February 2012 Archives by author

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

Starting: Mon Feb 6 13:29:58 UTC 2012
Ending: Wed Feb 29 15:17:12 UTC 2012
Messages: 194

• [master] 09b77e3 backend cookies   Per Buer
• [master] ab47f76 Move VCL-controlled backend timeouts from struct worker to busyobj   Martin Blix Grydeland
• [master] a434ce4 Honor remove-flag also when processing comments in ESI parsing.   Martin Blix Grydeland
• [master] 60490c9 Increase backend_workspace of r01038.vtc from 7k to 8k.   Martin Blix Grydeland
• [master] 877f011 Wait a bit before trying to start/use varnishd again   Tollef Fog Heen
• [master] 646d185 Fixup printf format on 32bit platforms   Tollef Fog Heen
• [master] 8a75838 Fix up some sess_workspace leftovers   Tollef Fog Heen
• [master] d119819 32 bit default parameter adjustments   Tollef Fog Heen
• [master] b42a6d1 More 32 bit adjustments, now rename gzip_stack_buffer to gzip_buffer   Tollef Fog Heen
• [master] 71b981f Use %ju for uintmax_t printf   Tollef Fog Heen
• [master] 7d164e0 Most of these variables are not available in vcl_deliver   Andreas Plesner Jacobsen
• [master] edf72b4 Documentation fixes from Federico G. Schwindt <fgsch at lodoss.net>   Andreas Plesner Jacobsen
• [master] cd0f595 Add more info about changes from 2.1 to 3.0. Thanks to xcir.net for inspiration.   Andreas Plesner Jacobsen
• [master] 9f8a77d Fix syntax   Andreas Plesner Jacobsen
• [master] db0f3b2 Grammar   Andreas Plesner Jacobsen
• [master] 195b601 Correct value for 2.1   Andreas Plesner Jacobsen
• [master] 512dddc Hindsight is always so glaring sharp...   Poul-Henning Kamp
• [master] a2a2717 Move more gzip buffers off th wrk->stack and into malloc.   Poul-Henning Kamp
• [master] da2cdc5 Move the second G-ESI-G buffer away from the stack too.   Poul-Henning Kamp
• [master] 0eae9d0 Now that we have a input buffer which persists over calls to vfp->bytes, loose the dedicated "pending" buffer and just use the front of the input buffer. This simplifies the code quite a bit.   Poul-Henning Kamp
• [master] 42f03cc Move the last g(un)zip buffers from the wrk->stack to malloc(3).   Poul-Henning Kamp
• [master] 2a5cf8f Remove comment OBE.   Poul-Henning Kamp
• [master] 2362e93 Add a call-back function to memory pools, which can be used to update the desired size of objects from the guard thread.   Poul-Henning Kamp
• [master] 1366473 Allocate busyobj with a mempool, instead of a private one-item pool.   Poul-Henning Kamp
• [master] 4d4e9f7 Use the req workspace, not the worker workspace, as we cannot trust that to be the same throughout the request.   Poul-Henning Kamp
• [master] 5454314 Remove now unused ws variable.   Poul-Henning Kamp
• [master] e88413b Allocate HTTP.status from http's designated workspace.   Poul-Henning Kamp
• [master] 1cdbbff NB: Major Reorg Commit   Poul-Henning Kamp
• [master] cd2ce96 Allocate the vbo and http's out of the backend workspace to avoid wasting partial pages with likely parameter values.   Poul-Henning Kamp
• [master] 8c2fe8d Clarify that certain parameters are likely to be rounded to page-size by the operating system.   Poul-Henning Kamp
• [master] fc387c3 Eliminate the mempool poll function again, we don't need it anyway.   Poul-Henning Kamp
• [master] c2a2576 Make sure to free workers cached busyobj when threads are reaped.   Poul-Henning Kamp
• [master] 21eb82b Rename variable to not confuse people on monday mornings.   Poul-Henning Kamp
• [master] 866b419 Fix a printf format issue with time_t   Poul-Henning Kamp
• [master] fb3cc4b Make fmt_bytes take an uintmax_t   Poul-Henning Kamp
• [master] b600f1a Don't use HUGE_VAL to signify no upper limit, we run into infinity in conversions to uintmax_t. Use zero to disable upper limit check. Check separately for byte sizes larger than the architecture can do.   Poul-Henning Kamp
• [master] 2e8ff87 Print out the right number in our diagnostics   Poul-Henning Kamp
• [master] 23cffba Check that workspace_backend is big enough, rather than run on the mercy of mallocs rounding algorithm.   Poul-Henning Kamp
