accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From ivakegg <...@git.apache.org>
Subject [GitHub] accumulo pull request #134: ACCUMULO-4391 Added synchronization around the a...
Date Thu, 28 Jul 2016 00:31:00 GMT
Github user ivakegg commented on a diff in the pull request:

    https://github.com/apache/accumulo/pull/134#discussion_r72548605
  
    --- Diff: core/src/main/java/org/apache/accumulo/core/file/rfile/bcfile/BoundedRangeFileInputStream.java
---
    @@ -62,12 +62,15 @@ public BoundedRangeFileInputStream(FSDataInputStream in, long offset,
long lengt
     
       @Override
       public int available() throws IOException {
    -    int avail = in.available();
    -    if (pos + avail > end) {
    -      avail = (int) (end - pos);
    -    }
    +    final FSDataInputStream inLocal = in;
    +    synchronized (inLocal) {
    +      int avail = inLocal.available();
    --- End diff --
    
    Sorry Marc, apparently pushing a rebased branch lost your comments.  Here they are reproduced:
    
    Marc:
    We don't follow the paradigm prescribed by the interface. If used in a parallel environment
the block region may all be different. The underlying stream itself will likely be nowhere
near your sought position and therefore availability means nothing to the instance that is
being called; however, if we used the length provided by the Constructor at least we have
some semblance of what availability means for a bounded range input stream. Anyone making
a decision based on availability could have a completely incorrect meaning because seek always
occurs within a read. This will be further compounded due to threading.
    
    Ivan:
        Absolutely correct, as I had eluded to in the ticket.  This is why I tracked where
that method is being invoked.  The only place I found was in the hadoop CompressionInputStream
where the returned value is used in the CompressionInputStream.getPos() call which is never
actually used.  That being said, using any InputStream in multiple threads is inherently unsafe
given its API.
        So I could completely rewrite this mechanism to not reuse the same underlying stream.
 Alternatively I could create a map that keeps track of threads and the position in the stream.
 Do you think that is worth the overhead for a method whose value is not actually used?


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message