Re: Jenkins vs. EBCDIC

2014-12-24 Thread gilpeleg
Hi,
I think I found a workaround to this issue, which also relates to 
JENKINS-3629 https://issues.jenkins-ci.org/browse/JENKINS-3629. The 
workaround is based on z/OS UNIX function called Enhanced ASCII support.

First, you need to have the slave successfully connected and online on z/OS 
UNIX using the above mentioned java options -Dfile.enconding=ISO8859-1 and 
-text.

Then, I noticed that when you use the Execute Shell build step, Jenkins 
creates a shell script in my /tmp and tries to execute it using /bin/sh 
-xe... but this shell script is in ASCII so the text is garbled and the 
script doesn't execute properly.
What I did was to reroute Jenkins' /tmp directory to my own specified 
location using java option -Djava.io.tmpdir=/var/jenkins/tmp in the slave 
configuration.

Then I allocated and mounted a new ZFS data set at /var/jenkins/tmp but 
defined it to automatically tag all files in the file system as ASCII. An 
example of a mount command with automatic tagging is:
MOUNT FILESYSTEM('JENKINS.TMP.ZFS')
  MOUNTPOINT('/var/jenkins/tmp')
  TYPE(ZFS) MODE(RDWR) TAG(TEXT,819)

Finally, prior to the Execute Shell build step, I use the Inject 
environment variables build step to inject _BPXK_AUTOCVT=ON before the 
shell script is executed. This environment variable tells z/OS to 
automatically convert the ASCII-tagged files from /var/jenkins/tmp to 
EBCDIC.

The shell script then executes successfully and output is displayed in the 
console log.

Gil.

On Friday, May 2, 2014 10:06:34 PM UTC+3, wwwild...@gmail.com wrote:

 Follow-up:

 I've gotten Jenkins working fine on z/OS, pushing builds from Windows, 
 running the tests on z/OS, and publishing the results to tomcat running on 
 z/OS.

 There were three key things for me to get working/understand that may be 
 helpful to others.  Some or all of these may have already been mentioned by 
 others here or elsewhere:

 1. ssh on z/OS - the key here for me was these changes to the sshd_config 
 config file in /etc/sshd:
 StrictModes no
 But, also, I had to do a number of things around setting folder and file 
 ownership and permissions (see the z/OS redbook on ssh, very helpful) and 
 work to synch up ssh on my cygwin/Windows with ssh on z/OS (I assume all of 
 this is standard ssh) - this was all to get password-less login to work.

 2. When I setup z/OS slave(s) in Jenkins, using ssh to invoke the slave 
 jar, the script I have it invoke to run the jar has this command in it:  
 java -Dfile.encoding=ISO8859_1 -jar /usr/local/jenkins/slave.jar -text
 This seemed to resolve any ASCII/EBCDIC issues I was seeing.

 3. To run the Jenkins jobs I had to explicitly set the jobs to run on 
 Windows, using:
 Restrict where this project can be run
 setting the Label Expression to master.
 Then to execute the job on z/OS I used build Execute Windows batch 
 command with an ssh command to invoke the desired script on z/OS; e.g.:
 ssh USER@HOST su -s RACF_SURROGATE_USER -c SCRIPT_ON_ZOS_TO_RUN

 And voila, that's it!

 There are a lot of details underneath this and I had certain machinations 
 I had to go through for local requirements (e.g. RACF surrogate to allow me 
 to rebuild a Websphere profile file system).

 www

 On Monday, December 30, 2013 8:41:40 AM UTC-8, wwwild...@gmail.com wrote:


 I see a similar issue; I can't get as far as getting the commands from 
 Jenkins to run on z/OS.  That is I see errors like this:
 [zOS unit tests] $ /bin/sh -xe /tmp/hudson4437459291454251752.sh 
 + Ž ÍËÊ %?Ä/% ÂÑ 
 + 1  ¢Á 1 ,Ñ 1 Ë^Í 1 ÑÈÈÁËÈË ËÇ 
 Ž ÍËÊ %?Ä/% ÂÑ: /tmp/hudson4437459291454251752.sh 1: FSUM7351 not found 

 The slave looks to be running fine on z/OS:  java 
 -Dfile.encoding=ISO8859_1 -jar /usr/local/jenkins/slave.jar -text
 and I can get System Information fine for the node.

 I've gone through this thread, but I don't see any correspondence to the 
 classes in remoting-2.25-SNAPSHOT.jar and my 1.535 version of Jenkins.