• [master] d3f869a Remove sendfile(2) support, it doesn't seem to actually make any difference in practice and complicates the code and increases the size of storage data structures.   Poul-Henning Kamp
• [master] 7ec9794 Move res_mode from worker to req   Poul-Henning Kamp
• [master] 73b465a Integrate stream_ctx in struct req   Poul-Henning Kamp
• [master] ffea64d Shave 256 bytes of the worker thread footprint, by using the thread workspace for WRW also.   Poul-Henning Kamp
• [master] 1fd2741 Start abstracting the VSL handling a bit.   Poul-Henning Kamp
• [master] 892c92a Make refcount signed.   Poul-Henning Kamp
• [master] a5d9851 Introduce a function to initialize a VSL buffer   Poul-Henning Kamp
• [master] c9c0572 Use VSL_Setup() once more   Poul-Henning Kamp
• [master] ad277d2 Add a facility to give each session a unique number with a minimum of locking: Each pool fetches blocks of numbers to assign, then hands them out one by one using the existing lock.   Poul-Henning Kamp
• [master] c305b19 Forgot to add the lock type.   Poul-Henning Kamp
• [master] b762016 Give struct req its own buffer for batching VSL records   Poul-Henning Kamp
• [master] d98dee2 Give busyobj its own VSL staging buffer too.   Poul-Henning Kamp
• [master] f093a08 Tag vbc's with the vsl buffer they should log into. Update the buffers wid with the vbc's vsl_id. Flush the buffer when we recycle/close the vbc.   Poul-Henning Kamp
• [master] fe68764 Attach the vslbuffer to the vbc while in use and lend it the vbc's vsl_id. When the vbc is returned, restore the original vsl_id.   Poul-Henning Kamp
• [master] 75739f8 Split http_Setup() into http_Setup() and http_Teardown()   Poul-Henning Kamp
• [master] 628fdea Tag struct http with which vsl buffer to use.   Poul-Henning Kamp
• [master] 88fd1b2 Adjust testcases for changes struct http size   Poul-Henning Kamp
• [master] f7dbf53 Clean up a lot of now needless argument passing for struct http's VSL logging. More to come.   Poul-Henning Kamp
• [master] e9e3e2e Eliminate a lot of arguments to http_ functions which used to convey the VSL coords which are now a property of struct http.   Poul-Henning Kamp
• [master] f33a076 Give struct htc the same "tell it what vsl_buffer to use" treatment.   Poul-Henning Kamp
• [master] 2d49712 Change FetchError() to take busyobj as arg.   Poul-Henning Kamp
• [master] fe8ec80 Argument strength-reduction: from worker to dstats   Poul-Henning Kamp
• [master] ac85937 More argument strength reduction to reduce the spread of struct wrk   Poul-Henning Kamp
• [master] fe9d181 Align the iovec pointer   Poul-Henning Kamp
• [master] d2b5440 Clamp # of iovec to IOV_MAX   Poul-Henning Kamp
• [master] 40273a3 Clean up this file, and GC a unused counter.   Poul-Henning Kamp
• [master] bf5b757 Force this one into the same consistent layout.   Poul-Henning Kamp
• [master] f012e53 Make sure we have debug.sizeof in at least one test-case.   Poul-Henning Kamp
• [master] af2bfa9 Push struct worker further upwards   Poul-Henning Kamp
• [master] bd9df97 The VGZ api does not need a worker, it just needs a vsl_log   Poul-Henning Kamp
• [master] 071f86b Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Poul-Henning Kamp
• [master] b5f1878 Remove the current "ping-pong" streaming code, it is in the way of the "real streaming" code we are working on.   Poul-Henning Kamp
• [master] 7703a5f Lend busyobj the workers dstat for the duration of Fetch_Body()   Poul-Henning Kamp
• [master] d387098 Make FetchStorage() take a busyobj instead of a worker as arg.   Poul-Henning Kamp
• [master] 5d4e2ce Move VGZ entirely to busyobj instead of worker   Poul-Henning Kamp
• [master] 4076126 Push worker->busyobj change up through ESI   Poul-Henning Kamp
• [master] 9b65ebc Push worker->busyobj shift up through VFP's.   Poul-Henning Kamp
• [master] a9d35b4 Clarify comment, add assert + comment   Poul-Henning Kamp
• [master] d647764 Eliminate a single-use macro   Poul-Henning Kamp
• [master] 9a63616 Cleanup all the easy bits of the VSL api in varnishd.   Poul-Henning Kamp
