[geary] Fix crash in progress meter when populating search table. Bug 776383.

[Michael Gratton <mjog src gnome org>]

commit 933dae2ddaf6791f5cf6a65f43a9c18de727c51a
Author: Michael James Gratton <mike vee net>
Date:   Fri Jan 13 16:08:09 2017 +1100

    Fix crash in progress meter when populating search table. Bug 776383.
    
    * src/engine/imap-db/imap-db-account.vala
      (Account::populate_search_table_batch_async): total_unindexed was wildy
      inaccurate, since there may be been messages missing from MessageTable,
      MessageSearchTable, or both. When clamped to 0, but when there actually
      were messages to index, this would trigger an assert in the search
      progress meter. So for now instead use something wildly overinflated
      but guaranteed to not trigger the assert.

 src/engine/imap-db/imap-db-account.vala |   14 ++++++++------
 1 files changed, 8 insertions(+), 6 deletions(-)
---
diff --git a/src/engine/imap-db/imap-db-account.vala b/src/engine/imap-db/imap-db-account.vala
index 5fb8b60..86ce37c 100644
--- a/src/engine/imap-db/imap-db-account.vala
+++ b/src/engine/imap-db/imap-db-account.vala
@@ -1637,12 +1637,14 @@ private class Geary.ImapDB.Account : BaseObject {
             stmt.bind_uint(0, Geary.ImapDB.Folder.REQUIRED_FTS_FIELDS);
             stmt.bind_uint(1, Geary.ImapDB.Folder.REQUIRED_FTS_FIELDS);
             Gee.HashSet<int64?> message_ids = do_build_rowid_set(stmt.exec(cancellable), cancellable);
-            
-            // guesstimate at the number that need to be indexed ... technically if this is zero then
-            // we're done, but for safety allow the chaffing to go through, in case there are search
-            // rows that do not correspond to message rows (which is bad but not fatal)
-            total_unindexed = (message_ids.size - search_ids.size).clamp(0, int.MAX);
-            
+
+            // This is hard to calculate correctly without doing a
+            // join (which we should above, but is currently too
+            // slow), and if we do get it wrong the progress monitor
+            // will crash and burn, so just something too big to fail
+            // for now. See Bug 776383.
+            total_unindexed = message_ids.size;
+
             // chaff out any MessageTable entries not present in the MessageSearchTable ... since
             // we're given a limit, stuff messages req'ing search into separate set and stop when limit
             // reached