 So, obviously I have some kind of EBCDIC/ASCII mismatch between my 
 Windows Master and z/OS.
 As far as I know z/OS UNIX shell requires script files to be in EBCDIC, 
 otherwise it's reasonably easy to deal with ASCII and EBCDIC from their 
 (e.g. via Java), but EBCDIC has to be a starting point.

 So, if anyone knows what I'm missing please let me know.

 www



-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jenkinsci-users/83b5579c-f874-47c9-8264-a43133e571a2%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


Re: Jenkins vs. EBCDIC

2014-05-02 Thread wwwild . walsh
Follow-up:

I've gotten Jenkins working fine on z/OS, pushing builds from Windows, 
running the tests on z/OS, and publishing the results to tomcat running on 
z/OS.

There were three key things for me to get working/understand that may be 
helpful to others.  Some or all of these may have already been mentioned by 
others here or elsewhere:

1. ssh on z/OS - the key here for me was these changes to the sshd_config 
config file in /etc/sshd:
StrictModes no
But, also, I had to do a number of things around setting folder and file 
ownership and permissions (see the z/OS redbook on ssh, very helpful) and 
work to synch up ssh on my cygwin/Windows with ssh on z/OS (I assume all of 
this is standard ssh) - this was all to get password-less login to work.

2. When I setup z/OS slave(s) in Jenkins, using ssh to invoke the slave 
jar, the script I have it invoke to run the jar has this command in it:  
java -Dfile.encoding=ISO8859_1 -jar /usr/local/jenkins/slave.jar -text
This seemed to resolve any ASCII/EBCDIC issues I was seeing.

3. To run the Jenkins jobs I had to explicitly set the jobs to run on 
Windows, using:
Restrict where this project can be run
setting the Label Expression to master.
Then to execute the job on z/OS I used build Execute Windows batch 
command with an ssh command to invoke the desired script on z/OS; e.g.:
ssh USER@HOST su -s RACF_SURROGATE_USER -c SCRIPT_ON_ZOS_TO_RUN

And voila, that's it!

There are a lot of details underneath this and I had certain machinations I 
had to go through for local requirements (e.g. RACF surrogate to allow me 
to rebuild a Websphere profile file system).

www

On Monday, December 30, 2013 8:41:40 AM UTC-8, wwwild...@gmail.com wrote:


 I see a similar issue; I can't get as far as getting the commands from 
 Jenkins to run on z/OS.  That is I see errors like this:
 [zOS unit tests] $ /bin/sh -xe /tmp/hudson4437459291454251752.sh 
 + Ž ÍËÊ %?Ä/% ÂÑ 
 + 1  ¢Á 1 ,Ñ 1 Ë^Í 1 ÑÈÈÁËÈË ËÇ 
 Ž ÍËÊ %?Ä/% ÂÑ: /tmp/hudson4437459291454251752.sh 1: FSUM7351 not found 

 The slave looks to be running fine on z/OS:  java 
 -Dfile.encoding=ISO8859_1 -jar /usr/local/jenkins/slave.jar -text
 and I can get System Information fine for the node.

 I've gone through this thread, but I don't see any correspondence to the 
 classes in remoting-2.25-SNAPSHOT.jar and my 1.535 version of Jenkins.

 So, obviously I have some kind of EBCDIC/ASCII mismatch between my Windows 
 Master and z/OS.
 As far as I know z/OS UNIX shell requires script files to be in EBCDIC, 
 otherwise it's reasonably easy to deal with ASCII and EBCDIC from their 
 (e.g. via Java), but EBCDIC has to be a starting point.

 So, if anyone knows what I'm missing please let me know.

 www



-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


Re: Jenkins vs. EBCDIC

2013-12-30 Thread wwwild . walsh

I see a similar issue; I can't get as far as getting the commands from 
Jenkins to run on z/OS.  That is I see errors like this:
[zOS unit tests] $ /bin/sh -xe /tmp/hudson4437459291454251752.sh 
+ Ž ÍËÊ %?Ä/% ÂÑ 
+ 1  ¢Á 1 ,Ñ 1 Ë^Í 1 ÑÈÈÁËÈË ËÇ 
Ž ÍËÊ %?Ä/% ÂÑ: /tmp/hudson4437459291454251752.sh 1: FSUM7351 not found 

The slave looks to be running fine on z/OS:  java -Dfile.encoding=ISO8859_1 
-jar /usr/local/jenkins/slave.jar -text
and I can get System Information fine for the node.

I've gone through this thread, but I don't see any correspondence to the 
classes in remoting-2.25-SNAPSHOT.jar and my 1.535 version of Jenkins.

