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Poul-Henning Kamp
• [experimental-ims] 7d9ca7d Eliminate the "first" step, merge it with "wait"   Poul-Henning Kamp
• [experimental-ims] a1e3ed9 Apply the "alexandrian solution" to cache_center.c and split the state engine into two separate state engines, one for session and one for requests.   Poul-Henning Kamp
• [experimental-ims] c6b0aaa Eliminate the "DONE" state, it is now a session state.   Poul-Henning Kamp
• [experimental-ims] 42780ae move the STP_ values out of the session state functions   Poul-Henning Kamp
• [experimental-ims] 3325c6c Eliminate the "sp" argument to the request state engine states.   Poul-Henning Kamp
• [experimental-ims] 8e253db More firmly split session and request states.   Poul-Henning Kamp
• [experimental-ims] c4d56d7 Add test case for v-c.org ticket #1157 (connection close on pass and do_stream)   Martin Blix Grydeland
• [experimental-ims] 78ad7c4 Continue the move towards multiple concurrent requests per session   Poul-Henning Kamp
• [experimental-ims] 09bf441 Mostly move sp->wrk to req->wrk   Poul-Henning Kamp
• [experimental-ims] 6dfad3d Totally eliminate sp->wrk   Poul-Henning Kamp
• [experimental-ims] 94d7a68 Make SES_Charge() and SES_ReleaseReq() take struct req* argument instead of struct sess*   Poul-Henning Kamp
• [experimental-ims] 87f32cb Always release req before calling SES_Delete()   Poul-Henning Kamp
• [experimental-ims] 5894995 Always allocate a request before calling CNT_Session()   Poul-Henning Kamp
• [experimental-ims] 90846c4 Remove sp->req, sessions can now (almost) have more than one request at a time, paving the way for a lot of future fun...   Poul-Henning Kamp
• [experimental-ims] 1c82be0 Fix a regression in panic output.   Poul-Henning Kamp
• [experimental-ims] ea1df10 Compress this eyesore using CPP   Poul-Henning Kamp
• [experimental-ims] c477891 White-space cleanup   Poul-Henning Kamp
• [experimental-ims] 590299e Make Flexelint hate me less   Poul-Henning Kamp
• [experimental-ims] 7e25234 Overhaul the thread-pool herding code.   Poul-Henning Kamp
• [experimental-ims] 27d63c2 Eliminate the "init_done" and listen socket fields from struct sess by moving the VSL'ing of socket endpoints earlier in the processing.   Poul-Henning Kamp
• [experimental-ims] 4c4d6f6 Make sure that all sessions, requests and busyobj's have a unique transaction id 'vxid'   Poul-Henning Kamp
• [experimental-ims] ef1b8cb Don't hand out zero VXIDs we need that as a magic value.   Poul-Henning Kamp
• [experimental-ims] f41fbd1 Rename '-x dumprst' to '-x dumprstparam', we may have more RST coming in the future.   Poul-Henning Kamp
• [experimental-ims] 24d426d Add code to dump VSL descriptions in RST format   Poul-Henning Kamp
• [experimental-ims] 7aa0548 Turn some descriptions into intended RST format.   Poul-Henning Kamp
• [experimental-ims] e027388 Remove XXX comment overtaken by implementation.   Poul-Henning Kamp
• [experimental-ims] 72d3574 Log SessionOpen with VSL() so we can be sure it is always the first VSL record for a session.   Poul-Henning Kamp
• [experimental-ims] 0008c89 Sanitize and document the Sess{Open|Close} VSL messages.   Poul-Henning Kamp
• [experimental-ims] 673c50d Fix parsing of -T and -S arguments from shared memory in varnishadm when using the -n option.   Martin Blix Grydeland
• [experimental-ims] ad65eb5 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Martin Blix Grydeland
