• Dag Erling Smørgrav's avatar
    Merged revisions... · 032e6e84
    Dag Erling Smørgrav authored
    Merged revisions 1743-1744,1746-1749,1755-1775,1777,1779-1793,1795,1797-1798,1800-1808,1810-1815,1817 via svnmerge from 
    svn+ssh://projects.linpro.no/svn/varnish/trunk/varnish-cache
    
    ........
      r1743 | ssm | 2007-07-20 14:43:54 +0200 (Fri, 20 Jul 2007) | 1 line
      
      Split debian package into varnish, libvarnish and libvarnish-dev
    ........
      r1748 | des | 2007-07-24 15:51:58 +0200 (Tue, 24 Jul 2007) | 2 lines
      
      Correct a comment
    ........
      r1749 | des | 2007-07-24 15:52:40 +0200 (Tue, 24 Jul 2007) | 2 lines
      
      Use strlen() directly.
    ........
      r1761 | cecilihf | 2007-07-25 10:39:10 +0200 (Wed, 25 Jul 2007) | 9 lines
      
      Implemented http purge with regexp. Example vcl usage:
      sub vcl_recv {
      	if (req.request == "REPURGE") {
      		purge(req.url);
      		error 404 "Purged";
      	}
      }
    ........
      r1762 | cecilihf | 2007-07-25 10:53:16 +0200 (Wed, 25 Jul 2007) | 2 lines
      
      Updated man page
    ........
      r1765 | cecilihf | 2007-07-25 13:09:06 +0200 (Wed, 25 Jul 2007) | 2 lines
      
      Use purge in function names instead of repurge and move VRT_purge to cache_vrt.c.
    ........
      r1766 | cecilihf | 2007-07-25 13:10:59 +0200 (Wed, 25 Jul 2007) | 2 lines
      
      Style...
    ........
      r1780 | des | 2007-07-28 12:03:29 +0200 (Sat, 28 Jul 2007) | 5 lines
      
      Reverse the logic for sp->wantbody: assume it is always wanted unless
      req.request is "HEAD".  This is what broke the RePurge test case.
      
      Ideally, sp->wantbody would be controllable by VCL.
    ........
      r1783 | des | 2007-07-30 10:03:42 +0200 (Mon, 30 Jul 2007) | 4 lines
      
      Avoid referencing <sys/cdefs.h> and __{BEGIN,END}_DECLS.
      
      Based on Theo Schlossnagle's Solaris portability patch.
    ........
      r1784 | des | 2007-07-30 11:49:05 +0200 (Mon, 30 Jul 2007) | 2 lines
      
      Avoid using non-portable <err.h>.
    ........
      r1785 | des | 2007-07-30 15:50:16 +0200 (Mon, 30 Jul 2007) | 3 lines
      
      Use mktime() rather than the unportable timegm().  The downside is that we're
      at the mercy of the TZ environment variable.
    ........
      r1786 | des | 2007-07-30 15:54:48 +0200 (Mon, 30 Jul 2007) | 2 lines
      
      Explicitly set TZ to GMT.
    ........
      r1787 | des | 2007-07-30 16:16:23 +0200 (Mon, 30 Jul 2007) | 4 lines
      
      Use fcntl(2)-style locks instead of non-portable flock(2)-style locks.
      
      Based on Theo Schlossnagle's Solaris portability patch.
    ........
      r1788 | des | 2007-07-30 16:19:45 +0200 (Mon, 30 Jul 2007) | 2 lines
      
      pid_t is not necessarily compatible with int.
    ........
      r1789 | des | 2007-07-30 16:22:00 +0200 (Mon, 30 Jul 2007) | 2 lines
      
      Bogons in previous commit.  Pass me the pointy hat...
    ........
      r1790 | des | 2007-07-30 16:31:16 +0200 (Mon, 30 Jul 2007) | 3 lines
      
      I accidentally committed the wrong patch in r1787; this corrects the logic
      used to determine which lock type (shared or exclusive) to use.
    ........
      r1791 | des | 2007-07-30 16:39:03 +0200 (Mon, 30 Jul 2007) | 2 lines
      
      Note that MAX_IOVS is intentionally not equal to IOV_MAX.
    ........
      r1792 | des | 2007-07-31 08:06:28 +0200 (Tue, 31 Jul 2007) | 2 lines
      
      Fine-tune the vhost example.
    ........
      r1793 | des | 2007-08-03 20:46:43 +0200 (Fri, 03 Aug 2007) | 2 lines
      
      Synchronize these two files.
    ........
      r1795 | phk | 2007-08-05 21:37:44 +0200 (Sun, 05 Aug 2007) | 9 lines
      
      Synchronize the paramters after we call their accessor functions rather
      than when we don't find one.
      
