PVCS - Source repository File Server

In PVCS, when we try to open the project database, it typically lists File Servers to select. The file server name listed there is not the actual server's name. Recently, we had to find out what server it was connecting to. This can be done in Admin->File Servers menu option in PVCS. You can add/delete/update server names here.

[caption id="attachment_4952" align="alignnone" width="300"] Where to add PVCS server list[/caption]

So, if the PVCS server is changed in the future and you need to connect to the new server, change it as mentioned above. This needs to be done on each developer machine or the ini file could be distributed. (The name of the ini file is, C:\Program Files\Merant\vm\common\bin\win32\servers.ini)

[caption id="attachment_4951" align="alignnone" width="300"] Fig 2. server list is stored in servers.ini[/caption]

If you connect to PVCS from Powerbuilder (or any other MSSCCI compliant IDEs), you do not have to change anything there. PB actually reads the PVCS servers.ini file to get the PVCS Server name(s). After making the changes to File Server name in PVCS, we just need to disconnect and reconnect to merant version manager in Powerbuilder.

SQLPLUS - Spooling in HTML format

I used this recently and I thought of sharing this. If you don't know this already, SQL*PLUS has built in support for HTML. All you have to do is add markup option either on command line or inside the script. By the nature of it, it is only useful when you are spooling the output. Useful when you are dumping from multiple tables or generating simple reports. Just be aware that the output file will be larger than the usual text file spool.

Just add below to the script before spooling and the

SET MARKUP HTML ON SPOOL ON PREFORMAT OFF ENTMAP ON -
HEAD "<TITLE>E-App Validation Report</TITLE> -
<STYLE type='text/css'> -
<!-- BODY {background: #FFFFFF} --> -
</STYLE>" -
BODY "TEXT='#000FF'" -
TABLE "WIDTH='90%' BORDER='5'"

spool cleanup_eapp_data.html

SQLPlus - spool name with embedded timestamp

To be able to create a spool file, with filename embedded with timestamp:

— spool filename with timestamp
col sysdt noprint new_value sysdt_var
SELECT TO_CHAR(SYSDATE, ‘yyyymmdd_hh24miss’) sysdt FROM DUAL;
spool run_filename_&sysdt_var.Log

Update 12/07/12

This was confusing, as I had the same name for column name and the SQL variable created by new_value. I've renamed the two accordingly. This works.

Also, I posted about spooling to HTML some time back. You can combine the two, of course:

col spoolname new_value spoolname_var

select 'tablist_'||to_char(sysdate, 'yymmdd_hh24miss') || '.html' spoolname from dual;
prompt &spoolname_var

SET MARKUP HTML ON -
HEAD "<TITLE>Session Stats</TITLE> -
<STYLE type='text/css'> -
<!-- BODY {background: #FFFFFF} --> -
</STYLE>" -
BODY "TEXT='#000FF'" -
TABLE "WIDTH='90%' BORDER='5'"

spool &spoolname_var

SELECT username, logon_time, status FROM v$session;
spool OFF;

creates a nice HTML file with a list of all users currently logged on to the database. This is saved in a HTML file with timestamp in it's name!

TomCat + Sybase EA Server

Today another coworker of mine was trying to get XAMPP setup work with our EA Server. (We have PB application and Java objects deployed to this EA Server). He has a complete setup of XAMPP with Apache and Tomcat. He has a JSP page that uses the PB objects. This got to a point and crashed with "class not found" error. Clearly pointing to a classpath issue. But which one?

He mentioned that he had to set the EA Server first before he started XAMPP. Clearly there are some conflicts in the classes shared between the 2. EA Server is normally started using a batch file that reads the env classpath (windows settings in control panel) and adds its own classpaths locally. (SET LOCAL inside batch file). So whatever EA Server set is not visible to XAMPP but not the other way around. That's why he had to start EA Server first.

We then looked for the specific class that was "not found". This happened to a stub for the PB objects. In Jaguar manager (aka Sybase Central, EA Server admin) there is an option to generate stubs (and skeletons if needed). This typically points to SYBASEEASERVERHTMLCLASSES directory. We found out that he also similar classes in SYBASEEASERVERJAVACLASSES directory. We knew these were not used, as they were not in the class path. Then we searched for classes or jar files with similar names. We stumbled on a jar file that contains part of these stubs, inside XAMPP. (He uses OpenBD for coldfusion applications, which is why he has XAMPP in the first place). This jar was old and had part of the stubs classes and did not have the one we were having trouble with.

Apparently Tomcat was automatically loading this jar (Anything inside the lib directory gets loaded automatically). We found that when we tried to rename the file.

Now that we found out the classes in trouble, we thought we could just add the class paths needed to the environment classpath and hope Tomcat would pick it up. This did not happen. On the web, I saw a note about how Tomcat handles classpaths differently than regular Java. There was a suggestion to the classpath to the batch file that starts Tomcat. We didn't try that yet.

For now, we decided to copy the classes to the library folder, so Tomcat can automatically load the classes. So, we shut down XAMPP, regenerated the jar file (in EA Server stub generation, there is an option to generate a jar instead of class files) and copied to XAMPP lib folder and everything worked OK.

Linux: Wget, Yum

One of my co-workers was trying to install Oracle linux update to Fedora linux box. She was trying to do a wget command based on some web site instructions. Her wget failed and was not able to get the repository file at public-yum.oracle.com.

wget http://public-yum.oracle.com/public-yum-ol6.repo

Per suggestions on the web, we tried using host or nslookup for the website. They could not find the web site.

Where as, google site was always found. host google.com and nslookup google.com worked fine. But not to the oracle server. Not even oracle.com showed up when we tried host, nslookup and ping commands.

After googling it up and trying to find solutions, we got to a point, where it pointed to some kind of network issue, probably proxy? I asked her to download the above file manually (she was always able to go to the web page directly in browser) and save to /etc/yum.repos.d directory. Then we tried,

yum list command.

This gave a different error:

IOError: <urlopen error (-2, 'Name or service not known')>

Some more web searches (http://serverfault.com/questions/76421/wget-cant-resolve-host) revealed this to be related to DNS. That was kind of strange. As we were able to ping and nslookup to google.com and we were always able to go to the oracle site in browser.

Anyways, we decided to edit the /etc/resolve.conf file and add some public name servers to it.

nameserver <dns ip address>

This didn't help. Then she logged into a unix box and we looked at the /etc/resolve.conf there. It had only one name server. We copied to the file on Linux box. Bingo!! It worked. So, the problem was that the name servers she had listed in the resolve file didn't resolve oracle servers for some reason.

Yum, Wget etc

I really didn't know much about Yum and Wget until today. While working on this problem, learned a few things:

Wget is a cool utility to download files. Per the wikipedia page , wget (formerly geturl) is a non-interactive replacement for FTP, but it also works with other protocols like HTTP, HTTPS etc. It can download recursively.

Yum is a package manager/installer for Linux. See here for a definition:

http://www.wisegeek.com/what-is-yum.htm

Turns out the .repo file we were having problem with in the first place, is the repository that has some configurations to be used by Yum to download more files and install them.

Note: If you are editing any of the /etc/ files and you already have the console window open, chances the new settings will not be effective. Close all the console windows and start a new one.