• [experimental-ims] 74e2d3f Seems I have either hit a GCC bug or a lack of imagination in the C-standardisation. (Even money if you ask me...)   Poul-Henning Kamp
• [experimental-ims] 47dd5d2 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Martin Blix Grydeland
• [experimental-ims] 88c7737 Also update waiter_epoll to new SES_* api.   Poul-Henning Kamp
• [experimental-ims] 0c82d95 Distribute sess_close.h too   Tollef Fog Heen
• [experimental-ims] aa9e09c Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Martin Blix Grydeland
• [experimental-ims] dec9455 Fix for an issue where varnishlog reports an incorrect "Length" entry on gunzipped delivery   Dag Haavi Finstad
• [experimental-ims] e855267 Make sure the user understands that this is about the default configuration   Andreas Plesner Jacobsen
• [experimental-ims] c31f908 Add my official "statement of interest" in HTTP/2.0   Poul-Henning Kamp
• [experimental-ims] 9f1fd24 Hmm, that hid too much, try this instead...   Poul-Henning Kamp
• [experimental-ims] d105f98 Document the version requirements of the varnishreplay input file.   Martin Blix Grydeland
• [experimental-ims] 83ab485 Front is slightly easier to understand   Lasse Karstensen
• [experimental-ims] c288891 Fix the parallell make check   Martin Blix Grydeland
• [experimental-ims] b33efe4 Allow 'storage' in backend identifiers.   Dag Haavi Finstad
• [experimental-ims] ae3e5a3 Push allocation of request into the assigned worker thread when we come back from the waiter, to lessen amount of work waiter does.   Poul-Henning Kamp
• [experimental-ims] 3cc4a72 Introduce an enum to make the states of HTTP reception more readable.   Poul-Henning Kamp
• [experimental-ims] c34f10e Move the temporary accounting structure from wrk to req and eliminate two SES_Charge() calls that have nothing to charge.   Poul-Henning Kamp
• [experimental-ims] 1172fc9 Various minor cleanups for code clarity.   Poul-Henning Kamp
• [experimental-ims] c22afc8 Do not consider busy objects as candidates in EXP_NukeOne.   Martin Blix Grydeland
• [experimental-ims] 116393f Do not try to expire objects that are still busy.   Martin Blix Grydeland
• [experimental-ims] 0c9975c Make the ban lurker skip busy objects   Martin Blix Grydeland
• [experimental-ims] 8d1a7b8 Test commit to check trac hooks   Tollef Fog Heen
• [experimental-ims] 5d976a5 INSTALL: minor update   Tollef Fog Heen
• [experimental-ims] f9c42ca Extend cli_limit to 48k to make room for the full output of 'param.show -l'.   Martin Blix Grydeland
• [experimental-ims] acc8909 Regex set for doing device detection   Lasse Karstensen
• [experimental-ims] 73d989b Patch from Geoff:   Poul-Henning Kamp
• [experimental-ims] 15330c9 Make the sandbox interface more modular and tell the implementation which subprocess we are locking down.   Poul-Henning Kamp
• [experimental-ims] d5c5d5b Sigh...   Poul-Henning Kamp
• [experimental-ims] 610d3cb I have no idea how I ended up incrementting the fail counter by the size of the request, probably sleep copy&paste ?   Poul-Henning Kamp
• [experimental-ims] ffa30f0 Change our chosen "bad_ip" to 192.0.2.255 which is a broadcast address from RFC5737's TEST-NET-1 and put a comment there to explain.   Poul-Henning Kamp
• [experimental-ims] feac371 Don't do automatic backend retry if we have received anything from the backend at all. The retry is only (pseudo-)valid when we have not received anything at all, indicating that the backend closed before it got our request.   Poul-Henning Kamp
• [experimental-ims] 5536bb3 Further untangle req/session state engines   Poul-Henning Kamp
