- 21 May, 2016 12 commits
-
-
Poul-Henning Kamp authored
and the VCLI core serve code.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
resulting in alphabetically sorted command listing. Not what I set out to do right now, but somebody asked for this recently and I happened to spot a way to do it now...
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
whitespace) testfiles. Spotted by: Coverity
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
This makes it possible to add a "varnishd -x dumprstcli" command.
-
- 20 May, 2016 2 commits
-
-
Nicole Izumi authored
-
Poul-Henning Kamp authored
-
- 19 May, 2016 7 commits
-
-
Nils Goroll authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Handle symbols are ACL, BACKEND, PROBE and STEVEDORE, they are not variables, but "things which do stuff". The only real difference from variables (like resp.status) is that when they are stringified (resp.http.foo = BLA) they return the name of the thing (ie: backend, acl ... name) rather than a value, which they don't have. Handles are by definition available everywhere. Global symbols are handles available in any VCL (ie: STEVEDOREs).
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
- 18 May, 2016 6 commits
-
-
Poul-Henning Kamp authored
This means that if you: acl foobar { .... } sub vcl_recv { set req.http.acl = foobar; } Then the HTTP header gets set to the name of the ACL, ie: "foobar"
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
people to throw return values away if the want to, but not to use VOID in expressions.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Submitted by: rezan Fixes: #1931
-
- 17 May, 2016 1 commit
-
-
Nils Goroll authored
In particular, this avoids a later backend.list attempt on a file descriptor to a dead process and, consequently, avoids a long delay in varnishtest when varnishd cant be started (for instance, because it is not in PATH).
-
- 16 May, 2016 5 commits
-
-
Federico G. Schwindt authored
Related to #1924.
-
Federico G. Schwindt authored
Fixes #1924.
-
Federico G. Schwindt authored
-
Federico G. Schwindt authored
-
Nils Goroll authored
ban_lurker_work completes bans by moving ocs up the ban list. Unless there are REQ bans, this leaves a tail of completed bans which, previously, got marked by ban_mark_completed() only to get removed by ban_clean_tail() in the next step We now clean out the completed tail in the first place, avoiding the extra marking step.
-
- 14 May, 2016 1 commit
-
-
Nils Goroll authored
-
- 13 May, 2016 1 commit
-
-
Nils Goroll authored
-
- 11 May, 2016 4 commits
-
-
Poul-Henning Kamp authored
-
Nils Goroll authored
-
Nils Goroll authored
Fixes https://github.com/nigoroll/varnish-modules/issues/1
-
Nils Goroll authored
Fixes #1873
-
- 10 May, 2016 1 commit
-
-
Poul-Henning Kamp authored
-