Gracefully cope with coarse system clock

On hppa, we sometimes get a DivisionByZero error - presumably because
the system clock doesn't offer enough granularity.

Gbp-Pq: Name 0005-Gracefully-cope-with-coarse-system-clock.patch

1 files changed, 5 insertions, 2 deletions

diff --git a/src/s3ql/fs.py b/src/s3ql/fs.py
index 33c87fd..ce61180 100644
--- a/src/s3ql/fs.py
+++ b/src/s3ql/fs.py
@@ -409,8 +409,11 @@ class Operations(llfuse.Operations):
                 queue.append(id_p)
 
             dt = time.time() - stamp
-            batch_size = int(batch_size * GIL_RELEASE_INTERVAL / dt)
-            batch_size = min(batch_size, 200) # somewhat arbitrary...
+            if dt == 0:
+                batch_size *= 2
+            else:
+                batch_size = int(batch_size * GIL_RELEASE_INTERVAL / dt)
+                batch_size = min(batch_size, 200) # somewhat arbitrary...
             batch_size = max(batch_size, 20000)
             log.debug('Adjusting batch_size to %d and yielding', batch_size)
             llfuse.lock.yield_(100)