WT-2130 Don't round the split_pct to an allocation size.

The value is passed in via the API, and rounding makes it likely we
won't split as the user expects. Especially if the split percentage
is high, when we'll likely round up to 100%.

1 files changed, 11 insertions, 2 deletions

diff --git a/src/reconcile/rec_write.c b/src/reconcile/rec_write.c
index 332449027a9..0e25b4bde3c 100644
--- a/src/reconcile/rec_write.c
+++ b/src/reconcile/rec_write.c
@@ -1889,8 +1889,7 @@ __wt_split_page_size(WT_BTREE *btree, uint32_t maxpagesize)
 	 * we don't waste space when we write).
 	 */
 	a = maxpagesize;			/* Don't overflow. */
-	split_size = (uint32_t)
-	    WT_ALIGN((a * (u_int)btree->split_pct) / 100, btree->allocsize);
+	split_size = (uint32_t)((a * (u_int)btree->split_pct) / 100);
 
 	/*
 	 * If the result of that calculation is the same as the allocation unit
@@ -3286,6 +3285,16 @@ supd_check_complete:
 		}
 	}
 
+	bnd->entries = r->entries;
+	/* Output a verbose message if we create a page without many entries */
+	if (WT_VERBOSE_ISSET(session, WT_VERB_SPLIT) && r->entries < 6)
+		WT_ERR(__wt_verbose(session, WT_VERB_SPLIT,
+		    "Reconciling page with memory footprint %" PRIu64
+		    ", %" PRIu32 " th page has"
+		    " %" PRIu32 " entries, %s, split state: %d\n",
+		    r->page->memory_footprint, r->bnd_next, r->entries,
+		    F_ISSET(r, WT_EVICTING) ? "evict" : "checkpoint",
+		    r->bnd_state));
 	WT_ERR(__wt_bt_write(session,
 	    buf, addr, &addr_size, false, bnd->already_compressed));
 	WT_ERR(__wt_strndup(session, addr, addr_size, &bnd->addr.addr));