[geary/mjog/579-logging-deadlock] Fix deadlock when finalising log records generates more logging

[Michael Gratton <mjog src gnome org>]

commit 6d74f4e50dd776774cda193565a7779eb501ef99
Author: Michael Gratton <mike vee net>
Date:   Sat Sep 28 22:25:36 2019 +1000

    Fix deadlock when finalising log records generates more logging
    
    Make sure records are finalised outside the lock.
    
    Should fix #579

 src/engine/api/geary-logging.vala | 18 +++++++++++++++++-
 1 file changed, 17 insertions(+), 1 deletion(-)
---
diff --git a/src/engine/api/geary-logging.vala b/src/engine/api/geary-logging.vala
index faee162b..c633910b 100644
--- a/src/engine/api/geary-logging.vala
+++ b/src/engine/api/geary-logging.vala
@@ -369,11 +369,27 @@ public void init() {
 public void clear() {
     record_lock.lock();
 
+    // Keep the old first record around for two reasons. First, so we
+    // don't cause potentially thousands of record to be finalised
+    // under the lock, and hence causing a deadlock with
+    // default_log_writer if the finalisation causes more logging to
+    // be generated. Second, so finalising doesn't cause an extremely
+    // deep stack as the first record is finalised, causing the second
+    // to be finalised, causing the third..., and so on.
+    Record? old_first = first_record;
+
     first_record = null;
     last_record = null;
     log_length = 0;
-    
+
     record_lock.unlock();
+
+    while (old_first != null) {
+        // Per the second reason above, by manually clearing each log
+        // record here in a loop, finalisation of each is an iterative
+        // process, not recursive.
+        old_first = old_first.next;
+    }
 }
 
 /**