•  29-05-2022
  •  | 
  •  

Question

I am facing a problem with svnant. <svnModified /> takes over 3 minutes to complete on my working copy (svn metadata version 1.7: using the latest svnant from HEAD). If I try to check for modifications using TortoiseSVN 1.7.12 it takes under 3 seconds. If I do a svn status from the command prompt it also takes under 3 seconds.

My test build file looks like this:

<project name="tools" default="default" basedir=".">

  <path id="paths.classpath">
    <fileset dir="." caseSensitive="false">
      <include name="svnant/**/*.jar" />
    </fileset>
  </path>

  <typedef resource="org/tigris/subversion/svnant/svnantlib.xml" classpathref="paths.classpath" />

  <fileset id="filesets.svn.modified" dir="C:/Workspace/projects">
    <svnModified />
  </fileset>

  <target name="default">
    <pathconvert pathsep="&#xA;" property="target.output" refid="filesets.svn.modified" />
    <echo>${target.output}</echo>
  </target>

</project>

Am I doing something wrong? I tried upgrading svnkit to the latest version but it made no difference. Am I including too many or too few jars? Other details: Ant 1.9.0, Windows 7 x64, Java 1.7.0_21 x64

My libs:

+---lib
    antlr-runtime-3.4.jar
    jna-3.4.0.jar
    LICENSE-ANTLR.txt
    LICENSE-JAVAHL.txt
    LICENSE-JNA.txt
    LICENSE-SEQUENCE.txt
    LICENSE-SQLJET.txt
    LICENSE-TRILEAD.txt
    sequence-library-1.0.2.jar
    sqljet-1.1.7.jar
    svn-javahl.jar
    svnant.jar
    svnClientAdapter.jar
    svnkit-1.7.9.jar
    svnkit-cli-1.7.9.jar
    svnkit-javahl16-1.7.9.jar
    trilead-ssh2-1.0.0-build216.jar
Était-ce utile?

La solution

I don't know for sure, but my guess is that when using svnModified, svn is being queried for each file under C:/Workspace/projects. But when using svn status with the command line or with TortoiseSVN, everything happens within the svn software. In the first case there is probably a lot of little queries about one file, in the second case there is only one query about a whole file tree.

To improve the performance, the only way I can think of is:

  1. calling a svn status (via the exec or svnant tasks) and get its output

  2. filter the output with regexp to keep the interesting lines about modified files

  3. make it a list of file names and use resourcelist to build a resource collection from that flat list of names

Some links you might be interested in to implement it:

Licencié sous: CC-BY-SA avec attribution
Non affilié à StackOverflow
scroll top