Commits · 10563cf5544f900b7c729174971a0d34c6d5e7ff · uplex-varnish / varnishevent

11 Sep, 2015 2 commits
- restore the simpler version of VSTAILQ_PREPEND · 10563cf5
  Geoff Simmons authored Sep 11, 2015
  
  10563cf5
- flush until it's all down the drain · b34d5501
  Geoff Simmons authored Sep 11, 2015
  
  b34d5501
10 Sep, 2015 2 commits
- use poll() (not select()) to wait for ready output with timeout · 1ea3ed91
  Geoff Simmons authored Sep 10, 2015
```
on output streams that may block
```
  1ea3ed91
- writer thread does no header lookups -- indices into the header · 4eb7a7e2
  Geoff Simmons authored Sep 10, 2015
```
tables are stored in compiled formats
```
  4eb7a7e2
09 Sep, 2015 2 commits
- idle pause at and of log is set automatically to match the tx read · 4eecae11
  Geoff Simmons authored Sep 09, 2015
```
rate, and is no longer configurable. eol counter and idle_pause are
logged
```
  4eecae11
- writer thread returns free lists before the reader is completely · 24d5a8b1
  Geoff Simmons authored Sep 09, 2015
```
exhausted
```
  24d5a8b1
08 Sep, 2015 2 commits
- bump version to 4.1.0 · 3b722d03
  Geoff Simmons authored Sep 08, 2015
  
  3b722d03
- - bugfixes, make check passes · 5f440d2a
  Geoff Simmons authored Sep 08, 2015
```
- Varnish:hitmiss and :handling are working again
- states of data objects are tracked more strictly
```
  5f440d2a
30 Jun, 2015 3 commits

complete refactoring of the internal data structures, so that the · e0413fdc

Geoff Simmons authored Jun 19, 2015

reader thread only copies the data needed for formatting -- the first
matching record --, and all accesses are indexed table lookups

'make check' does *not* succeed in this commit

e0413fdc

Revert "complete refactoring of the internal data structures, so that the" · 81593688

Geoff Simmons authored Jun 30, 2015

Currently failing assertions due to dirty records (with the occupied
flag set) being returned from the record free list.

This reverts commit a2af8368.

81593688

Revert "set a write memory barrier when clearing data before returning data" · 1e0264de

Geoff Simmons authored Jun 30, 2015

This did not resolve the problem with dirty records (with the occupied flag
set) being returned from the free list.

This reverts commit 182d509d.

1e0264de

19 Jun, 2015 3 commits
- set a write memory barrier when clearing data before returning data · 182d509d
  Geoff Simmons authored Jun 19, 2015
```
to free lists to ensure that cleared fields are seen by the reader
thread
```
  182d509d
- complete refactoring of the internal data structures, so that the · a2af8368
  Geoff Simmons authored Jun 19, 2015
```
reader thread only copies the data needed for formatting -- the first
matching record --, and all accesses are indexed table lookups

'make check' does *not* succeed in this commit
```
  a2af8368
- Revert "make the occupancy stats private to MON_, tighten the lock, and" · 529d30f3
  Geoff Simmons authored Jun 19, 2015
```
This reverts commit c4cbd278.
```
  529d30f3
02 Jun, 2015 1 commit
- make the occupancy stats private to MON_, tighten the lock, and · c4cbd278
  Geoff Simmons authored Jun 02, 2015
```
check by assert that they don't go negative
```
  c4cbd278
15 May, 2015 1 commit

Formatters now search for the first tag and/or header in a transaction · b1b22418

Geoff Simmons authored May 15, 2015

that are needed for the format, rather than the last. This is necessary
because Varnish 4.0.x logs are still mixing the URLs and headers of
ESI-included requests into the same log transaction, so that we were
getting the URL/headers of the last ESI include, rather than those
of the root request.