So, obviously I have some kind of EBCDIC/ASCII mismatch between my Windows 
Master and z/OS.
As far as I know z/OS UNIX shell requires script files to be in EBCDIC, 
otherwise it's reasonably easy to deal with ASCII and EBCDIC from their 
(e.g. via Java), but EBCDIC has to be a starting point.

So, if anyone knows what I'm missing please let me know.

www

-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


Re: Jenkins vs. EBCDIC

2013-11-22 Thread Dorcus Malorcus
Any updates on this?  I'm in the same boat.  The slave.jar connects just 
fine with no special encoding, but any artifacts created on z/OS USS side 
come back to the Jenkins server as EBCDIC and so does the log file found on 
the server (even though it won't display on the web page). 

Thanks! 
Dave (dorcus)




-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


Re: Jenkins vs. EBCDIC

2013-11-22 Thread Kevin Fleming (BLOOMBERG/ 731 LEXIN)
It doesn't seem practical for build artifacts to have their encodings changed 
automatically; doing so would require that the Jenkins slave code somehow be 
told which files are safe to convert and which are not. The log is a different 
situation, of course; that would be safe to convert (and to do so on the fly).

- Original Message -
From: jenkinsci-users@googlegroups.com
To: jenkinsci-users@googlegroups.com
At: Nov 22 2013 16:50:04

Any updates on this?  I'm in the same boat.  The slave.jar connects just fine 
with no special encoding, but any artifacts created on z/OS USS side come back 
to the Jenkins server as EBCDIC and so does the log file found on the server 
(even though it won't display on the web page). 

Thanks! 
Dave (dorcus)


-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


Re: Jenkins vs. EBCDIC

2013-11-22 Thread Dorcus Malorcus
I agree. I just found the iconv utility.  I run this as the last steps in 
my shell script on any artifacts I want brought back to Jenkins.  This 
worked beautifully! 

Using Jenkins with z/OS opens up a lot of possibilities! 






-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


Re: Jenkins vs. EBCDIC

2013-06-24 Thread Dirk Haun
Kohsuke,

just checking: I sent you the build.xml and a logfile last week. Did you
have a chance to look into it yet?

Anything else I could do? Open a bug report, maybe?

Thanks.




On Mon, Jun 17, 2013 at 9:47 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 It should be able to, actually. Any chance you can send me that log file
 and build.xml?

 Also, give us the whole stack trace for that error, not just the error
 message, please.


 2013/6/17 Dirk Haun 68k...@gmail.com

 Thanks for making me check the logfile :)

 Indeed, it's not empty but contains the output in EBCDIC. I guess Jenkins
 simply couldn't display that in the browser?

 From the looks of it, we can now run commands (without having to
 translate them to EBCDIC manually), but we seem to have a problem with SVN
 authentication:

 Checking out svn://our-svn-server/cpmffmeta/trunk
 ERROR: Subversion checkout has been canceled
 hudson.util.IOException2: remote file operation failed:
 /.../jenkins/workspace/zpdtCommandTest at hudson.remoting.Channel@5d0a3a93
 :sva-zpdt-emul
 ...
 Caused by: org.tmatesoft.svn.core.SVNCancelException: svn: No credential
 to try. Authentication failed


 We'll investigate some more ...




 On Sun, Jun 16, 2013 at 2:06 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 Thanks for helping us troubleshoot this.

 If you look at the log file inside $JENKINS_HOME, is it really empty? It
 should contain console log in EBCDIC, then you should see
 charsetCP037/charset (or something like that) in build.xml.


 2013/6/13 Dirk Haun 68k...@gmail.com

 Thanks.

 The good news: We can now connect the slave without having to set the
 file.encoding.

 The bad news: Our jobs don't produce any console output any more.
 Neither a simple set command nor attempting to check out from SVN (which
 worked before) leave anything in the console - it's simply empty.

 Any ideas?



 On Thu, Jun 13, 2013 at 2:20 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 No, it's not. I created
 https://ci.jenkins-ci.org/job/remoting/lastBuild/


 2013/6/13 Dirk Haun 68k...@gmail.com

 Kohsuke,

 thanks for trying to fix it. But I'm afraid there's no difference -
 it still can only connect when we set the file.encoding to ISO8859.

 I took the remoting-1.42-SNAPSHOT.jar from here:
 https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

 Was that even the correct file? But I didn't see any other version
 anywhere.




 On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi 
 k...@kohsuke.orgwrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting
 fixes that. Any chance you can try a new remoting jar to see if you can
 connect via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but
 you wouldn't need that since your master runs in an ASCII-compatible
 encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any 
 case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the
 username and password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML
 prolog (the ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from
 that without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to
 something other than EBCDIC so we can even get a connection to our 
 master.
 

