[ http://issues.apache.org/jira/browse/HADOOP-352?page=comments#action_12419691 ]
Doug Cutting commented on HADOOP-352: ------------------------------------- These scripts were initially declared to use sh, however they didn't work correctly under all versions of sh. The problem is that most developers of these scripts develop on Linux, where sh and bash are the same program. Thus it is *very* easy for bash-specific features to creep into these scripts. So, if we switch them back to sh but don't have folks who are daily testing the trunk on a non-bash-based sh, then we are likely to deliver releases where these fail in strange ways. Right now they fail in a predictable way if you don't have bash installed, so there's no confusion. So I am reluctant to make this change unless I am convinced there are several developers who track trunk changes using a non-bash-based sh. How hard is it to install bash on FreeBSD? > Portability of hadoop shell scripts for deployment > -------------------------------------------------- > > Key: HADOOP-352 > URL: http://issues.apache.org/jira/browse/HADOOP-352 > Project: Hadoop > Type: Bug > Versions: 0.4.0 > Reporter: Jean-Baptiste Quenot > Attachments: hadoop-sh-patch > > Hadoop shell scripts are based on /bin/bash, which is a "standard" shell only > on GNU/Linux. On other Unix systems like FreeBSD however the "standard" > shell is /bin/sh. The attached patch addresses these compatiblity issues. > Note that Solaris support is not yet tested. > Also, the best way to set the HADOOP_HOME variable upon ssh connection in a > portable way is to set it in .ssh/environment. The bash startup script > ".bashrc" is not an option on systems where "bash" is not installed. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
