more insight re. ob_start bug

diff --git a/TODO_SEGFAULTS b/TODO_SEGFAULTS
index 2afa90ad54ed0aa9ca74bb8593e36fdee06d73a7..f4b1026c29c7b21f1e7c5d93d37971c1711a0a4c 100644 (file)
--- a/TODO_SEGFAULTS
+++ b/TODO_SEGFAULTS
@@ -40,10 +40,13 @@ Open:
 
        format was zero.  The zval appeared to be fubar'ed.
 
-    The current do_crash script does not cause a segfault.  An older
-    version does however:
+    When one comments out the (unused) line
 
-           <URL:http://schumann.cx/do_crash_ob_start.txt>
+        $maxa = array_fill(100,1000,&$maxn);
+
+    in the current test script, the segfault can be reproduced (it probably
+    influences the heap layout, so that ob_start does not scribble over 
+    important memory).
 
 
 Methodology