• [experimental-ims] a6419c8 Fix HTC_Reinit() return type   Poul-Henning Kamp
• [experimental-ims] f81aedd More untangling of req vs. session state.   Poul-Henning Kamp
• [experimental-ims] d47db77 Add the version identifier to the CLI banner.   Poul-Henning Kamp
• [experimental-ims] 38a7199 Add updated Solaris Sandbox from Nils Goroll   Poul-Henning Kamp
• [experimental-ims] 57412eb Disable CLANG for now, until some auto* magic happens to remove the -g option from linker-commands.   Poul-Henning Kamp
• [experimental-ims] fe2ad14 Finally split cache_center.c into a HTTP/1 and a Request Finite State Machine.   Poul-Henning Kamp
• [experimental-ims] 1226479 Clean up the HTTP1 fsm and document it with a small dot-graph.   Poul-Henning Kamp
• [experimental-ims] 27f24b5 Fix the dot graph of the request fsm   Poul-Henning Kamp
• [experimental-ims] 279bc2e Move most of CNT::start into HTTP1_fsm where it belongs.   Poul-Henning Kamp
• [experimental-ims] d11b853 Shave another couple of lines out of cnt_start{}   Poul-Henning Kamp
• [experimental-ims] 5ce8786 Expose SES_GetReq()   Poul-Henning Kamp
• [experimental-ims] d9a9ecd Doing rollback in a esi:include request would roll back to the original ESI processed request, rather than to the included requests.   Poul-Henning Kamp
• [experimental-ims] e9a47e8 Spelling   Andreas Plesner Jacobsen
• [experimental-ims] cc710d5 Explicitly sync the header of the VSM so other processes can read(2) it on kernels without coherent VM/buf (OpenBSD, still ?, really ?)   Poul-Henning Kamp
• [experimental-ims] efaba47 Recognize OpenBSD as platform. Federico says it almost works.   Poul-Henning Kamp
• [experimental-ims] 1b3b48e Remove long gone cli.debug cli calls   Poul-Henning Kamp
• [experimental-ims] a47e651 Remove long since gone debug.backend cli.   Poul-Henning Kamp
• [experimental-ims] 5d82c73 Remove comment OBE.   Poul-Henning Kamp
• [experimental-ims] 09c459c Eliminate the cnt_start state by folding it into cnt_recv   Poul-Henning Kamp
• [experimental-ims] e7d5afe Allocate a unique "vxid" to all sessions, requests and backend requests (busyobj).   Poul-Henning Kamp
• [experimental-ims] 4670e05 Add a couple of counters to keep track of how many requests are on the busy-obj waitinglists   Poul-Henning Kamp
• [experimental-ims] dc79130 If people only specify the Transient storage, only run on the Transient storage.   Poul-Henning Kamp
• [experimental-ims] 58596b2 Sigh, this one is the working part of the fix:   Poul-Henning Kamp
• [experimental-ims] 158b428 Not sure what I was thinking: Have SLT_SessOpen log the fd#, not the xid, we already have that.   Poul-Henning Kamp
• [experimental-ims] f5fe417 Move INCOMPL() into the VAS family and apply a bit of polish while there.   Poul-Henning Kamp
• [experimental-ims] 3c53c1a Replace the fd# with the VXID in VSL records.   Poul-Henning Kamp
• [experimental-ims] 5d3dc29 Include compat/srandomdev.h in cache_main.c to fix build on Linux.   Martin Blix Grydeland
• [experimental-ims] a591999 Update changes for 3.0.2 rc 1   Tollef Fog Heen
• [experimental-ims] ad169d8 Document 3.0.2 changes   Tollef Fog Heen
• [experimental-ims] 7be15ba Changelog for 3.0.3   Tollef Fog Heen
• [experimental-ims] 043e7dc Duplicate changelog entry   Tollef Fog Heen
• [experimental-ims] 33f5bae Another duplicate changelog entry, thanks fgs   Tollef Fog Heen
• [experimental-ims] 7c784d5 Add long time missing error handling of gunzip'ing fetched objects for ESI processing.   Poul-Henning Kamp
