Automatic VectorCAST with Build Servers from Workstation
The customer has a Linux build server that gets accessed from a Windows workstation. All operations from a user standpoint is executing from the workstation, including downloading to the target. Files are available in a shared folder (Samba).
A full remote build from Windows by installing VectorCAST and executing any preprocessing, compiling, and linking operations on the build server is possible.
Some of the times, all those steps can be overkill for what you want to achieve.
For these “easier cases”, you can usually (1) installed VCAST on Linux, and (2) scripted both Windows and Linux so the execution is 100% automatic.
How it works:
- When a test (or range test) needs to run in an environment on the build machine (either in stdin/stdout or stdout only mode), a script launches. That script launches a countdown in vpython (Pyside) for X seconds (from the Execute command). It waits either for a specific file to show up in the shared directory or the timeout to be done.
- On the workstation, VectorCAST agent sees a new test should be run because the executable showed up in the shared dir and runs it on the board. Four modes are implemented: Cygwin, openSSH, puTTY (which lets you input passwords), and customized (if we ever want to do anything else besides these). The menu changes when you select another mode, and the agent only starts listening when you have no red fields (so all fields are in a legal state).
- When the execution finishes, the data is saved in VCAST_STDOUT.DAT in the shared drive, then a finished file is also created. On the build server side, the timeout mechanism sees it, copies the data back to the working directory, and releases.
Find the build server agent here.