Re: svn commit: r1501654 - in /subversion/branches/tristate-chunked-request/subversion: include/svn_config.h libsvn_ra_serf/options.c libsvn_ra_serf/ra_serf.h libsvn_ra_serf/serf.c libsvn_ra_serf/util.c libsvn_subr/config_file.c

On Wed, Jul 10, 2013 at 1:59 AM, <breser_at_apache.org> wrote:
>...
> * subversion/libsvn_ra_serf/ra_serf.h
> (svn_ra_serf__session_t): Replace using_chunked_requests and busted_proxy
> members with set_CL and using_chunked_requests members respectively.
> The new using_chunked_requests member is now a tristate instead of boolean.

I don't think you want to do it this way.

* keep the (now-renamed) "detect_chunking" flag separate, and in the
session. it may be "local" to open() right now, but in the future it
may be used as part of an operation's "first request"

* detect_chunking remains a boolean, and it can be set to FALSE when
the first probe is done (whether that is OPTIONS or a REPORT or...)

* don't alter the probe or the request logic. it is fine. consider:
  - "auto": set detect_chunking, set using_chunked_requests
  - "yes": clear detect_chunking, set using_chunked_requests
  - "no": clear detect_chunking, clear using_chunked_requests
  In other words, you can get away with a simple change to serf.c for
the "auto" work

And sorry: my r1502097 change conflicts with your branch. But
considering how simple this change *should* be, I think you can just
toss the branch and suggest a one-file patch.

(to be clear: I'm still feeling about -0.7 on backporting an "auto"
patch to 1.8.x; my current feeling is that a knob/upgrade
recommendation is sufficient, and we improve the situation in the
1.9.x series)

Cheers,
-g
Received on 2013-07-11 06:34:36 CEST