[
https://issues.apache.org/jira/browse/HADOOP-1689?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12521596
]
Doug Cutting commented on HADOOP-1689:
--------------------------------------
These scripts all used to specify /bin/bash, since that's what /bin/sh maps to
for most Hadoop developers, and its thus hard to keep bash-isms from getting
committed. They got switched in HADOOP-352, and I think my concerns about that
patch have been confirmed.
I vote to switch all scripts to specify /bin/bash once more, and document
/bin/bash as a requirement for Hadoop. Note that Owen, in HADOOP-1553,
recently changed the logging code to require that 'bash' is on $PATH, and I did
something similar in HADOOP-1717. So we currently, in 0.14, rely on bash being
installed. On Solaris bash is not normally installed as /bin/bash, but that
seems a small requirement to add.
> .sh scripts do not work on Solaris
> ----------------------------------
>
> Key: HADOOP-1689
> URL: https://issues.apache.org/jira/browse/HADOOP-1689
> Project: Hadoop
> Issue Type: Bug
> Components: scripts
> Affects Versions: 0.13.1
> Environment: SunOS acl301 5.10 Generic_118855-33 i86pc i386 i86pc
> Reporter: David Biesack
> Priority: Minor
>
> the EXPORT commands in the scripts didn't play nicely w/the default shell &
> Hadoop wouldn't start ("... is not an
> identifier"). I changed the first #! line of bin/hadoop-daemon.sh and
> bin/hadoop to specify bash, then everything worked:
> #!/bin/bash
> ksh is probably another option that will work, but we did not try that
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
