avro-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Patrick Linehan (JIRA)" <j...@apache.org>
Subject [jira] Commented: (AVRO-697) Install SIGPIPE handler in the Avro tools main method.
Date Thu, 18 Nov 2010 06:18:13 GMT

    [ https://issues.apache.org/jira/browse/AVRO-697?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12933318#action_12933318

Patrick Linehan commented on AVRO-697:

Shoot.  Just noticed that javac in Ant has the "-Werror" flag enabled, so I won't be able
to directly access the signal-related code in sun.misc.  This is actually a completely reasonable
setting for javac, but makes this particular bug a bit trickier.

The workaround I've used for this previously is to use basic reflection and an InvocationHandler
to bypass javac's static checks.  If I coded this up, would it get accepted?

> Install SIGPIPE handler in the Avro tools main method.
> ------------------------------------------------------
>                 Key: AVRO-697
>                 URL: https://issues.apache.org/jira/browse/AVRO-697
>             Project: Avro
>          Issue Type: Improvement
>          Components: java
>            Reporter: Patrick Linehan
> The following command should complete quickly, no matter how many records "myfile.avro"
>   java -jar avro-tools.jar tojson myfile.avro | head
> However, because the tool JAR's main method does not install a SIGPIPE handler, this
command can be quite slow for large files.
> Suggested fix is to install a handler which calls "System.exit(128 + signal.getNumber())"
in response to SIGPIPE.  Typically the signal number ("getNumber") is 13, but no need to hard
code it.
> Potential issues:
> * Tests which invoke Main.main() will necessarily force the signal handler to be installed.
 This may be undesirable, as the entire test JVM will be "infected" with the handler.  Workarounds
exist.  Does not bother me, personally, however.
> * Installing signal handlers generates compile-time warnings, if I'm not mistaken.  Workarounds
exist.  Does not bother me, personally, however.
> If this suggestion sounds reasonable, I'll post a patch.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message