to avoid dropping an object on out of storage conditions.

I belive this...

Fixes	#953

Otherwise please reopen.

Ensure that stats are correct if allocations fail.

Don't rely on the VSC counter for checking the limit.

It seems most RPM-based distros has the -p switch to killproc and
status those days, so only blacklist the ones we know about that does
not have it.

Fixes: #969

Make sure we compensate for sizeof(int) and the stuff we have already
read.

Fixes: #848

LLVM C-compiler.

You have no idea how much or how long (20+ years!) I have missed having
an alternative to GCC.

Remember:

  Mono cultures are particularly bad for you, when they are good for you

Kudos to:       The LLVM team.

Patch by:	scoof
Fixes		#961

as per discussions in httpbis WG

We're seeing at least up to 4k in the wild, so increase the limit
somewhat.

See: #939

Switch back from beta2 to trunk meaning RPMs can be built directly
from the tarballs again.

Fixes: #955

"-" is correct according to the Apache LogFormat documentation on
http://httpd.apache.org/docs/current/mod/mod_log_config.html

This reverts commit fe7e315c.

See #949

Print out the rest of the format, changing the error from

  "Unknown format character: {"

to:

  "Unknown format starting at: %{asdf}x"

Thanks to new GCC for spotting those.

Fall back to "-" if the handling has not been decided yet.

Fixes: #950
Fixes: #944
Fixes: #918

utils expect to receive SIGHUP).

Fixes: #947

Fixes: #949

different functions.

I have not managed to write a vtc case for this one, but I am pretty
sure this:

Fixes:	#945

a21746d2.

Fixes: #941

of source code.

This is substantially DocWilcos fix, but instead of adding yet another
unique numbering variable, I have collapsed the once we have into
a single one.

Fixes #948

Use it to hold the derived vary string for the request (if any) and
skip busy objects with non-matching Vary during hash lookup.

Idea by:	sky

memlevel 8.

Fix testcases which depend on older values.