Spotted by: coredump

Writing to bereq.uncacheable in backend_response and backend_error would only
make sense in the context of a retry. In most real world cases, writing to
be*resp*.uncacheable would be intended in backend_response and backend_error.

This was mentioned in the user guide, but not in the man
page itself.

Pointed out by:		Brett Fitzgerald

This fixes a race where the client started freeing the body (which it
does in the case of new hit-for-pass object being inserted) before the
sanity functions were run.

Fixes: #1596

it is sane and pass it to streaming clients, provided we don't munge
the body along the way.

The ungzip'ed length stored as attribute has not been backported to
4.0, meaning on-the-fly ungzip will be delivered chunked.

Fixes: #1506

Conflicts:
	bin/varnishd/cache/cache_fetch.c

Conflicts:
	bin/varnishd/cache/cache_http1_deliver.c

Always interpret C-L in decimal

Fixes #1612
Testcase by: 	daghf

Conflicts:
	bin/varnishd/cache/cache_http1_deliver.c

And also don't disable the default compiler optimization.

Noted by:	Matt Robenolt

Implement -k argument in vut for use with varnishlog. This takes as an
argument the number of log transactions to be processed before
exiting. Useful to stop varnishlog after a given set of log data has
been produced.

Drop -s from the 'not implemented yet' list as this option has very
limited usefullness.

copy&paste.

Spotted by:	scn

After re-reading the patch, this isn't important enough to
warrant a changelog entry.

Spotted by: 	fgsch

implementation

dedicated setup function we have.

Fixes	#1553

Fixes #1561

that need to be purged we iterate until done.

Fixes #1551

Conflicts:
	bin/varnishd/cache/cache_hash.c

This reverts commit f54e14f6.

linebreaks in panic messages.

Conflicts:
	lib/libvmod_std/vmod_std.c

Conflicts:
	lib/libvmod_std/vmod_std.c

in all VCL_methods.

Apologies for the previous breakage.

Fixes #1546

Fixes: #1591