Build clean removes features from compiled orphans

HFCDev · 2021-05-11T14:23:01+00:00

There was an error rendering this rich post.

Hi,

I've seen this behaviour for a number of years now but it's very non-determininstic and very hard to pin down. It also seems to only occur with orphaned objects. The scenario is something like this:

* I orphan a request handler (or a Utils pkg) into my oscript project;

* In the source, I add a new method to the orphan taking the form:

public function myNewFunction(Object prgCtx, ....)

end

* On building, I now have an orphan with this new method;

* I later do a build clean after making changes elsewhere (the key to this is that the changes are made elsewhere);

* Customer goes to use module - they get stack trace because in the orphaned object, the stack trace says that method myNewFunction not found.

* On examination in CSIDE, the new method is missing from the Ospace view, the Module view, but it shows up in the Overview.

* There are NO reported errors or problems with the source - the function simply gets ignored;

This is obviously alarming. Partners are depending on CSIDE being somewhat reliable. THis is all with the latest version of the plug-in (21.2) and the latest version of Eclipse installed using the Eclipse installer.

The worst thing about this is the seeming randomness of it. You could try reproducing and never see this, and then a year later, you will be plagued. When this occurs, it's always the same orphaned object (BTW it's always a problem with orphans).

This might be part of the problem but when you add a method to an orphan that is later used to override an object (using the Chris Webster method), you see in your ospace the orphan with the function name twice - once as a new feature, and once as modified.

I will provide screenshots to OpenText development but there are customer confidentiality issues I need to sort out first. And once again, as a 3rd party Technology partner. I AM UNABLE TO OPEN A SUPPORT TICKET!!!!!!!

-Hugh

Find more posts tagged with

Comments

HFCDev

PS, if anyone from OpenText is watching, I would be happy to send the screenshots. They contain information about partner IP, which is why I would rather share them with OpenText exclusively rather than on the forum. I can also do a screen share where I can show this bug in motion.

Dave Carpeneto

Hi Hugh - I've not seen this before. I'll take screen shots ( dcarpene@opentext.com ) . Alternately / ideally: screen share of the steps would be fantastic (even if it's not reproducing at the moment: the specific steps taken would hopefully give hints as to what's going on here).

Thx, Dave

David Henshaw

@Hugh - what's the "Chris Webster" method you're referring to?

HFCDev

@David_Henshaw I posted it in Slack, but in essence, it's a technique where you do the override by orphaning the object in question, then in the __Init() method, you compare feature by feature the new or overridden features on your orphan to its parent. For any feature on the orphan that you find that is net new or overridden, you write that feature to the parent and if it's an existing feature, you first rename its original method to methodname_ospacename(). So if you had an ospace called abcospace, and you have overridden a method called Execute() in the orphan, you rename the Execute() method on the original to Execute_abcospace() so your code gets executed when orgmodule.Execute() gets called. In your ospace orphan you can then "wrap" the method, do your thing and call Execute_abcospace() so that the original code gets executed.

The WebReportsWF module used to do this (before they brought it into core), and I think this is not far off from how customizationsRT worked. Chris Webster is often credited as the guy who originally came up with the technique although I'm not 100% sure who invented it.

-Hugh

HFCDev

PS I would love it if there was a frank discussion between OT and the larger 3rd party tech partner community about how we should be doing overrides when we need to.

Chris Meyer

On examination in CSIDE, the new method is missing from the Ospace view, the Module view, but it shows up in the Overview.

Does it show up in the .os file if you open it with a text editor?

Appu Nair

I think it is somehow related to the Super in OOP. It is easy to understand which is the module that overrides which is a convention Chris Webster, who many of us call our Oscript Guru, might be able to explain it better but essentially I just follow what he says is a simple utility Oscript and plug in a few things all the rest of his "mind" does. It has one advantage over CustomizationsRT which makes you load an extra module, this one can stay as part of your module. It also successively can change e the web lingo compiler script similar to what one does with HTMLMap.

@sıɹɥɔ

#AOTD1D

HFCDev

@Chris Meyer Yes it shows up in the editor; @Appu Nair The method you describe in WebReports is exactly what I'm talking about. If there is a slicker way of doing overrides where you need them, I'm all ears. In the meantime, the original problem occurs seldom, but it's somewhat random. That means you can do a build clean, ship your module, thinking all is well in the world, and a method will be missing from the ospace.

-Hugh

John Simon

I have this issue as well. For me it only shows up in some modules, not all. I have always associated it with the override method discussed above AND the implementation of a new method. So in @Appu Nair's example if my .Patch() method called a new ._MyNewMethod() method then this behavior may show up. I clean the module then check the Module Explorer tab to see if my methods exist and if they do not then I clean the module again and it always fixes it. Kind of a pain as the clean can take quite a while (10 minutes).

Cheers...