May 2015 Archives by date

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

Starting: Mon May 4 09:05:09 UTC 2015
Ending: Sun May 31 06:33:42 UTC 2015
Messages: 184

• [master] 9ea6149 Add a tiny bit of compile time type-safety, just because we can.   Poul-Henning Kamp
• [master] 419f983 Enable PCRE JIT-er by default   Tollef Fog Heen
• [master] e012bc3 Introduce a req->resp_len field to hold the number of bytes we actually expect to deliver, and use it to eliminate most of the C-L header munging.   Poul-Henning Kamp
• [master] 6acf177 Major tour through the Range code to make us more strict and RFC7233 conforming.   Poul-Henning Kamp
• [master] a8d0651 One "no backend connection" is enough   Federico G. Schwindt
• [master] 4e5c8ab Fail gracefully if we can't get a backend on pipe   Federico G. Schwindt
• [master] dd54ea0 Testcase for #1576   Federico G. Schwindt
• [master] cfd288a Add "recv N" action   Poul-Henning Kamp
• [master] 24f88f8 Add a session close value for clients asking for bogus ranges on streaming objects.   Poul-Henning Kamp
• [master] cbb0356 Communicate the session close reason as a negative in the sess->fd so the SessClose VSL can reporte it correctly.   Poul-Henning Kamp
• [master] 97434c7 Handle Range requests on streaming objects better.   Poul-Henning Kamp
• [master] 8552e75 Use consistent type for buffers.   Poul-Henning Kamp
• [master] ffe29fe Workaround: Compile even if PCRE is too old to have JIT   Poul-Henning Kamp
• [master] ce89ba7 Only enable PCRE-JIT for PCRE version >= 8.32   Poul-Henning Kamp
• [master] bb99710 Let varnishtest know if we use pcre_jit   Poul-Henning Kamp
• [master] f6d7149 Add an assert so I don't again botch the SessClose duration field.   Poul-Henning Kamp
• [master] b43f31e Give chunked precedence over C-L per RFC7230   Poul-Henning Kamp
• [master] 9ab5669 Remove any C-L header if T-E: chunked is present.   Poul-Henning Kamp
• [master] 18a06da Add a test for HTTP/0.9 deprecation   Poul-Henning Kamp
• [master] 93820a1 Implement a resp.is_streaming status variable   Martin Blix Grydeland
• [master] ea7a162 If there is a C-L in the object, trust it for Range header purposes.   Poul-Henning Kamp
• [master] 16bc126 Try to make the new tests more robust   Poul-Henning Kamp
• [master] cd0329b Take another shot at stability in this case   Poul-Henning Kamp
• [master] e2e4542 Give up synchronizing the last test.   Poul-Henning Kamp
• [master] 00da3e1 I guess none of these two test can be reliably sync'ed across platforms, but I'm not quite sure I know why...   Poul-Henning Kamp
• [master] 5aaf7db Like any test involving streaming this looks like it will become a unstable beast :-/   Poul-Henning Kamp
• [master] 42c33b0 Prepare skeleton changes and upgrade sections.   Lasse Karstensen
• [master] f12e24e Remove extra whitespace in generated C code.   Lasse Karstensen
• [master] d135985 Use PROXY syntax.   Lasse Karstensen
• [master] cf08566 Document new -a format.   Lasse Karstensen
• [master] 668be11 Update to regular style and prose.   Lasse Karstensen
• [master] 86300e2 Remove some debugging, polish   Poul-Henning Kamp
• [master] 26c460c Turn ESI_Deliver() inside-out, so that we iterate over the storage, consuming ESI string as we go, rather than the other way around.   Poul-Henning Kamp
• [master] e76ad2d Make the meaty part of ESI_Deliver() look like a VDP function.   Poul-Henning Kamp
• [master] 0868f6d white space ocd   Poul-Henning Kamp
• [master] d8d9cd0 Whitespace ocd   Poul-Henning Kamp
• [master] dc4e747 Call ESI_Deliver() as a VDP.   Poul-Henning Kamp
• [master] 092ab4d Get closer to req->resp_len meaning "bytes to actually send, if known"   Poul-Henning Kamp
• [master] b86ffb9 Move 304/C-L/Body logic earlier, making it simpler   Poul-Henning Kamp
• [master] 7333dde Pull the protocol-agnostic bits of VDP (esi, range, gunzip) out of V1D and plunk it down in req_fsm for now.   Poul-Henning Kamp
• [master] 8c0af33 Fix an attempt to double close a session.   Poul-Henning Kamp
• [master] d0012bd Retire the backend_toolate counter.   Poul-Henning Kamp
• [master] d573f35 Only close backend connection on retry if it has a body   Poul-Henning Kamp
• [master] 735523e Fail if backend workspace doesn't have room for htc   Poul-Henning Kamp
