[
https://issues.apache.org/jira/browse/CRYPTO-135?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Marcelo Vanzin resolved CRYPTO-135.
-----------------------------------
Resolution: Won't Fix
See discussion in the PR for this:
https://github.com/apache/commons-crypto/pull/74
Short summary: it's very hard to provide non-blocking semantics and still
return reasonable values to the callers of the WritableByteChannel API. So
instead I just commented on the class javadocs that they should only be used
with blocking I/O.
> CryptoOutputStream is always blocking
> -------------------------------------
>
> Key: CRYPTO-135
> URL: https://issues.apache.org/jira/browse/CRYPTO-135
> Project: Commons Crypto
> Issue Type: Bug
> Components: Stream
> Affects Versions: 1.1.0
> Reporter: Marcelo Vanzin
> Priority: Major
>
> CRYPTO-125 described a bug where not all data written to the
> CryptoOutputStream would make it to the underlying channel if that channel
> was non-blocking.
> The fix added in that bug makes CryptoOutputStream blocking by doing a busy
> loop. This makes it a poor match (not to say unusable) for applications that
> use non-blocking channels, since it will turn non-blocking streams into
> blocking streams that waste a lot of CPU.
> Instead, CryptoOutputStream should correctly implement non-blocking semantics
> in its implementation of WritableByteChannel.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
