[
https://issues.apache.org/jira/browse/HDFS-9446?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15124454#comment-15124454
]
Colin Patrick McCabe commented on HDFS-9446:
--------------------------------------------
The new hdfsStreamBuilder API added in HDFS-9541 allows users to create files
with a block size bigger than 2 GB.
For the other places where tSize is used, a 4-byte number is appropriate. It
doesn't make sense to read/write more than 2 GB in a single call to
{{hdfsRead}} or {{hdfsWrite}} (this is why the Java code uses {{int}} for those
values).
Meanwhile, the size of files returned by {{hdfsListDirectory}} uses {{tOffset}}
rather than {{tSize}}. Unlike {{tSize}}, {{tOffset}} is typedef'ed as an
{{int64_t}}
Closing as a duplicate of HDFS-9541. Thanks, all.
> tSize of libhdfs in hadoop-2.7.1 is still int32_t
> -------------------------------------------------
>
> Key: HDFS-9446
> URL: https://issues.apache.org/jira/browse/HDFS-9446
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Glen Cao
>
> Issue (https://issues.apache.org/jira/browse/HDFS-466) says what I mentioned
> in the title is fixed. However, I find that in the source
> (hadoop-hdfs-project/hadoop-hdfs/src/main/native/libhdfs/hdfs.h) of
> hadoop-2.7.1, tSize is still typedef-ed as int32_t and I don't find any
> compilation option about that.
> In hdfs.h:
> 75 typedef int32_t tSize; /// size of data for read/write io ops
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
