Friday, November 26, 2010

SharpSSH: A Recompiled Version Compatible with Latest 64-bit Windows Server System

One of my recent project works is to integrate our CRM solution with a few data sources from different vendors, which requires me to develop an ETL component to download incremental update files from SFTP sites before processing them in our CRM application.

Since there is no support of SFTP in .NET framework itself, I have to look for alternative solution, so I ended up with the open source library called SharpSSH, which is pretty promising.

I was very happy that it worked with my first try on my local development VPC image which is on Windows Server 2003, 32-bit edition. But as soon as I promoted my ETL component to our UAT environment which is on Windows Server 2008 R2 (of course, 64-bit system), I ran into a "Bad Data" error when my component was trying to make connection to the SFTP site.

I did a little Google search, and found the solution here. I followed snowcoder's instructions, made a bunch of changes, and recompiled the library. Then everything worked for me on the 64-bit system.

Here are the download links, I hope that they could save you some time if you ever need this component (There were some errors in snowcode's original code he pasted there probably due to HTML formatting issue):

SharpSSH- Library (DLLs only)

SharpSSH- Source Code

[DISCLAIMER] I am by no means an expert on hardcore networking or cryptology programming (I have no intention to become one either at this moment), I am posting the recompiled library just for your convenience. All credit goes to Tamir Gal, snowcoder, and Mono Project. The license should stay as whatever it is originally.

[Further Note - Updated December 2011] There is another open source library in the community called SSH.NET, which you might want to check out. The library has got support for .NET 3.5, .NET 4.0, Windows Phone, Silverlight respectively. It looks very promising, and it has been actively maintained by the author.

Hope this helps!

Thursday, November 25, 2010

Non-sense CRM4 Customization Import Error

Here is another CRM4 customization import error that I recently ran into, but it's a non-sense and probably a nasty one.

What the problem is, I have been trying to update my development VPC image from rollup 9 to a more recent one. As soon as my VPC image is updated to any rollup after 10, I start to receive an error message when I try to import CRM customizations. The weird thing is, this error even happens in the same environment. By the same environment, I mean that it happens even I was exporting the CRM customizations and then re-importing the same customization file on the same server box.

Here is what the error message looks like:
Failure: incident_activity_parties: Cascade link type 'NoCascade' is invalid for Delete.
Nonsense Customization Import Error
Since I had a little bit cycle today and I was aware that Microsoft has just released Rollup 14 last week, I thought it would be a good chance to give it another try. I usually had pretty good confidence at Microsoft team's fast turnaround in terms of bug fixing. Unfortunately it's not the case this time, the same error persisted after I updated to Rollup 14.

I was naive enough, I thought that I could become a hero if I could find a solution for this issue. So I turned on CRM server trace log, and re-ran the import process. After about two tries (first try failed for a different reason), I got 115 log files in my CRM trace folder (they consumed 1.1GB hard drive space on my poor VPC image). However those log files didn't give me much helpful information, except another error message that was thrown before the previously mentioned message, which stated that "Failed to import entity relationship incident_activity_parties", which I don't have any clue why it has happened either.

After about 30 minutes of experimenting and Google search, I gave up on this. A few folks in the community have been suggesting to muck around the customization xml file, which I am not so encouraged to go this path, as I believe that could possibly cause more problems in the long run that it actually solves.

[Update - Feb 28, 2011] I finally figured out what the problem is, I had Data Migration Manager installed on my server. As soon as I uninstalled DMM and the problem went away. Still a mystery for me, but I can now finally go with more recent rollups.

Wednesday, November 24, 2010

JavaScript Snippet: Test Whether a JavaScript Variable is Regular Expression

Here is a short snippet if you ever need to check whether a JavaScript variable is regular expression.

function isRegExp(obj) {
   return === '[object RegExp]';

You might be wondering why on the earth you ever need this function. It could be when you want to develop a validation framework that takes a variable for validation purpose, and you want to make the validation framework flexible as flexible as possible, so that the variable could be a function, string or even a regular expression. Your validation framework would do different thing based on the type of the validation variable that is passed to the framework. Hope this makes sense to you now, maybe you have a totally different reason for using it, which is absolutely up to you... :-)

[UPDATE - Dec 10, 2010]: A couple of hours after I published the original blog post, I figured that I should be able to use JavaScript instanceof operator to achieve the same purpose, which could be an easier solution. But further reading indicated that will not work correctly if you are evaluating a cross-frame RegExp object. If you are interested, you may continue reading a very detailed explanation with regard to a similar testing of JavaScript Array object. I have also updated my code a little bit so it looks pretty much the same as kangax's isArray function now, shamelessly. ^_^

A lesson learned from this is, you should only trust JavaScript instanceof operator if your code never has the need to go cross frames. If you want a robust solution to evaluate the type of a JavaScript object, you better take the similar approach as I have just illustrated you.

You can always expect some surprise from JavaScript. Stay tuned until next time.