I am testing this patch at the moment.
I am running the bz2 compress with the new code, uncompessing
using the command line bunzip and comparing the
files.
The test is over ~100, 000 files, and I have had to
use the computer for other things - so the test
is not yet complete.
<project name="testbz" default="t"
xmlns:ac="antlib:net.sf.antcontrib" >
<import file="src/ant/simple.xml"/>
<property name="space" location="/media/disk/preilly/space"/>
<property name="checkstatus" value="if [ ! $? == 0 ] ; then exit 1; fi"/>
<macrodef name="testbz">
<attribute name="file"/>
<sequential>
<local name="bname"/>
<local name="target"/>
<basename file="@{file}" property="bname"/>
<property name="target" location="${space}/${bname}"/>
<delete quiet="yes" file="${target}"/>
<delete quiet="yes" file="${target}.bz2"/>
<bzip2 src="@{file}" destfile="${target}.bz2"/>
<ac:bash failonerror="true" dir="${space}">
bunzip2 '${bname}.bz2'
${checkstatus}
cmp '${bname}' '@{file}'
${checkstatus}
rm '${bname}'
true
</ac:bash>
</sequential>
</macrodef>
<target name="t">
<ac:for param="file">
<fileset dir="/media/disk" excludes="preilly/**,**/*$*"/>
<ac:sequential>
<testbz file="@{file}"/>
</ac:sequential>
</ac:for>
</target>
</project>
It looks good at the moment.
Peter
On Thu, Mar 26, 2009 at 1:20 PM, Stefan Bodewig <[email protected]> wrote:
> Hi all,
>
> with Ant 1.7.0 we changed the bzip2 code to make it a lot faster and
> reverted the change in 1.7.1 because it was creating corrupt archives.
>
> Meanwhile the Hadoop folks have been using the 1.7.0 code and claim
> they have found and fixed the problem (a single + 1 missing somewhere
> IIUC).
>
> I think we have a unit test that failed with the 1.7.0 code and passes
> with 1.7.1 - should we give the Hadoop fixed code a second chance if
> it passes all our tests?
>
> Stefan
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