• [master] a34c900 Eliminate some wrk->sp usage.   Poul-Henning Kamp
• [master] 33a2242 Push struct worker out of STV_NewObject()   Poul-Henning Kamp
• [master] ce3ca92 Fix a detail I didn't catch during code review:   Poul-Henning Kamp
• [master] 91d192c Eliminate wrk->sp   Poul-Henning Kamp
• [master] 6633013 Make RFC2616_Ttl() take a busyobj   Poul-Henning Kamp
• [master] 3544388 More RFC2616 related argument cleanup   Poul-Henning Kamp
• [master] 06b6c78 Detect client crashing during startup   Poul-Henning Kamp
• [master] 1bda658 Hide the hash preallocation in cache_hash.c again, by tasking persistent with allocating its own objcores.   Poul-Henning Kamp
• [master] 219aa02 Eliminate struct sess from HSH_Insert()   Poul-Henning Kamp
• [master] e632739 Eliminate the ban-lurkers need for struct sess   Poul-Henning Kamp
• [master] a1cdab1 Remove struct sess entirely from bgthreads   Poul-Henning Kamp
• [master] ef74b85 Move the busyobj reference from worker to req   Poul-Henning Kamp
• [master] e65be4b Belive it or not, I'm still not used to the guts of varnishd being exposed.   Poul-Henning Kamp
• [master] 2608703 Give WRW a hint about which vsl to log to.   Poul-Henning Kamp
• [master] 4ba2faf Eliminate WSL() usage.   Poul-Henning Kamp
• [master] 42f4e95 Give the expiry thread its own VSL, and get rid of WSL() entirely.   Poul-Henning Kamp
• [master] 8140011 Renovate the ban lurkers usage of VSL to cause less synchronous VSL activity.   Poul-Henning Kamp
• [master] 41a719c Urg, forgot the outher loop of the ban-lurker:   Poul-Henning Kamp
• [master] efb159e Put delivery G(un)zip VSL on req->vsl   Poul-Henning Kamp
• [master] 89558df Allocate the req early and VSL the SessionOpen there.   Poul-Henning Kamp
• [master] c53ab05 Get rid of the workthread VSL buffer.   Poul-Henning Kamp
• [master] 27fd76a Don't use 8k stackbuffer, now that thread-stacks get smaller.   Poul-Henning Kamp
• [master] a6de126 Use tweak_bytes() for wthread params.   Poul-Henning Kamp
• [master] cdcaade Remove unused worker arguments to VDI/close/recycle   Poul-Henning Kamp
• [master] ea9d288 silence flexelint   Poul-Henning Kamp
• [master] c082835 Drop the body of hit-for-pass objects once we have delivered them to the original requester.   Poul-Henning Kamp
• [master] 96e2a95 Avoid taking the saintmode lock if the list empty.   Poul-Henning Kamp
• [master] 4d047a9 Accept-filters on FreeBSD has been broken for ages, fix them.   Poul-Henning Kamp
• [master] 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
• [master] f3a85e9 typo   Lasse Karstensen
• [master] 9a9a9cd typo   Lasse Karstensen
• [master] db8b1d3 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [master] 21beea8 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [master] 4e529da Use american english   Lasse Karstensen
• [master] a502bae Use american english   Lasse Karstensen
• [master] bf2da4e forgot to delete, thank you scoof   Lasse Karstensen
• [master] 3a00261 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [master] e995b1a Add short example on how to get Websockets to work   Lasse Karstensen
• [master] 064077f Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Lasse Karstensen
• [master] 36fca1e Add short example on how to get Websockets to work (#2)   Lasse Karstensen
• [master] e090127 Add Cookie and Set-Cookie to the HTTP headers list   Rogier Mulhuijzen
• [experimental-ims] f3a85e9 typo   Geoff Simmons
• [experimental-ims] 512dddc Hindsight is always so glaring sharp...   Geoff Simmons
• [experimental-ims] 99f0d07 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache into experimental-ims   Geoff Simmons
• [experimental-ims] a2a2717 Move more gzip buffers off th wrk->stack and into malloc.   Geoff Simmons
• [experimental-ims] da2cdc5 Move the second G-ESI-G buffer away from the stack too.   Geoff Simmons
• [experimental-ims] 0eae9d0 Now that we have a input buffer which persists over calls to vfp->bytes, loose the dedicated "pending" buffer and just use the front of the input buffer. This simplifies the code quite a bit.   Geoff Simmons