• [master] ade0db0 Stop the epoll_waiter from holding on to fd's while stuff happens, it is supected of being the cause of the race in #1675 amongst other evils.   Poul-Henning Kamp
• [master] 7f71fe6 Avoid a length=0 panic where the length can actually be zero.   Poul-Henning Kamp
• [master] 4e5107d Attempt to allocate the htc before the backend connection, its cheaper to eliminate the exceedingly unlikely resource leak Coverity spotted that way.   Poul-Henning Kamp
• [master] f1501b5 try to get more comprehensive core dumps on test timeouts   Poul-Henning Kamp
• [master] 303f560 Eliminate ws argument from VCL_bla_method() functions.   Poul-Henning Kamp
• [master] d9c1916 Add a per-method specific pointer to the vrt-context, and use it to slim down req/busyobj a tiny bit.   Poul-Henning Kamp
• [master] b588787 Fix comment.   Poul-Henning Kamp
• [master] 3cc846e Introduce "transports" which is how responses to requests gets delivered.   Poul-Henning Kamp
• [master] ca50ccd Start using the "transport" mechanism for esi includes   Poul-Henning Kamp
• [master] 8b1f899 Make ESI delivery less magic.   Poul-Henning Kamp
• [master] 6764fbb Add a new test-case.   Poul-Henning Kamp
• [master] 01bd798 Shuffle order of functions   Poul-Henning Kamp
• [master] 68dc19d More reordering   Poul-Henning Kamp
• [master] e3712a4 Slight improvement to our symbol lookup hack.   Poul-Henning Kamp
• [master] 3341de5 Straighten out ESI delivery some more.   Poul-Henning Kamp
• [master] fc53691 Make this test-case even more strange.   Poul-Henning Kamp
• [master] ede0ff5 Gross hack, but could be handy another time.   Poul-Henning Kamp
• [master] 4ccf3b5 Remove the ESI/gzip CRC fields from struct req, and pass them as priv's inside ESI's delivery code.   Poul-Henning Kamp
• [master] b0bea23 Further simplicfication of ESI delivery   Poul-Henning Kamp
• [master] de91486 Add explanation for not streaming   Poul-Henning Kamp
• [master] 3bd29b4 Be much more cautions about create C-L headers from OA_GZIPBITS   Poul-Henning Kamp
• [master] 6eb3f1e Update the magic gzip bit positions whenever we spot them changing.   Poul-Henning Kamp
• [master] 2679bb6 Allow return(abandon) in vcl_backend_error{}   Poul-Henning Kamp
• [master] 2e964eb Document new TCP endpoint variables.   Lasse Karstensen
• [master] eb8f35b Close a race condition in the backend/tcp_pools interface.   Poul-Henning Kamp
• [master] c2cb5e7 Add read-only access to the top level request in an ESI tree.   Poul-Henning Kamp
• [master] d7b9296 Certain failure scenarios, for instance PROXY protocol syntax errors, are handled so fast that there may be no numerical difference between the start and end session timestamps.   Poul-Henning Kamp
• [master] 24d8af2 Improve stability of this test case   Poul-Henning Kamp
• [master] 415c7d8 Improve EOF detection.   Poul-Henning Kamp
• [master] b2e9410 Improve stability of these two testcases.   Poul-Henning Kamp
• [master] f8dd42a Expand the SIGSEGV handler to also cover SIGBUS   Poul-Henning Kamp
• [master] b59bf49 Alignment may push the size up   Poul-Henning Kamp
• [master] b703b6b Rewrite for more robust handling of server write failures   Poul-Henning Kamp
• [master] 0c3eaeb Increase alignment to uintmax_t to cater for 32bit userlands on 64bit processors.   Poul-Henning Kamp
• [master] cd55aa7 Make the ban header a multiple of 8 bytes for alignment   Poul-Henning Kamp
• [master] f17bd99 Brute force align ban-strings, to see if we can make the sparc64 box happy.   Poul-Henning Kamp
• [master] ab38b9b Fix a brain-o uncovered by running our test-cases thousands of times over the weekend:   Poul-Henning Kamp
• [master] 0f3be9f Tweak l00002.vtc   Martin Blix Grydeland
• [master] 1de529d Tweak l00005.vtc to make it less error prone   Martin Blix Grydeland
• [master] 2b1ac1c Ref: https://tools.ietf.org/html/rfc7231#section-6.1   Nils Goroll
• [master] be28935 Avoid passing a NULL z_stream msg to vsnprintf via VSLb and VFP_Error   Nils Goroll
• [master] a8b61cc fix an error message   Poul-Henning Kamp
• [master] 51f4c23 White-space ocd   Poul-Henning Kamp
• [master] b1dd667 Add a new mode to server instance 'b0', where it acts like "a normal webserver" and dispatches a thread for each incomming connection.   Poul-Henning Kamp
