- 10 Jun, 2014 10 commits
-
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Fixes #1518
-
Lasse Karstensen authored
Add a bit more detail while cleaning up the language. Reformat for tabs instead of 7 spaces.
-
Lasse Karstensen authored
Using multiple names for the instanciateed directors in this document just makes it more confusing, since .add_backend() and .backend() is mostly identical for all of them.
-
Lasse Karstensen authored
-
Dag Haavi Finstad authored
Fixes: #1519
-
Poul-Henning Kamp authored
keep them in sync. If you want a custom reason field, set status first, then reason. With this in place, HTTP1_Write() doesn't need to worry about them being OK, and we can polish that code a fair bit. Turn a couple of boolean arguments into more readable magics.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
to follow shortly.
-
- 05 Jun, 2014 2 commits
-
-
Lasse Karstensen authored
-
Lasse Karstensen authored
The rst2man utility exists only in "newer" versions of python-docutils. Require at least 0.7, released July 2010.
-
- 04 Jun, 2014 5 commits
-
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
the pthread implementation in *every* test case.
-
Poul-Henning Kamp authored
Move policy decisions out where they belong and let the http_Filter function send the first like through unmolested. Rather than value specific functions, make a generic http_ForceField() function, because the name is cool.
-
Poul-Henning Kamp authored
"response/reason" confusion for good.
-
Poul-Henning Kamp authored
one function which does what you'd expect. Start allowing internal VCL use of *resp.status > 999. The idea is that we ignore the higher bits but it still has to end up >= 100, so 10404 becomes 404 and 20099 is illegal. (This change is not yet complete it will panic if you use it yet.) More http->failed stuff.
-
- 03 Jun, 2014 3 commits
-
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Usually, (probably always) this means failure to get workspace for modifications of the http. The intent is that this flag will have the same "latching" behaviour as error handling in VSB's: Once set, it stays set and nobody reads or writes the struct http any more. Setting the flag causes a SLT_Error message. Rewrite http_CollectHdr() to respect failed and optimize it slightly while were here.
-
Poul-Henning Kamp authored
-
- 02 Jun, 2014 4 commits
-
-
Martin Blix Grydeland authored
Fixes: #1514
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
- 27 May, 2014 6 commits
-
-
Poul-Henning Kamp authored
-
Martin Blix Grydeland authored
Other minor cleanups
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
This adds %I and %O formatters to varnishncsa, which are total bytes received and total bytes sent. These are similar to the same formatters in apache (with mod_logio). Useful for bandwidth accounting. Patch by: David Robertson
-
Poul-Henning Kamp authored
-
- 26 May, 2014 5 commits
-
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Dag Haavi Finstad authored
Verify that s_resp_bodybytes is indeed incremented the correct number of bytes on partial content responses.
-
Poul-Henning Kamp authored
Pointed out by: Dridi Boukelmoune
-
Poul-Henning Kamp authored
Fixes #1510
-
- 20 May, 2014 5 commits
-
-
Poul-Henning Kamp authored
is a contention point and scales negatively with number of thread-pools. Sum into a per-pool dstat structure instead, and have the first idle thread summ that into the global counters without blocking the pool mutex. Fixes #1495 (I hope)
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
also takes account of fetch jobs.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
what the real solution is going to be.
-