stephen-smith wrote on Friday, February 04, 2011:
Discussion here, so we can have BBCode.
This should be the last PHP lint error in the tree. I am also working on hooks and/or an automated system to help prevent PHP lint errors from being introduced to the tree in the future, in a dependent branch.
stephen-smith wrote on Friday, February 04, 2011:
This is just another one-line patch, but it’s already near the end of the work-week here. With that in mind, I’m going to hold off on merging this into the main repository until Monday afternoon, ~72 hours.
As far as the hook goes, it works but IMO it is far too slow for a pre-commit hook. I’m working on making it fast enough to use for most developers.
bradymiller wrote on Saturday, February 05, 2011:
hi,
Not very excited by the hook(just don’t want something that will further deplete resources of the code integrators). Would be nice to instead have a auto test of the build that testing each new build and threw the error on the test server. Like the test server here:
http://www.openmedsoftware.org/wiki/Automated_Testing
-brady
stephen-smith wrote on Saturday, February 05, 2011:
The hook would be something that is run automatically by git on the local developer’s machine each time (git commit) is run. This pushes errors down to the person that should be responsible for fixing them.
Local hooks, in particular pre-commit, can’t be enforced at clone time. (Security issue, as you can imagine.) So, even with a hook provided, I’d still need to try and do some automated checks. Still, I’d like to encourage developers to use some hooks to catch potential errors as soon as possible. Most likely, I’d even provide a hook installer script to make it as easy as possible to use them.
Looks like the automated testing is offline for now. I’m getting forbidden / not found errors on all the links. What I’m doing is not really testing, per se. It’s static code analysis, so it doesn’t actually involve running the code. I’m also a big fan of testing, particularly unit testing, but I’ve seen some pretty remarkable results from adding static code analysis to the mix as well.
Do we have infrastructure to set up something an automated process could report to? It looks like some sort of automated post-push process is actually preferred by the project, so I may switch my focus to that.
bradymiller wrote on Monday, February 07, 2011:
Hi,
Placing the hooks on the primary developers git repo is much more palatable.
The online testing when offline awhile ago after lots of work in the Installer class. This was contributed by acmoore; hopefully he’ll weigh in on a status update. Would be nice to do the unit testing along with the static testing on each new build.
-brady
stephen-smith wrote on Tuesday, February 08, 2011:
Merged into main repository in commit cbd0.
(Just the patch attached to the tracker item. The hook and/or automated testing is not ready yet and I’ll open another tracker item for it when it is ready.)