• [master] 51c0b18 Don't use libgzip internal #definery outside libgzip   Poul-Henning Kamp
• [master] ffc28ea A random attempt at debugging that shouldn't have been committed   Poul-Henning Kamp
• [master] ba2d2e0 Add a needless return for GCC's pointless warning.   Poul-Henning Kamp
• [master] 490a552 We cannot expect there to be a full second between the Bereq/Beresp timestamps, when the Bereq timestamp is taken after our write(2) to the socket, which may have caused us to loose the CPU.   Poul-Henning Kamp
• [master] 85729c5 Squash trailing whitespaces.   Lasse Karstensen
• [master] 730e185 clarifying comments   Nils Goroll
• [master] 54fbace Split waiter functions into the ones which deal with waiters (Waiter_*) and the ones which deal with fd's being waited on (Wait_*)   Poul-Henning Kamp
• [master] c960cd1 sort functions into Wait_ and Waited_   Poul-Henning Kamp
• [master] 77dd6bc Disable all waiters but the 'poll' waiter.   Poul-Henning Kamp
• [master] 5947441 Add a FALL-THROUGH comment   Poul-Henning Kamp
• [master] aeee5b5 If we rely on VSL order, space the requests a fraction of a second to ensure this order   Poul-Henning Kamp
• [master] cbc829f Scrape back the waiter API, and rewrite the poll waiter to handle sparse fd-sets efficiently.   Poul-Henning Kamp
• [master] f2c1e64 Patch from dridi to replace CLI with mutext for locking backend config.   Poul-Henning Kamp
• [master] 385930c Rework the backend/waiter stuff.   Poul-Henning Kamp
• [master] 64fbc76 Bring back the kqueue waiter.   Poul-Henning Kamp
• [master] 9500bd7 Also delete the mutex.   Poul-Henning Kamp
• [master] dc698d7 one fall through comment should be enough for everyone   Nils Goroll
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] 49712a5 vmods running vmodtool require python, so check for it   Federico Schwindt
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] dc698d7 one fall through comment should be enough for everyone   Poul-Henning Kamp
• [master] 49712a5 vmods running vmodtool require python, so check for it   Federico Schwindt
• [master] 476589b Stall until waiter is drained.   Poul-Henning Kamp
• [master] 50c9933 Improved shutdown   Poul-Henning Kamp
• [master] d74fd18 Polish   Poul-Henning Kamp
• [master] 27bab72 Update and reenable epoll(2) waiter to new reality.   Poul-Henning Kamp
• [master] 12ace86 Stabilize this test a bit   Poul-Henning Kamp
• [master] dc698d7 one fall through comment should be enough for everyone   Nils Goroll
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] da26134 Stabilize this test   Poul-Henning Kamp
• [master] 71dfce8 Only explicitly show param default if it not set to that default   Poul-Henning Kamp
• [master] 0cba6f8 Convert the parameter array to a list of parameters.   Poul-Henning Kamp
• [master] 3327577 Only look for accept-filters if platform supports them.   Poul-Henning Kamp
• [master] 39e0bc5 Start tabelizing the trivial parameters.   Poul-Henning Kamp
• [master] b1ef42b Fix unterminated strings in #if 0 part of file.   Poul-Henning Kamp
• [master] cda99e3 Move variable inside #ifdef   Poul-Henning Kamp
• [master] d154379 Fix a loop doing shutdown   Poul-Henning Kamp
• [master] 49712a5 vmods running vmodtool require python, so check for it   Federico Schwindt
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] eccb26a Revert alignment to sizeof(void*). When I have a chance, I'll add the necessary #ifdef for the sparc64/32bit_userland thing.   Poul-Henning Kamp
• [master] 3f1704f Be more permissive about timing.   Poul-Henning Kamp
• [master] 940d545 Be consistent about jit on/off flag.   Poul-Henning Kamp
• [master] 581f438 Reduce the default for pcre_match_limit_recursion to 20.   Poul-Henning Kamp
• [master] 341d470 Remove the "feature pcre_jit" we don't use it anymore.   Poul-Henning Kamp
• [master] 4871058 Be even more tolerant of timing (for the jenkins::arm box)   Poul-Henning Kamp
• [master] fd2cdbb Remove the magic variable used to inform varnishtests now defunct "feature pcre_jit".   Poul-Henning Kamp
• [master] a54f85b Try to make this test run on el5-x86_64   Poul-Henning Kamp
• [master] a2be22d Cosmetic   Federico G. Schwindt
• [master] 2c41db8 Make this show under apropos varnish   Federico G. Schwindt
• [master] 5941d35 Spelling and minor changes   Federico G. Schwindt