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp-auth 
 username:password
 

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave
 can not get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi 
 k...@kohsuke.org wrote:

 This should work out of the box, because we take the encoding of
 the slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class,
 which uses the FilePath.createTextTempFile method to create a shell 
 script.
 This file takes contents as String, then uses the encoding of the 
 slave
 when it does new FileWriter(f). So it should write the shell script 
 in
 EBCDIC.

 ** **

 Does your slave JVM have a proper default encoding configured

Re: Jenkins vs. EBCDIC

2013-06-17 Thread Dirk Haun
Thanks for making me check the logfile :)

Indeed, it's not empty but contains the output in EBCDIC. I guess Jenkins
simply couldn't display that in the browser?

From the looks of it, we can now run commands (without having to translate
them to EBCDIC manually), but we seem to have a problem with SVN
authentication:

Checking out svn://our-svn-server/cpmffmeta/trunk
ERROR: Subversion checkout has been canceled
hudson.util.IOException2: remote file operation failed:
/.../jenkins/workspace/zpdtCommandTest at hudson.remoting.Channel@5d0a3a93
:sva-zpdt-emul
...
Caused by: org.tmatesoft.svn.core.SVNCancelException: svn: No credential to
try. Authentication failed


We'll investigate some more ...




On Sun, Jun 16, 2013 at 2:06 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 Thanks for helping us troubleshoot this.

 If you look at the log file inside $JENKINS_HOME, is it really empty? It
 should contain console log in EBCDIC, then you should see
 charsetCP037/charset (or something like that) in build.xml.


 2013/6/13 Dirk Haun 68k...@gmail.com

 Thanks.

 The good news: We can now connect the slave without having to set the
 file.encoding.

 The bad news: Our jobs don't produce any console output any more. Neither
 a simple set command nor attempting to check out from SVN (which worked
 before) leave anything in the console - it's simply empty.

 Any ideas?



 On Thu, Jun 13, 2013 at 2:20 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 No, it's not. I created
 https://ci.jenkins-ci.org/job/remoting/lastBuild/


 2013/6/13 Dirk Haun 68k...@gmail.com

 Kohsuke,

 thanks for trying to fix it. But I'm afraid there's no difference - it
 still can only connect when we set the file.encoding to ISO8859.

 I took the remoting-1.42-SNAPSHOT.jar from here:
 https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

 Was that even the correct file? But I didn't see any other version
 anywhere.




 On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting fixes
 that. Any chance you can try a new remoting jar to see if you can connect
 via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but
 you wouldn't need that since your master runs in an ASCII-compatible
 encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any 
 case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the username
 and password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog
 (the ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from
 that without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something
 other than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can
 not get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the
 slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which
 uses the FilePath.createTextTempFile method to create a shell script. 
 This
 file takes contents as String, then uses the encoding of the slave when 
 it
 does new FileWriter(f). So it should write the shell script in EBCDIC.
 

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using
 the Hercules emulator, not real hardware, but still). This OS uses 
 EBCDIC
 as its native character set, not ASCII. We've successfully managed to 
 check
 out from our SVN repository (yay!), but now we would need to run 
 commands
 to build the software. And this is where we run into problems ...***
 *

 The problem is that commands that we type into the text field of the
 Execute shell build

Re: Jenkins vs. EBCDIC

2013-06-17 Thread Kohsuke Kawaguchi
It should be able to, actually. Any chance you can send me that log file
and build.xml?

Also, give us the whole stack trace for that error, not just the error
message, please.


2013/6/17 Dirk Haun 68k...@gmail.com

 Thanks for making me check the logfile :)

 Indeed, it's not empty but contains the output in EBCDIC. I guess Jenkins
 simply couldn't display that in the browser?

 From the looks of it, we can now run commands (without having to translate
 them to EBCDIC manually), but we seem to have a problem with SVN
 authentication:

 Checking out svn://our-svn-server/cpmffmeta/trunk
 ERROR: Subversion checkout has been canceled
 hudson.util.IOException2: remote file operation failed:
 /.../jenkins/workspace/zpdtCommandTest at hudson.remoting.Channel@5d0a3a93
 :sva-zpdt-emul
 ...
 Caused by: org.tmatesoft.svn.core.SVNCancelException: svn: No credential
 to try. Authentication failed


 We'll investigate some more ...




 On Sun, Jun 16, 2013 at 2:06 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 Thanks for helping us troubleshoot this.

 If you look at the log file inside $JENKINS_HOME, is it really empty? It
 should contain console log in EBCDIC, then you should see
 charsetCP037/charset (or something like that) in build.xml.


