[
https://issues.apache.org/jira/browse/HADOOP-9271?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Nauroth updated HADOOP-9271:
----------------------------------
Attachment: HADOOP-9271-branch-trunk-win.1.patch
I'm attaching a patch to remove the Python scripting. I tested the build on
Mac, Windows, and Ubuntu, including native support and bundling Snappy on
Ubuntu.
For the most part, this is just a revert back to the trunk version of each of
these pom.xml files. There are a few differences for Windows compatibility
though:
{code}
- ROOT=`cd ${basedir}/..;pwd`
+ ROOT=`cd ../..;pwd`
{code}
This command was not working on Windows because it would generate a mix of \
and / in the destination path. (i.e. cd C:\foo\bar/..) Switching to a
relative path and using only / as the path delimiter worked cross-platform.
{code}
- run tar czf hadoop-${project.version}.tar.gz
hadoop-${project.version}
+ run tar cf hadoop-${project.version}.tar
hadoop-${project.version}
+ run gzip hadoop-${project.version}.tar
{code}
The 'z' flag for compression causes tar to fork a separate process for gzip.
GnuWin32 tar has a limitation in that fork was never implemented, so this would
fail on Windows with "Cannot fork: Function not implemented". Splitting this
into separate tar and gzip commands works cross-platform.
Another option here would have been to control the pipeline explicitly using a
shell pipeline (tar | gzip), but the "run" helper function used here isn't
compatible with passing a command that has a pipe.
{code}
- if [ $? = 1 ]; then
+ if [ $? != 0 ]; then
{code}
On Windows, failure of the prior command might result in some non-zero exit
code other than 1, so I changed this to check for any non-zero exit code.
{code}
- tar xzf
${basedir}/downloads/apache-tomcat-${tomcat.version}.tar.gz
+ gzip -cd
../../downloads/apache-tomcat-${tomcat.version}.tar.gz | tar xf -
{code}
See explanation above about tar failure to fork on Windows. This one was
solveable using gzip | tar.
{code}
+ BIN_DIR="${BUILD_DIR}/bin"
+ if [ -d $${BIN_DIR} ] ; then
+
TARGET_BIN_DIR="${BUILD_DIR}/${project.artifactId}-${project.version}/bin"
+ mkdir -p $${TARGET_BIN_DIR}
+ cd $${BIN_DIR}
+ $$TAR * | (cd $${TARGET_BIN_DIR}/; $$UNTAR)
+ fi
{code}
This is required to bundle the native Windows components (winutils.exe and
hadoop.dll) from the bin directory.
> revert Python build scripts from branch-trunk-win
> -------------------------------------------------
>
> Key: HADOOP-9271
> URL: https://issues.apache.org/jira/browse/HADOOP-9271
> Project: Hadoop Common
> Issue Type: Bug
> Components: build
> Affects Versions: trunk-win
> Reporter: Chris Nauroth
> Assignee: Chris Nauroth
> Attachments: HADOOP-9271-branch-trunk-win.1.patch
>
>
> On branch-trunk-win, some work was committed to convert custom build
> scripting to Python as a cross-platform solution to support builds on both
> Linux and Windows. The community has since rejected this solution. This
> issue tracks removal of the Python build code from branch-trunk-win.
> This will restore branch-trunk-win to the same state as trunk, which uses
> shell for custom build scripting. This means that developers working on
> branch-trunk-win will once again need sh.exe on the PATH to build the
> distribution. This will be a temporary state. A subsequent jira will add
> build scripts specific to Windows, using a language more common in a Windows
> development environment. The community has approved minimal use of multiple
> platform-specific scripting languages.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
