Re: CCSID of objects in IFS

Scott,

This gets a little more interesting.

I started on a system with no prior installation of this particular tarball. I initially ran the tar xvf mytarball.tar command and then looked at the CCSID's of the extracted files. It was 37 on this particular system (V5R4M0). I attempted to run the shell script and got an error, which I had gotten before, that indicated it couldn't find a class file that was needed. THAT was due to some "bad" characters in the script. using EDTF, you could see some "funky" characters in the file because they were just highlighted, reverse image, blocks in the file. Removing those characters allowed the script to run, but I ran into other errors. Replacing several files in this installation from another successful installation eventually fixed the problems.

So, it *seems* that the extracted tarball contains data that cannot be properly "converted" when run inside of QShell. Not sure why. So, I deleted the whole thing and started over. This time I started QSH and then issued the command:

export QIBM_CCSID=819

before I extracted the tarball. Again all the CCSID's extracted as 37 on all the files and again I had the same errors due to the same reason as before: The files, when viewed with EDTF would show "odd" characters that needed to be removed.

Once again I deleted the folder and files. Curiously, I took a different approach and instead mapped a network drive to the IFS and then used a zipped version of the tarball (I could choose a .tar.gz version or a .zip version and this time I downloaded the .zip version) I unzipped the file to a folder and, after doing so, I started QShell and ran the shell scripts and they ran perfectly. No errors or problems. The CCSID is 1252.

As for my 270 that has 819 as the CCSID. I think you are correct. I must have extracted these in a PASE session although I don't recall doing so. But, then again, I have attempted the installation a few times and may have lost track of what I had done.

At this point I am going to go with the option to map a network drive and then extract the zipped version of the file to the IFS rather than going into QSH and using tar to extract the tarball. At least this part of the mystery has been solved.

At the end of the day I may just use SAV to save the installation and then have the user use the RST command at the other end to deploy it. That should eliminate the extract issues.

Pete

Pete Helgren wrote:

Thanks again Scott.
I delayed my response in hopes of doing some testing with setting CCSID's and then extracting the files but I haven't had a chance because I am traveling.

If I get all of it sorted out, I think WILL just use the sav and rst commands and provide a zipped save file for deployment in the future. First though, I need to sort out some of the issues I am seeing and determine what is causing them. I'll have access to a "clean" box this afternoon and I am going to walk through each step. That should clarify the issues for me.

If I run into any unanswerable questions though, or finally sort this out, I'll be sure to post back here.

Pete


Scott Klement wrote:

Hi Pete,

What I am trying to test is a deployment of these files in both a
V5R3M0 environment and in a V5R4M0 environment. I am trying to do
exactly the same steps on the two platforms so that if there are any
differences I'll document them.

If you do the following from QShell, I would expect them to work the same:

export QIBM_CCSID=819 (or 37, or whatever your data is)
tar xvf mytarball.tar

This will force the CCSID of the extracted files to be 819 on both systems.

Having said that, I don't think the JDBC error you reported has anything whatsoever to do with the CCSID of the IFS files you're extracting. Instead, it's more likely that the job has a 65535 CCSID like Marty suggested.

Though, if this is a System i deployment (as opposed to extracting a Unix archive you downloaded from someone else) I wonder why you don't use the SAV and RST CL commands? That would preserve all of the file information (including the CCSID) instead of just the ones that make sense under Unix (like TAR).

But, again, I don't think this'll solve the JDBC error...