• [experimental-ims] 42f03cc Move the last g(un)zip buffers from the wrk->stack to malloc(3).   Geoff Simmons
• [experimental-ims] 2a5cf8f Remove comment OBE.   Geoff Simmons
• [experimental-ims] 2362e93 Add a call-back function to memory pools, which can be used to update the desired size of objects from the guard thread.   Geoff Simmons
• [experimental-ims] 1366473 Allocate busyobj with a mempool, instead of a private one-item pool.   Geoff Simmons
• [experimental-ims] 4d4e9f7 Use the req workspace, not the worker workspace, as we cannot trust that to be the same throughout the request.   Geoff Simmons
• [experimental-ims] 5454314 Remove now unused ws variable.   Geoff Simmons
• [experimental-ims] e88413b Allocate HTTP.status from http's designated workspace.   Geoff Simmons
• [experimental-ims] 1cdbbff NB: Major Reorg Commit   Geoff Simmons
• [experimental-ims] cd2ce96 Allocate the vbo and http's out of the backend workspace to avoid wasting partial pages with likely parameter values.   Geoff Simmons
• [experimental-ims] 8c2fe8d Clarify that certain parameters are likely to be rounded to page-size by the operating system.   Geoff Simmons
• [experimental-ims] fc387c3 Eliminate the mempool poll function again, we don't need it anyway.   Geoff Simmons
• [experimental-ims] c2a2576 Make sure to free workers cached busyobj when threads are reaped.   Geoff Simmons
• [experimental-ims] 21eb82b Rename variable to not confuse people on monday mornings.   Geoff Simmons
• [experimental-ims] 877f011 Wait a bit before trying to start/use varnishd again   Geoff Simmons
• [experimental-ims] 646d185 Fixup printf format on 32bit platforms   Geoff Simmons
• [experimental-ims] 8a75838 Fix up some sess_workspace leftovers   Geoff Simmons
• [experimental-ims] d119819 32 bit default parameter adjustments   Geoff Simmons
• [experimental-ims] b42a6d1 More 32 bit adjustments, now rename gzip_stack_buffer to gzip_buffer   Geoff Simmons
• [experimental-ims] 866b419 Fix a printf format issue with time_t   Geoff Simmons
• [experimental-ims] fb3cc4b Make fmt_bytes take an uintmax_t   Geoff Simmons
• [experimental-ims] b600f1a Don't use HUGE_VAL to signify no upper limit, we run into infinity in conversions to uintmax_t. Use zero to disable upper limit check. Check separately for byte sizes larger than the architecture can do.   Geoff Simmons
• [experimental-ims] 2e8ff87 Print out the right number in our diagnostics   Geoff Simmons
• [experimental-ims] 23cffba Check that workspace_backend is big enough, rather than run on the mercy of mallocs rounding algorithm.   Geoff Simmons
• [experimental-ims] d3f869a Remove sendfile(2) support, it doesn't seem to actually make any difference in practice and complicates the code and increases the size of storage data structures.   Geoff Simmons
• [experimental-ims] 7ec9794 Move res_mode from worker to req   Geoff Simmons
• [experimental-ims] 73b465a Integrate stream_ctx in struct req   Geoff Simmons
• [experimental-ims] ab47f76 Move VCL-controlled backend timeouts from struct worker to busyobj   Geoff Simmons
• [experimental-ims] ffea64d Shave 256 bytes of the worker thread footprint, by using the thread workspace for WRW also.   Geoff Simmons
• [experimental-ims] 1fd2741 Start abstracting the VSL handling a bit.   Geoff Simmons
• [experimental-ims] 892c92a Make refcount signed.   Geoff Simmons
• [experimental-ims] a5d9851 Introduce a function to initialize a VSL buffer   Geoff Simmons
• [experimental-ims] c9c0572 Use VSL_Setup() once more   Geoff Simmons
• [experimental-ims] ad277d2 Add a facility to give each session a unique number with a minimum of locking: Each pool fetches blocks of numbers to assign, then hands them out one by one using the existing lock.   Geoff Simmons
• [experimental-ims] c305b19 Forgot to add the lock type.   Geoff Simmons
• [experimental-ims] b762016 Give struct req its own buffer for batching VSL records   Geoff Simmons
• [experimental-ims] d98dee2 Give busyobj its own VSL staging buffer too.   Geoff Simmons