      This makes changing runtime paramters work again, without the need to
      ask for a nonexistent parameter to trigger the update.
      
      Ticket: 136
    ........
      r1797 | phk | 2007-08-05 22:17:49 +0200 (Sun, 05 Aug 2007) | 4 lines
      
      Make the nonexistence of "nobody" and "nogroup" users and groups nonfatal
      
      Ticket: 140
    ........
      r1798 | phk | 2007-08-05 22:26:09 +0200 (Sun, 05 Aug 2007) | 5 lines
      
      We need the math library on some systems.
      
      Ticket 138
    ........
      r1800 | phk | 2007-08-05 22:57:20 +0200 (Sun, 05 Aug 2007) | 2 lines
      
      Add a missing return: we always return after detecting the first error.
    ........
      r1801 | phk | 2007-08-05 22:57:32 +0200 (Sun, 05 Aug 2007) | 2 lines
      
      vrt.h is not needed here.
    ........
      r1802 | phk | 2007-08-05 23:06:41 +0200 (Sun, 05 Aug 2007) | 2 lines
      
      Add comment to remind myself.
    ........
      r1803 | phk | 2007-08-06 10:07:18 +0200 (Mon, 06 Aug 2007) | 14 lines
      
      Have the VCL compiler provide a hint about the worst case number of
      operations on the req.hash variable.
      
      It is only a hint, because it merely counts how many times the parser
      saw something being added to the req.hash variable.  If the operation
      was in a subroutine which was called multiple times, the hint will not
      reflect the number of actual operations.
      
      For now we will deal with that at runtime, at the expense of a
      failed transaction every time we run short.  If this becomes an issue,
      an extensive topological analysis of the VCL program can give us
      a definitive count.
    ........
      r1804 | phk | 2007-08-06 11:19:20 +0200 (Mon, 06 Aug 2007) | 3 lines
      
      Add piecemal crc32 function back, we will need it for the scatter/gather
      hash string.
    ........
      r1805 | phk | 2007-08-06 11:25:20 +0200 (Mon, 06 Aug 2007) | 13 lines
      
      Rewrite the req.hash implmentation:
      
      Instead of assembling the entire hash-string in the workspace, use
      a scatter gather approach, hinted by the VCL compiler.
      
      This eliminates the workspace reservation which prevented regsub() from
      working in vcl_hash, and reduces the size of the necessary workspace a
      fair bit as well, at the cost of a little bit of complexity in the
      hash implmentations.
      
      Closes ticket 137 and possibly 141
    ........
      r1806 | phk | 2007-08-06 11:28:44 +0200 (Mon, 06 Aug 2007) | 4 lines
      
      Forgot an assert.
      
      WS_Return() is now unused, comment it out.
    ........
      r1807 | phk | 2007-08-07 08:55:52 +0200 (Tue, 07 Aug 2007) | 4 lines
      
      When vcl_miss() take error action, remember to discard backend request.
      
      Ticket: 139
    ........
      r1808 | phk | 2007-08-07 09:23:10 +0200 (Tue, 07 Aug 2007) | 3 lines
      
      Implement "error" action in vcl_fetch() and vcl_deliver()
    ........
      r1810 | phk | 2007-08-08 10:49:11 +0200 (Wed, 08 Aug 2007) | 5 lines
      
      Elminiate a spurious message when sigchild looses a race to popen(2).
      
      This fixed #143
    ........
      r1811 | des | 2007-08-08 12:01:50 +0200 (Wed, 08 Aug 2007) | 3 lines
      
      Remove FreeBSD workaround; the FreeBSD ports tree has had working autotools
      for two weeks now.  Also fix inconsistent indentation.
    ........
      r1817 | des | 2007-08-09 13:19:20 +0200 (Thu, 09 Aug 2007) | 2 lines
      
      Rename libcompat to libvarnishcompat, and make it dynamic.
    ........
    
    
    git-svn-id: http://www.varnish-cache.org/svn/branches/1.1@1818 d4fa192b-c00b-0410-8231-f00ffab90ce4
    032e6e84
varnishd.c 13.4 KB