On Dec 13, 2004, at 11:10 AM, Jason Gessner wrote:
David Cantrell wrote:Jason Gessner wrote:heheheh. unless you have already added a bunch of failing tests in anticipation of the changes.Changing the module without adding a test for the changes is kinda lame too!Andy Lester wrote:
So it's sort of adding make functionality with prove. The way the check
is running in the patch, the only criteria for updating it is changes
in the .t file, but what if what you're updating is the source file?
Detecting a change in the mod time for a .t file is less than optimal, I agree. If you have to go and touch your test file to tell it about a change in your module, that is kinda lame.
I agree with Andy, this is not something that should go into prove, but something which should wrap around prove.
It seems to me that a MANIFEST file could be used for this. Since the process is continually running, it would only need to read the MANIFEST once, and keep those values in memory. If there is no MANIFEST, a simple up-front directory traversal should do. I assume there would be some kind of timeout in between attempting test runs, after which it could loop through all the files and look for new mod times. Seems simple enough.
Keeping test histories and such get a little more complex though, at that point I would think we might need to capture output before Test::Harness gets to it.
Steve