• [master] e05ac94 Braino.   Poul-Henning Kamp
• [master] 49712a5 vmods running vmodtool require python, so check for it   Federico Schwindt
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] 49712a5 vmods running vmodtool require python, so check for it   Poul-Henning Kamp
• [master] 49712a5 vmods running vmodtool require python, so check for it   Nils Goroll
• [master] 2c022ed The changes to the backend/waiter has changed the dyamics of closing backend connections enough to make previously almost stable test-cases more unstable.   Poul-Henning Kamp
• [master] 720d0d5 Move waiter selection from param to command-line and pave the road for them taking parameters.   Poul-Henning Kamp
• [master] 330cd26 Concentrate waiter stuff in waiter includes   Poul-Henning Kamp
• [master] d0b4ba4 Add a missing line continuation   Poul-Henning Kamp
• [master] 26f6726 Be consistent about filenames: mgt_ is a prefix   Poul-Henning Kamp
• [master] 0d9759c Introduce the struct "waitfor" so that the same waiter can be used for more than one kind of waiting.   Poul-Henning Kamp
• [master] 49712a5 vmods running vmodtool require python, so check for it   Tollef Fog Heen
• [master] 49712a5 vmods running vmodtool require python, so check for it   Dridi Boukelmoune
• [master] 49712a5 vmods running vmodtool require python, so check for it   Federico Schwindt
• [master] 68cf6f8 Split 'struct pool' into a private include file.   Poul-Henning Kamp
• [master] f71729a Move pool stuff related to the workthreads of each pool to cache_wrk.c, but retain stats stuff (since it runs across pools) in cache_pool.c   Poul-Henning Kamp
• [master] 841abfe Absorb struct sesspool into struct pool   Poul-Henning Kamp
• [master] 49712a5 vmods running vmodtool require python, so check for it   Dridi Boukelmoune
• [master] 15d4d99 remove two unused fields   Poul-Henning Kamp
• [master] f2d75dc Constify the comparison function prototype   Poul-Henning Kamp
• [master] 0f03834 Give the kqueue waiter a binheap to track timeouts.   Poul-Henning Kamp
• [master] 754f386 Make the poll waiter also use the binheap   Poul-Henning Kamp
• [master] 1225301 Ditch the per tcp-pool waiters and recycle backend connections on the pool waiter for the working thread doing the recycling.   Poul-Henning Kamp
• [master] 8b5670d Disable epoll waiter for now.   Poul-Henning Kamp
• [master] 427e00c Eliminate the list of struct waited, a simple count will do.   Poul-Henning Kamp
• [master] c037d5b Rewrite the epoll waiter to the new reality.   Poul-Henning Kamp
• [master] b1b3c93 init jails when called with no arguments   Nils Goroll
• [master] c3d7467 T_arg has a sensible default   Nils Goroll
• [master] a942313 minor docfix   Nils Goroll
• [master] 98491be http is 80/tcp   Nils Goroll
• [master] b2065e8 listen to *:80 by default as documented in varnishd.rst   Nils Goroll
• [master] a200018 we can run without any arguments   Nils Goroll
• [master] 74e2dd7 trying to guess phks intention   Nils Goroll
• [master] 28964cf Add back the ports waiter with the existing (struct waited *) passing though port_send   Nils Goroll
• [master] 7ed785a Put waiter.h in the right autocrap-category.   Poul-Henning Kamp
• [master] 9123362 put the session's struct waited on the ws while waiting   Nils Goroll
• [master] 9597fcd improve the note on concurrency   Nils Goroll
• [master] 020aaff save another 8 bytes in struct sess by removing the waited pointer   Nils Goroll
• [master] c4462c3 Add an overbroad #ifdef to catch the 32bit userland on 64bit SPARC config we have in Jenkins.   Poul-Henning Kamp
• [master] fa84722 Staticize.   Poul-Henning Kamp
• [master] 519fca1 Move more parameters over to the new table format.   Poul-Henning Kamp
• [master] ec38fd3 Remove the busyobj_worker_cache parameter, which has defaulted to "off" for a very long time and never been shown to actually improve performance.   Poul-Henning Kamp
• [master] 0710944 Staticize.   Poul-Henning Kamp
• [master] e93ea04 Add a facility to schedule at task on the thread pools in round-robin order.   Poul-Henning Kamp
• [master] 8b122ca Add a needless return for gcc   Poul-Henning Kamp

Last message date: Sun May 31 06:33:42 UTC 2015
Archived on: Wed Feb 11 18:06:37 UTC 2026

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