I have to use some third-party programs to compare exported image files because they are too big for TestComplete to compare. TC quietly fails to do the compare and actually reports that the compare was successful. I reported this and was told the images are too big for memory. I have 4GB on this machine. The images are 1275x41220 pixels.
The way I have it organized, the baseline files and my test-generated files to compare have paths and filenames containing the Project and Keyword Test names.
The command-line programs that I use to split and compare the image files don't run on the WinXP 64bit machine where I use TC, so the Keyword Test pauses with a manual checkpoint instructing to copy/paste a command-line into a (32bit XP) VMware virtual machine. The VM can access the host directory where all the files are. Before comparing externally, I split the image into chunks small enough for TC to handle. I then click "Success" in the Manual Checkpoint when complete.
Then the Keyword Test continues and runs a TC script that checks the directory for any chunks of the image that miscompare (the directory will be empty if they all compare OK). I run TC's compare method on the chunks that miscompare, which puts them nicely into the log file with the GUI tools to look at them. I obviously need to find replacement image tools that run on 64bit windows. I'm using a suite called Alchemy.
So, one thing I'd like to do is modify the Manual Checkpoint text on the fly so that the command line that's pasted into the VM has the Project name and Keyword Test name as parameters for the DOS batch file. Even if I find 64bit replacement tools, my script still needs the Project and Keyword Test names to operate on the image files. I call this script from an if-then-else block that I copy/paste into each Keyword Test thast needs it. These lines contain the name of the Keyword Test (as the name of the Manual Checkpoint and the directory where the files are to be found). The Project name is already a variable in several places. This is where I'd code the Keyword Test name as a variable.
I have (will have) hundreds of tests for a suite of 4 similar applications. I plan to clone the Project for each new version tested. I expect some number of Stores (Regions, mostly) to legitimately change from version to version, and I need to replace these stored Regions with the new image from the log (I wish there was a one-click way to do this, BTW) but still have the Stores intact for the older versions. That's why the Project name is part of the path for various baseline and temporary files. I currently set the Project name in the Project variables, which as I said is not too bad.
