- 11 May, 2011 22 commits
-
-
Tollef Fog Heen authored
-
Tollef Fog Heen authored
-
Tollef Fog Heen authored
-
Tollef Fog Heen authored
-
Tollef Fog Heen authored
It does not make sense to specify less that 1M of storage, so error out if a user specifies -smalloc,512 for instance. They probably meant -smalloc,512M, but require them to be explicit. Adjust C00027 test case to match
-
Tollef Fog Heen authored
reworked from a commit to 2.1 by Bjørn Ruberg.
-
Kristian Lyngstol authored
-
Kristian Lyngstol authored
-
Kristian Lyngstol authored
Also removed the $Id$, which was both incorrect and now obsolete.
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5447 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5438 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5437 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5430 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5424 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5054 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5049 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@4766 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@4725 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Tollef Fog Heen authored
git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@4711 d4fa192b-c00b-0410-8231-f00ffab90ce4
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Expose vreadfd(), also with a size pointer return argument. Use vreadfd() in std.fileread()
-
- 10 May, 2011 3 commits
-
-
Poul-Henning Kamp authored
A couple of constifications while I wait for the tests to run. Spotted by: DocWilco
-
Poul-Henning Kamp authored
Submitted by: ehocdet Fixes #911
-
Poul-Henning Kamp authored
still need the object. Instead, add a WRW_error() function to query if there are WRW trouble. Fixes #904
-
- 09 May, 2011 1 commit
-
-
Poul-Henning Kamp authored
not gunzip, send content-length to client rather than chunked. This makes progress-bars possible in browsers.
-
- 06 May, 2011 7 commits
-
-
Tollef Fog Heen authored
Add -O option to turn off grouping. Get rid of blank line separating requests.
-
Tollef Fog Heen authored
Rename -o added in last commit to -m, add documentation.
-
Tollef Fog Heen authored
Make -o tag:regex only act on transactions where the regex matches tag, generalised from what varnishlog and varnishncsa already had support for. In addition, support multiple -o options which are then and-ed together.
-
-
Poul-Henning Kamp authored
an object at the same time a request-lookup does. The functional solution to this race is to hold the objhdr mutex, which we already hold briefly to get the refcount, also when we do the check. In terms of souce code this inlines the problematic HSH_FindBan() function in the lurker. And since that was major surgery af few other acts of improvement was carried out also. Most notably, we will now scan all applicable bans in the lurker and not give up on the first ban that tests req.* variables.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
-
- 05 May, 2011 1 commit
-
-
Tollef Fog Heen authored
Update docs for purge → ban rename plus various typos and grammatical errors.
-
- 03 May, 2011 1 commit
-
-
Tollef Fog Heen authored
Commit 3bd239e9 makes it so we set the object's last-modified timestamp to the current timestamp if there's no Last-Modified header. This makes it so we can do conditional gets even when there's no Last-Modified or ETag header from the backend. However, the time entered was a float, so we effectively ended up with an off-by-up-to-one-minus-epsilon error. This lead to requests ending up in a 304/200 response code round-dance. We now round the timestamp down to the nearest whole second instead, which should fix this bug. Thanks to lrowe for a helpful bug report. Fixes: #907
-
- 02 May, 2011 5 commits
-
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
we simply have to abandon it.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Edging closer to streap+gunzip
-
Poul-Henning Kamp authored
requests. This should make it possible to deliver arbitrarily sized objects with pass mode.
-