• [experimental-ims] f093a08 Tag vbc's with the vsl buffer they should log into. Update the buffers wid with the vbc's vsl_id. Flush the buffer when we recycle/close the vbc.   Geoff Simmons
• [experimental-ims] fe68764 Attach the vslbuffer to the vbc while in use and lend it the vbc's vsl_id. When the vbc is returned, restore the original vsl_id.   Geoff Simmons
• [experimental-ims] 75739f8 Split http_Setup() into http_Setup() and http_Teardown()   Geoff Simmons
• [experimental-ims] 628fdea Tag struct http with which vsl buffer to use.   Geoff Simmons
• [experimental-ims] 88fd1b2 Adjust testcases for changes struct http size   Geoff Simmons
• [experimental-ims] f7dbf53 Clean up a lot of now needless argument passing for struct http's VSL logging. More to come.   Geoff Simmons
• [experimental-ims] e9e3e2e Eliminate a lot of arguments to http_ functions which used to convey the VSL coords which are now a property of struct http.   Geoff Simmons
• [experimental-ims] f33a076 Give struct htc the same "tell it what vsl_buffer to use" treatment.   Geoff Simmons
• [experimental-ims] 2d49712 Change FetchError() to take busyobj as arg.   Geoff Simmons
• [experimental-ims] fe8ec80 Argument strength-reduction: from worker to dstats   Geoff Simmons
• [experimental-ims] ac85937 More argument strength reduction to reduce the spread of struct wrk   Geoff Simmons
• [experimental-ims] fe9d181 Align the iovec pointer   Geoff Simmons
• [experimental-ims] d2b5440 Clamp # of iovec to IOV_MAX   Geoff Simmons
• [experimental-ims] f1dfb81 Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache into experimental-ims   Geoff Simmons
• [experimental-ims] ee03e4a Removed superfluous parameters in cache_http.c, or narrowed their scope, following similar cleanup in master.   Geoff Simmons
• [experimental-ims] 40273a3 Clean up this file, and GC a unused counter.   Geoff Simmons
• [experimental-ims] bf5b757 Force this one into the same consistent layout.   Geoff Simmons
• [experimental-ims] a434ce4 Honor remove-flag also when processing comments in ESI parsing.   Geoff Simmons
• [experimental-ims] 28026e2 Merge master -> experimental-ims   Geoff Simmons
• [experimental-ims] 71b981f Use %ju for uintmax_t printf   Geoff Simmons
• [experimental-ims] f012e53 Make sure we have debug.sizeof in at least one test-case.   Geoff Simmons
• [experimental-ims] af2bfa9 Push struct worker further upwards   Geoff Simmons
• [experimental-ims] 60490c9 Increase backend_workspace of r01038.vtc from 7k to 8k.   Geoff Simmons
• [experimental-ims] bd9df97 The VGZ api does not need a worker, it just needs a vsl_log   Geoff Simmons
• [experimental-ims] 071f86b Merge branch 'master' of ssh://git.varnish-cache.org/git/varnish-cache   Geoff Simmons
• [experimental-ims] b5f1878 Remove the current "ping-pong" streaming code, it is in the way of the "real streaming" code we are working on.   Geoff Simmons
• [experimental-ims] 7703a5f Lend busyobj the workers dstat for the duration of Fetch_Body()   Geoff Simmons
• [experimental-ims] d387098 Make FetchStorage() take a busyobj instead of a worker as arg.   Geoff Simmons
• [experimental-ims] 5d4e2ce Move VGZ entirely to busyobj instead of worker   Geoff Simmons
• [experimental-ims] 4076126 Push worker->busyobj change up through ESI   Geoff Simmons
• [experimental-ims] 9b65ebc Push worker->busyobj shift up through VFP's.   Geoff Simmons
• [experimental-ims] 7d164e0 Most of these variables are not available in vcl_deliver   Geoff Simmons
• [experimental-ims] a18ff47 Merge master -> experimental-ims Since STV_alloc() was changed in master so that it assumes allocation is only for busyobj->fecthobj, there's a problem with the duplication of storage from stale_obj into the newly fetched obj -- for copying, it needs to allocate storage, and now just calls stv->alloc, without STV_alloc's logic to check for exhausted storage, nuke if necessary, etc. This is *not* safe now.   Geoff Simmons

Last message date: Wed Feb 29 15:17:12 UTC 2012
Archived on: Wed Feb 11 18:06:23 UTC 2026

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