-
Bug
-
Resolution: Fixed
-
Major
-
None
-
Windows Server 2003 R2, Cygwin 1.7.2, git 1.7.0.4, Hudson 1.353, GIT plugin 0.8.2
[Problem]
When we look at changes, commit log is garbled.
[Cause]
Git assumes that commit log is written by utf-8, if we don't set "i18n.commitencoding" property.
hudson.plugins.git.GitChangeLogParser class uses java.io.FileReader class at 27.
java.io.FileReader class uses the default character encoding to read file.
Japanese version of Windows uses "MS932" as default character encoding.
So we run hudson on it, commit log is garbled.
[Solution]
use java.io.FileInputStream and java.io.InputStreamReader classes instead of java.io.FileReader class.
java.io.InputStreamReader class has a constructor that is able to set the encoding.
[Example]
BufferedReader rdr = null;
try {
// fetch encoding out of configuration
rdr = new BufferedReader(new InputStreamReader(new FileInputStream(changelogFile), encoding));
/* ... */
} finally {
// close rdr whether constructors throw exception or not
if (rdr != null) rdr.close();
}
- is related to
-
-
- Closed
-
[JENKINS-6203] Git plugin uses default encoding to read change log file
Alexander Radionov
added a comment - Any update? The call to FileUtils.lineIterator() in GitChangeLogParser.java lacks the second parameter(encoding), which seems to cause the fallback to default encoding.
lineIterator = FileUtils.lineIterator(changelogFile);
Since changelog is written in UTF-8 (in GitSCM.java), following change should clear the issue (worked for me).
lineIterator = FileUtils.lineIterator(changelogFile, "UTF-8");
Gennady Evstratov
added a comment - This can be avoided if you set default file encoding to UTF-8 in system properties by adding -Dfile.encoding=UTF-8 to java opts while starting Jenkins or setting JAVA_TOOL_OPTIONS environment variable
Gennady Evstratov
added a comment - This can be avoided if you set default file encoding to UTF-8 in system properties by adding -Dfile.encoding=UTF-8 to java opts while starting Jenkins or setting JAVA_TOOL_OPTIONS environment variable 
Jörgen Lundberg
added a comment - We also have this problem. We're running Jenkins as a Windows server
and I added -Dfile.encoding=UTF-8 to the start-up parameters and restarted
the service. The log messages till shows the UTF-8 characters instead
of for example å ä ö.
Jörgen Lundberg
added a comment - We also have this problem. We're running Jenkins as a Windows server
and I added -Dfile.encoding=UTF-8 to the start-up parameters and restarted
the service. The log messages till shows the UTF-8 characters instead
of for example å ä ö. Jörgen Lundberg
added a comment - It was my bad. I didn't realize the log was written to disk. So, the utf-8 setting works for all builds done after the server restart.
Jörgen Lundberg
added a comment - It was my bad. I didn't realize the log was written to disk. So, the utf-8 setting works for all builds done after the server restart. I'm running Jenkins on Windows 7 (Korean version, encoding is CP949) with msysgit-1.9.5-xxx. The git repository is on remote Linux machine.
I have the issue that recent change message is corrupted.
After some testing and debugging git-plugin, I found, at least in my case, that the log message received from standard output of external git.exe was already corrupted. So, changelog.xml file in builds directory stores corrupted text.
git-client plugin execute external git.exe like this command, "git.exe whatchanged --no-abbrev -M --pretty=raw df1cca6135b7019dbd583693b59f6b97f408f5c5 ", and git.exe out change log to standard out. git-client plugin takes that output. In this point, the wrongfully converted message was received to git-client plugin.
The original message was UTF-8, but this message was assumed to current OS's encoding(CP949 in my computer) and then wrong encoding conversion (CP949 to UCS-2) is occured. I don't yet know which one is doing this wrong conversion; hudson Locallauncher, get.exe or other.
hayarobi Park
added a comment - - edited I'm running Jenkins on Windows 7 (Korean version, encoding is CP949) with msysgit-1.9.5-xxx. The git repository is on remote Linux machine.
I have the issue that recent change message is corrupted.
After some testing and debugging git-plugin, I found, at least in my case, that the log message received from standard output of external git.exe was already corrupted. So, changelog.xml file in builds directory stores corrupted text.
git-client plugin execute external git.exe like this command, "git.exe whatchanged --no-abbrev -M --pretty=raw df1cca6135b7019dbd583693b59f6b97f408f5c5 ", and git.exe out change log to standard out. git-client plugin takes that output. In this point, the wrongfully converted message was received to git-client plugin.
The original message was UTF-8, but this message was assumed to current OS's encoding(CP949 in my computer) and then wrong encoding conversion (CP949 to UCS-2) is occured. I don't yet know which one is doing this wrong conversion; hudson Locallauncher, get.exe or other. I might have a working fix for this issue. Here is the fix: https://github.com/gennady/git-client-plugin/commit/aef7fff3ff765e2f8fd2b270d89e3f6b462cc2de
Give it a try if you don't mind.
You can compile the plugin yourself with
mvn package
or try already compiled version https://github.com/gennady/git-client-plugin/raw/8383bd7c222b52e26b0d1b395b2eb26766f86cf7/compiled-plugin/git-client.hpi
How to try:
- stop jenkins
- remove git-client, git-client.hpi, git-client.jpi from the plugins folder
- copy git-client.hpi to the plugins folder
- start jenkins
Code changed in jenkins
User: Gennady Trafimenkov
Path:
src/main/java/org/jenkinsci/plugins/gitclient/CliGitAPIImpl.java
src/test/java/org/jenkinsci/plugins/gitclient/GitAPITestCase.java
src/test/resources/unicodeCharsInChangelogRepo.zip
src/test/resources/unicodeCharsInChangelogRepoCreate.sh
http://jenkins-ci.org/commit/git-client-plugin/c99c91fcf497e784204398761be5c10f438d0e55
Log:
Fixed garbled commit messages on Windows
On windows changelog commit messages with unicode characters are
not saved correctly to changelog.xml when CliGitAPI
implementation is in use.
That happens because "git whatchanged" gives byte stream of data.
Commit messages in that stream are encoded in UTF-8. It is
necessary to explicitly decode bytestream to strings using UTF-8
encoding, otherwise default system encoding will be used.
This should fix issues:
https://issues.jenkins-ci.org/browse/JENKINS-6203
https://issues.jenkins-ci.org/browse/JENKINS-14798
https://issues.jenkins-ci.org/browse/JENKINS-23091
The same: http://i.imgur.com/63mSQKC.jpg