I use a homemade program written in Python, which has enough editing capabilities that a lot of puzzles I make never appear on a sheet of paper. wxWidgets does the GUI and also used to do the drawing until I changed to Cairo in a rewrite to get vector graphics. It does have its limitations, mostly when the grid is irregular. I've had intentions to clean it up enough to be released for over a year, but it seems like every time I strike something off the todo list, two more things take its place.

A screenshot of the program to give an impression of how it works: http://mellowmelon.files.wordpress.com/2011/12/screenshot.png
Should be noted that one major item on the aforementioned todo list is a revised interface made to be intuitive for someone other than me, not to mention documentation.

I know CorelDraw is very effective to use for doing that, but Adobe programs are more understandable to me, maybe because I am more accustomed to them, so I still prefer to work with Adobe Software.

My workflow uses 3 (technically 4) pieces of software and two devices:

Step 1: Draft the puzzle by hand in Sketchbook for iPad, export as PNG to Dropbox
Step 2: Draw nicely using Inkscape on my PC, save editable SVG and PDF copy for posting, save PDF to Dropbox
Step 3: Test-solve the PDF version in UPAD for iPad, checking for errors and comparing against the original drawing when errors are found.