 2013/6/13 Dirk Haun 68k...@gmail.com

 Thanks.

 The good news: We can now connect the slave without having to set the
 file.encoding.

 The bad news: Our jobs don't produce any console output any more.
 Neither a simple set command nor attempting to check out from SVN (which
 worked before) leave anything in the console - it's simply empty.

 Any ideas?



 On Thu, Jun 13, 2013 at 2:20 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 No, it's not. I created
 https://ci.jenkins-ci.org/job/remoting/lastBuild/


 2013/6/13 Dirk Haun 68k...@gmail.com

 Kohsuke,

 thanks for trying to fix it. But I'm afraid there's no difference - it
 still can only connect when we set the file.encoding to ISO8859.

 I took the remoting-1.42-SNAPSHOT.jar from here:
 https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

 Was that even the correct file? But I didn't see any other version
 anywhere.




 On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi 
 k...@kohsuke.orgwrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting
 fixes that. Any chance you can try a new remoting jar to see if you can
 connect via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but
 you wouldn't need that since your master runs in an ASCII-compatible
 encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any 
 case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the
 username and password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog
 (the ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from
 that without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something
 other than EBCDIC so we can even get a connection to our master.***
 *

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp-auth 
 username:password
 

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave
 can not get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of
 the slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class,
 which uses the FilePath.createTextTempFile method to create a shell 
 script.
 This file takes contents as String, then uses the encoding of the slave
 when it does new FileWriter(f). So it should write the shell script 
 in
 EBCDIC.

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using
 the Hercules emulator, not real hardware, but still). This OS uses 
 EBCDIC
 as its native character set, not ASCII. We've successfully managed to 
 check

Re: Jenkins vs. EBCDIC

2013-06-16 Thread Kohsuke Kawaguchi
Thanks for helping us troubleshoot this.

If you look at the log file inside $JENKINS_HOME, is it really empty? It
should contain console log in EBCDIC, then you should see
charsetCP037/charset (or something like that) in build.xml.


2013/6/13 Dirk Haun 68k...@gmail.com

 Thanks.

 The good news: We can now connect the slave without having to set the
 file.encoding.

 The bad news: Our jobs don't produce any console output any more. Neither
 a simple set command nor attempting to check out from SVN (which worked
 before) leave anything in the console - it's simply empty.

 Any ideas?



 On Thu, Jun 13, 2013 at 2:20 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 No, it's not. I created https://ci.jenkins-ci.org/job/remoting/lastBuild/


 2013/6/13 Dirk Haun 68k...@gmail.com

 Kohsuke,

 thanks for trying to fix it. But I'm afraid there's no difference - it
 still can only connect when we set the file.encoding to ISO8859.

 I took the remoting-1.42-SNAPSHOT.jar from here:
 https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

 Was that even the correct file? But I didn't see any other version
 anywhere.




 On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting fixes
 that. Any chance you can try a new remoting jar to see if you can connect
 via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but you
 wouldn't need that since your master runs in an ASCII-compatible encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any case, 
 I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the username
 and password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog
 (the ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from
 that without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something
 other than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can
 not get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the
 slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which
 uses the FilePath.createTextTempFile method to create a shell script. 
 This
 file takes contents as String, then uses the encoding of the slave when 
 it
 does new FileWriter(f). So it should write the shell script in EBCDIC.
 

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using
 the Hercules emulator, not real hardware, but still). This OS uses EBCDIC
 as its native character set, not ASCII. We've successfully managed to 
 check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC 
 when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into 
 the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII
 equivalents, so this is not an option in practice (besides it being a 
 pain
 in the back to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a
 script that's already on the OS/390 side? Please note that again, there's
 the problem of character set translation - this time with the file name;
 but since the SVN plugin does work as expected, maybe some part of 
 Jenkins

Re: Jenkins vs. EBCDIC

2013-06-13 Thread Dirk Haun
Kohsuke,

thanks for trying to fix it. But I'm afraid there's no difference - it
still can only connect when we set the file.encoding to ISO8859.

I took the remoting-1.42-SNAPSHOT.jar from here:
https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

Was that even the correct file? But I didn't see any other version anywhere.




On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting fixes
 that. Any chance you can try a new remoting jar to see if you can connect
 via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but you
 wouldn't need that since your master runs in an ASCII-compatible encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the username and
 password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog (the
 ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from that
 without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something
 other than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can not
 get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the
 slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which
 uses the FilePath.createTextTempFile method to create a shell script. This
 file takes contents as String, then uses the encoding of the slave when it
 does new FileWriter(f). So it should write the shell script in EBCDIC.
 