b1b22418

07 May, 2015 1 commit
- fix a bug that caused formatters to overwrite header payloads if · 5b67c574
  Geoff Simmons authored May 07, 2015
```
there is more than one in a record spanning more than one chunk
```
  5b67c574
23 Apr, 2015 5 commits
- get rid of the max.vcl_log parameter -- we can estimate one record · 142a1c14
  Geoff Simmons authored Apr 23, 2015
```
per VCL_Log formatter, since they are filtered to match the "header"
```
  142a1c14
- get rid of unused config params · 69dbb6a9
  Geoff Simmons authored Apr 23, 2015
  
  69dbb6a9
- VSB config params require VSB_finish() after intitialization · a1fb3063
  Geoff Simmons authored Apr 23, 2015
  
  a1fb3063
- writer thread returns free lists if its local list is larger than · 1972e1c6
  Geoff Simmons authored Apr 23, 2015
```
half of the global total for the data type
```
  1972e1c6
- writer thread returns free lists before re-opening output, and after · e59bc293
  Geoff Simmons authored Apr 23, 2015
```
a ready output timeout
```
  e59bc293
22 Apr, 2015 1 commit
- updated the documentation · 73d12385
  Geoff Simmons authored Apr 22, 2015
  
  73d12385
21 Apr, 2015 9 commits
- 'make clean' removes files created by the regression test · 9bee32fa
  Geoff Simmons authored Apr 21, 2015
  
  9bee32fa
- consistently name the element "record" in the sources (not "line" or · d16c8b3f
  Geoff Simmons authored Apr 21, 2015
```
"logline")
```
  d16c8b3f
- refactored log levels used for types of information: · d0b3925e
  Geoff Simmons authored Apr 21, 2015
```
NOTICE: lifecycle events
INFO: monitoring logs and informational messages at startup
CRIT: fatal errors at startup
ERR: recoverable errors after startup
ALERT: fatal errors after startup
```
  d0b3925e
- handle buffer sizes safely in handler.c · 77a06de3
  Geoff Simmons authored Apr 21, 2015
  
  77a06de3
- handle buffer lengths safely in varnishevent.c · 5b55fee7
  Geoff Simmons authored Apr 21, 2015
  
  5b55fee7
- handle buffer sizes safely in format.c · 24b34da8
  Geoff Simmons authored Apr 21, 2015
  
  24b34da8
- handle buffer sizes for config parameters safely · b0696adb
  Geoff Simmons authored Apr 21, 2015
  
  b0696adb
- don't used fixed-size buffers to read config files, unless we have · 551b0fe1
  Geoff Simmons authored Apr 21, 2015
```
to due to absence of getline()
```
  551b0fe1
- writer returns free lists without idling when reader free lists are · cb2cdd4f
  Geoff Simmons authored Apr 21, 2015
```
exhausted
```
  cb2cdd4f
20 Apr, 2015 8 commits
- reader thread uses -i filters for tags whose payloads do not need · bf1e6a72
  Geoff Simmons authored Apr 20, 2015
```
to be matched against a regex
```
  bf1e6a72
- refactor the declaration and initialization of VSBs in format.c · d84cc955
  Geoff Simmons authored Apr 20, 2015
  
  d84cc955
- remove an unnecessary condition on returning freelists in the writer · c1ffcdd6
  Geoff Simmons authored Apr 20, 2015
```
thread
```
  c1ffcdd6
- add a write memory barrier before submitting transactions, and a read · 3faa1fd7
  Geoff Simmons authored Apr 20, 2015
```
mb before reading them to format for log output
```
  3faa1fd7
- add a full memory barrier before the main loop in the writer thread, · 6ba2caa8
  Geoff Simmons authored Apr 20, 2015
```
to ensure that all initializations from the main thread are stored
```
  6ba2caa8
- test_format.c: more checking that the occupied field is set when required · 3d55a75d
  Geoff Simmons authored Apr 20, 2015
```
(and not "by accident" for stack-local data)
```
  3d55a75d
- in test_format.c before the get_hdr() test, init records and chunks to · 338795ca
  Geoff Simmons authored Apr 20, 2015
```
zero, and set the occupied field
```
  338795ca
- initialize, clear and dump the transaction parent vxid field · dd55e8ec
  Geoff Simmons authored Apr 20, 2015
  
  dd55e8ec