[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gnulib-bug
Subject:    fix nap.h logic
From:       Bruno Haible <bruno () clisp ! org>
Date:       2017-04-24 22:04:25
Message-ID: 2645110.4lLJQ9W3pq () omega
[Download RAW message or body]

With MSVC 14, test-utimens fails:

../../gltests/nap.h:69: assertion 'nanosleep (&delay_spec, 0) == 0' failed

Obviously a negative delay has been passed to nanosleep.
The code in nap.h is wrong: it multiplies the delay by 2, but a signed integer
eventually wraps around and becomes negative.

This fixes it.


2017-04-23  Bruno Haible  <bruno@clisp.org>

	nap.h: Fix logic.
	* tests/nap.h (nap): Avoid signed integer overflow in loop.

diff --git a/tests/nap.h b/tests/nap.h
index b2726d7..ce8f95a 100644
--- a/tests/nap.h
+++ b/tests/nap.h
@@ -117,9 +117,18 @@ nap (void)
     delay = delay / 2;  /* Try half of the previous delay.  */
   ASSERT (0 < delay);
 
-  for ( ; delay <= 2147483647; delay = delay * 2)
-    if (nap_works (nap_fd, delay, old_st))
-      return;
+  for (;;)
+    {
+      if (nap_works (delay, old_st))
+        return;
+      if (delay <= (2147483647 - 1) / 2)
+        {
+          delay = delay * 2 + 1;
+          continue;
+        }
+      else
+        break;
+    }
 
   /* Bummer: even the highest nap delay didn't work. */
   ASSERT (0);


[prev in list] [next in list] [prev in thread] [next in thread]