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII
 equivalents, so this is not an option in practice (besides it being a pain
 in the back to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

 --
 You received this message because you are subscribed to the Google
 Groups Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send
 an email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  



 

 ** **

 --
 Kohsuke Kawaguchi 

 --
 You received this message because you are subscribed to the Google
 Groups Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send
 an email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  

 ** **

 --
 You

Re: Jenkins vs. EBCDIC

2013-06-13 Thread Kohsuke Kawaguchi
No, it's not. I created https://ci.jenkins-ci.org/job/remoting/lastBuild/


2013/6/13 Dirk Haun 68k...@gmail.com

 Kohsuke,

 thanks for trying to fix it. But I'm afraid there's no difference - it
 still can only connect when we set the file.encoding to ISO8859.

 I took the remoting-1.42-SNAPSHOT.jar from here:
 https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

 Was that even the correct file? But I didn't see any other version
 anywhere.




 On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting fixes
 that. Any chance you can try a new remoting jar to see if you can connect
 via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but you
 wouldn't need that since your master runs in an ASCII-compatible encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the username
 and password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog
 (the ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from that
 without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something
 other than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can
 not get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the
 slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which
 uses the FilePath.createTextTempFile method to create a shell script. This
 file takes contents as String, then uses the encoding of the slave when it
 does new FileWriter(f). So it should write the shell script in EBCDIC.
 

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII
 equivalents, so this is not an option in practice (besides it being a pain
 in the back to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

 --
 You received this message because you are subscribed to the Google
 Groups Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send
 an email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  



 

 ** **

 --
 Kohsuke Kawaguchi 

 --
 You received this message because you are subscribed to the Google
 Groups Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send
 an email to jenkinsci-users

Re: Jenkins vs. EBCDIC

2013-06-13 Thread Dirk Haun
Thanks.

The good news: We can now connect the slave without having to set the
file.encoding.

The bad news: Our jobs don't produce any console output any more. Neither a
simple set command nor attempting to check out from SVN (which worked
before) leave anything in the console - it's simply empty.

Any ideas?



On Thu, Jun 13, 2013 at 2:20 PM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 No, it's not. I created https://ci.jenkins-ci.org/job/remoting/lastBuild/


 2013/6/13 Dirk Haun 68k...@gmail.com

 Kohsuke,

 thanks for trying to fix it. But I'm afraid there's no difference - it
 still can only connect when we set the file.encoding to ISO8859.

 I took the remoting-1.42-SNAPSHOT.jar from here:
 https://ci.jenkins-ci.org/view/All/job/jenkins_main_trunk/2599/

 Was that even the correct file? But I didn't see any other version
 anywhere.




 On Wed, Jun 12, 2013 at 5:14 PM, Kohsuke Kawaguchi k...@kohsuke.orgwrote:

 I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting fixes
 that. Any chance you can try a new remoting jar to see if you can connect
 via JNLP without overriding the encoding?

  There's a corresponding change in the server for a proper fix, but you
 wouldn't need that since your master runs in an ASCII-compatible encoding.


 2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the username
 and password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog
 (the ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from that
 without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something
 other than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can
 not get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the
 slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which
 uses the FilePath.createTextTempFile method to create a shell script. This
 file takes contents as String, then uses the encoding of the slave when it
 does new FileWriter(f). So it should write the shell script in EBCDIC.
 

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into 
 the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII
 equivalents, so this is not an option in practice (besides it being a pain
 in the back to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

 --
 You received this message because you are subscribed to the Google
 Groups Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from

Re: Jenkins vs. EBCDIC

2013-06-12 Thread Kohsuke Kawaguchi
This should work out of the box, because we take the encoding of the slave
into account for obvious reasons.

the Execute Shell feature uses the CommandInterpreter class, which uses
the FilePath.createTextTempFile method to create a shell script. This file
takes contents as String, then uses the encoding of the slave when it does
new FileWriter(f). So it should write the shell script in EBCDIC.

Does your slave JVM have a proper default encoding configured?



2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII equivalents,
 so this is not an option in practice (besides it being a pain in the back
 to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

  --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.






-- 
Kohsuke Kawaguchi

-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.




Re: Jenkins vs. EBCDIC

2013-06-12 Thread Dirk Haun
Ah, the default encoding. It seems we *have* to set it to something other
than EBCDIC so we can even get a connection to our master.

The slave is connected via Java Web Start and started like so:

java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp
-auth username:password

Without the -Dfile.encoding (or setting it to EBCDIC), the slave can not
get a connection to the master.