• [experimental-ims] a3eb213 Remove the FAQ from sphinx. It was never properly maintained.   Per Buer
• [experimental-ims] 77d35ab remove warning (invalid comment   Per Buer
• [experimental-ims] f656558 Try to make osx build happier.   Poul-Henning Kamp
• [experimental-ims] 7bced23 Quiet build by default   Tollef Fog Heen
• [experimental-ims] 652cdeb Showcase !~ in the regex example   Per Buer
• [experimental-ims] 07d9e57 Pass -g to nm(1), it makes backtraces on OSX marginally less useless   Poul-Henning Kamp
• [experimental-ims] 83387b8 Initialize pthreads stuff before we use it.   Poul-Henning Kamp
• [experimental-ims] 86619ad A couple of cleanups to get OSX to be more happy.   Poul-Henning Kamp
• [experimental-ims] 4711b4b Fix a race condition, the child thread could get to the first assert before the parent carried out the assignment of the thread id to the variable.   Poul-Henning Kamp
• [experimental-ims] bbdf093 Use expect_close after we munged the gzip stream, varnishd isn't going to like it...   Poul-Henning Kamp
• [experimental-ims] 4f9c5b9 Trim chunk headers to force the bogus data into varnishd's processing   Poul-Henning Kamp
• [experimental-ims] 7ad9ae5 Fix the CLI ref documentation (including a bug in the built in docs).   Per Buer
• [experimental-ims] e60db5b Pass -Wc, -fstack-protector to the libtool ld stage, to make the stack protector work on omnios   Tollef Fog Heen
• [experimental-ims] 9d052cb Clarify storage backend usage Some more text describing the behaviour on the various storage backend + some reformatting for readability   Per Buer
• [experimental-ims] 1ecffcd remove indent warning   Per Buer
• [experimental-ims] 71a1806 Add a "vsl_mask" parameter, that allows individual SLT tags to be suppressed so the never take up space/bandwidth in the VSL.   Poul-Henning Kamp
• [experimental-ims] cf4da22 Reimplement param log_hashstring using vsl_mask   Poul-Henning Kamp
• [experimental-ims] 9232cd6 Generalize the param/bits code, we can do more with it that way.   Poul-Henning Kamp
• [experimental-ims] 73632b2 Add a debug parameter and start moving diag_bitmap stuff into it.   Poul-Henning Kamp
• [experimental-ims] 4dbf7e7 Move 3 bits more from diag_bitmaps to debug   Poul-Henning Kamp
• [experimental-ims] 609c67d Move another bit from diag_bitmap to debug   Poul-Henning Kamp
• [experimental-ims] 495b9f5 Move 3 more debug bits from diag_bitmaps to debug   Poul-Henning Kamp
• [experimental-ims] a1781ca Minor polish to bitmap params   Poul-Henning Kamp
• [experimental-ims] 13c8fed Slightly better coverage testing.   Poul-Henning Kamp
• [experimental-ims] 0b7cc04 Add a new "feature" parameter and move certain bits from diag_bitmaps there. These bits are legitimate feature-ettes which might make sense in a production environment.   Poul-Henning Kamp
• [experimental-ims] a187a0a Elimiante diag_bitmaps and the mutex-profiling/measuring code.   Poul-Henning Kamp
• [experimental-ims] 132e912 eliminate unused VSC field   Poul-Henning Kamp
• [experimental-ims] 57b4594 Don't overrun the VSL segment looking for the end.   Poul-Henning Kamp
• [master] 2ddab72 %hh length modifier added for non-printabale chars.   arianna-aondio
• [master] 544749c Whitespace   Federico G. Schwindt
• [master] 3e8cabc Correct synopsis   Federico G. Schwindt
• [master] cd66576 Update tags   Federico G. Schwindt

Last message date: Tue Dec 30 03:17:27 UTC 2014
Archived on: Wed Feb 11 18:06:34 UTC 2026

• Messages sorted by: [ subject ] [ author ] [ date ]
• More info on this list...