Trivial, but I'm trying to avoid a context switch:
[[[
Index: subversion/libsvn_ra_neon/options.c
===================================================================
--- subversion/libsvn_ra_neon/options.c (revision 980674)
+++ subversion/libsvn_ra_neon/options.c (working copy)
@@ -144,6 +144,8 @@ parse_capabilities(ne_request *req,
*youngest_rev = SVN_INVALID_REVNUM;
/* Start out assuming all capabilities are unsupported. */
+ apr_hash_set(ras->capabilities, SVN_RA_CAPABILITY_PARTIAL_REPLAY,
+ APR_HASH_KEY_STRING, capability_no);
apr_hash_set(ras->capabilities, SVN_RA_CAPABILITY_DEPTH,
APR_HASH_KEY_STRING, capability_no);
apr_hash_set(ras->capabilities, SVN_RA_CAPABILITY_MERGEINFO,
Index: subversion/libsvn_ra_serf/options.c
===================================================================
--- subversion/libsvn_ra_serf/options.c (revision 980674)
+++ subversion/libsvn_ra_serf/options.c (working copy)
@@ -394,6 +394,8 @@ options_response_handler(serf_request_t *request,
serf_bucket_t *hdrs = serf_bucket_response_get_headers(response);
/* Start out assuming all capabilities are unsupported. */
+ apr_hash_set(orc->session->capabilities, SVN_RA_CAPABILITY_PARTIAL_REPLAY,
+ APR_HASH_KEY_STRING, capability_no);
apr_hash_set(orc->session->capabilities, SVN_RA_CAPABILITY_DEPTH,
APR_HASH_KEY_STRING, capability_no);
apr_hash_set(orc->session->capabilities, SVN_RA_CAPABILITY_MERGEINFO,
]]]
How would the lack of these lines cause neon/serf to behave when run against
old servers? (should they just loop because apr_hash_get() would return NULL?
but that's /not/ how they actually behave...)
Daniel
(it's trivial enough that I'll probably commit it later even if I can't test it)