Connecting via ssh is not an option, as it's not available on the emulator.

Any suggestions?



On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org wrote:

 This should work out of the box, because we take the encoding of the slave
 into account for obvious reasons.

 the Execute Shell feature uses the CommandInterpreter class, which uses
 the FilePath.createTextTempFile method to create a shell script. This file
 takes contents as String, then uses the encoding of the slave when it does
 new FileWriter(f). So it should write the shell script in EBCDIC.

 Does your slave JVM have a proper default encoding configured?



 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII equivalents,
 so this is not an option in practice (besides it being a pain in the back
 to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

  --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.






 --
 Kohsuke Kawaguchi

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.




-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.




RE: Jenkins vs. EBCDIC

2013-06-12 Thread James Nord (jnord)
Looks like jenkins serves up its JNLP pages without an XML prolog (the ?xml 
version=1.0 encoding=UTF-8? bit)...

If you save the file and add an XML prolog for UTF-8 and run from that without 
setting the encoding - does that work?

/James

From: jenkinsci-users@googlegroups.com 
[mailto:jenkinsci-users@googlegroups.com] On Behalf Of Dirk Haun
Sent: 12 June 2013 15:03
To: jenkinsci-users
Subject: Re: Jenkins vs. EBCDIC

Ah, the default encoding. It seems we *have* to set it to something other than 
EBCDIC so we can even get a connection to our master.
The slave is connected via Java Web Start and started like so:

java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl 
http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlphttp://%3cip-of-master%3e/computer/sva-zpdt-emul/slave-agent.jnlp
 -auth username:password
Without the -Dfile.encoding (or setting it to EBCDIC), the slave can not get a 
connection to the master.

Connecting via ssh is not an option, as it's not available on the emulator.
Any suggestions?

On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi 
k...@kohsuke.orgmailto:k...@kohsuke.org wrote:
This should work out of the box, because we take the encoding of the slave into 
account for obvious reasons.

the Execute Shell feature uses the CommandInterpreter class, which uses the 
FilePath.createTextTempFile method to create a shell script. This file takes 
contents as String, then uses the encoding of the slave when it does new 
FileWriter(f). So it should write the shell script in EBCDIC.

Does your slave JVM have a proper default encoding configured?


2013/6/12 Dirk Haun 68k...@gmail.commailto:68k...@gmail.com
We're trying to use Jenkins to build our software on OS/390 (using the Hercules 
emulator, not real hardware, but still). This OS uses EBCDIC as its native 
character set, not ASCII. We've successfully managed to check out from our SVN 
repository (yay!), but now we would need to run commands to build the software. 
And this is where we run into problems ...
The problem is that commands that we type into the text field of the Execute 
shell build step need to be translated from ASCII to EBCDIC when run on the 
OS/390 side. We confirmed that this is the problem by manually translating a 
simple command to EBCDIC and typing those characters into the text input field 
- the command was then executed.
However, a lot of EBCDIC characters have non-printable ASCII equivalents, so 
this is not an option in practice (besides it being a pain in the back to 
maintain).
Is there any any plugin that can handle this situation?
Alternatively, is there a plugin that would simply let us run a script that's 
already on the OS/390 side? Please note that again, there's the problem of 
character set translation - this time with the file name; but since the SVN 
plugin does work as expected, maybe some part of Jenkins is already doing the 
right thing.
We're stuck at this point, so any help is appreciated. Thank you.
bye, Dirk
--
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to 
jenkinsci-users+unsubscr...@googlegroups.commailto:jenkinsci-users%2bunsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.





--
Kohsuke Kawaguchi
--
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to 
jenkinsci-users+unsubscr...@googlegroups.commailto:jenkinsci-users%2bunsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.



--
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to 
jenkinsci-users+unsubscr...@googlegroups.commailto:jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.




Re: Jenkins vs. EBCDIC

2013-06-12 Thread Kohsuke Kawaguchi
XML without prolog is parsed as UTF-8, so I think that part is OK.

I'm spotting a few places in JNLP handshaking that relies on
ASCII-compatible encoding, so I wonder if that's causing it. In any case, I
think we need more detailed error messages.

Another thing I spotted is the BASIC auth header. How is the username and
password supposed to be encoded?


2013/6/12 James Nord (jnord) jn...@cisco.com

  Looks like jenkins serves up its JNLP pages without an XML prolog (the
 ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from that
 without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something other
 than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can not
 get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the emulator.
 

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the slave
 into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which uses
 the FilePath.createTextTempFile method to create a shell script. This file
 takes contents as String, then uses the encoding of the slave when it does
 new FileWriter(f). So it should write the shell script in EBCDIC.

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII equivalents,
 so this is not an option in practice (besides it being a pain in the back
 to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  



 

 ** **

 --
 Kohsuke Kawaguchi 

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  

 ** **

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.






-- 
Kohsuke Kawaguchi

-- 
You received this message because you are subscribed to the Google Groups 
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to jenkinsci

Re: Jenkins vs. EBCDIC

2013-06-12 Thread Kohsuke Kawaguchi
I claim f8916a839c5a7bfe6d0de100a33cf23de33de6e4 in the remoting fixes
that. Any chance you can try a new remoting jar to see if you can connect
via JNLP without overriding the encoding?

 There's a corresponding change in the server for a proper fix, but you
wouldn't need that since your master runs in an ASCII-compatible encoding.


2013/6/12 Kohsuke Kawaguchi k...@kohsuke.org

 XML without prolog is parsed as UTF-8, so I think that part is OK.

 I'm spotting a few places in JNLP handshaking that relies on
 ASCII-compatible encoding, so I wonder if that's causing it. In any case, I
 think we need more detailed error messages.

 Another thing I spotted is the BASIC auth header. How is the username and
 password supposed to be encoded?


 2013/6/12 James Nord (jnord) jn...@cisco.com

   Looks like jenkins serves up its JNLP pages without an XML prolog (the
 ?xml version=1.0 encoding=UTF-8? bit)…

 ** **

 If you save the file and add an XML prolog for UTF-8 and run from that
 without setting the encoding – does that work?

 ** **

 /James

 ** **

 *From:* jenkinsci-users@googlegroups.com [mailto:
 jenkinsci-users@googlegroups.com] *On Behalf Of *Dirk Haun
 *Sent:* 12 June 2013 15:03
 *To:* jenkinsci-users
 *Subject:* Re: Jenkins vs. EBCDIC

 ** **

 Ah, the default encoding. It seems we *have* to set it to something other
 than EBCDIC so we can even get a connection to our master.

 The slave is connected via Java Web Start and started like so:

 java -Dfile.encoding=ISO8859_1 -jar slave.jar -text -jnlpUrl
 http://ip-of-master/computer/sva-zpdt-emul/slave-agent.jnlp -auth
 username:password

 Without the -Dfile.encoding (or setting it to EBCDIC), the slave can not
 get a connection to the master.

 ** **

 Connecting via ssh is not an option, as it's not available on the
 emulator.

 Any suggestions?

 ** **

 On Wed, Jun 12, 2013 at 11:14 AM, Kohsuke Kawaguchi k...@kohsuke.org
 wrote:

 This should work out of the box, because we take the encoding of the
 slave into account for obvious reasons.

 ** **

 the Execute Shell feature uses the CommandInterpreter class, which uses
 the FilePath.createTextTempFile method to create a shell script. This file
 takes contents as String, then uses the encoding of the slave when it does
 new FileWriter(f). So it should write the shell script in EBCDIC.

 ** **

 Does your slave JVM have a proper default encoding configured?

 ** **

 ** **

 2013/6/12 Dirk Haun 68k...@gmail.com

 We're trying to use Jenkins to build our software on OS/390 (using the
 Hercules emulator, not real hardware, but still). This OS uses EBCDIC as
 its native character set, not ASCII. We've successfully managed to check
 out from our SVN repository (yay!), but now we would need to run commands
 to build the software. And this is where we run into problems ...

 The problem is that commands that we type into the text field of the
 Execute shell build step need to be translated from ASCII to EBCDIC when
 run on the OS/390 side. We confirmed that this is the problem by manually
 translating a simple command to EBCDIC and typing those characters into the
 text input field - the command was then executed.

 However, a lot of EBCDIC characters have non-printable ASCII equivalents,
 so this is not an option in practice (besides it being a pain in the back
 to maintain).

 Is there any any plugin that can handle this situation?

 Alternatively, is there a plugin that would simply let us run a script
 that's already on the OS/390 side? Please note that again, there's the
 problem of character set translation - this time with the file name; but
 since the SVN plugin does work as expected, maybe some part of Jenkins is
 already doing the right thing.

 We're stuck at this point, so any help is appreciated. Thank you.

 bye, Dirk

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  



 

 ** **

 --
 Kohsuke Kawaguchi 

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  

 ** **

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group and stop receiving emails from it, send an
 email to jenkinsci-users+unsubscr...@googlegroups.com.
 For more options, visit https://groups.google.com/groups/opt_out.

  

 --
 You received this message because you are subscribed to the Google Groups
 Jenkins Users group.
 To unsubscribe from this group