Class and example reloading #31
Conversation
Before it would log an error and keep on going anyway. The Containment#protect block now throws the symbol that instructs guard that there was an error and to stop testing.
Additional custom_reloaders can also be passed in via options to support other frameworks like Rails 2.x or custom class reloaders. Closes jkutner#29
|
Excellent. Looking at it now. |
|
I should note that both the new |
| end | ||
|
|
||
| context 'with a custom error_handler' do | ||
| subject(:containment) { described_class.new(:error_handler => lambda { @custom_handler_called = true }) } |
There was a problem hiding this comment.
This spec failed for me, with wrong number of arguments (1 for 0). But changing the lambda to lambda { |e| ... fixed it
There was a problem hiding this comment.
Ah, I'll add a commit to fix that. I forgot that my JRuby was running in 1.8 mode. Thanks.
|
I'm finding that on subsequent runs, the count of examples is increased. So if the first run is |
|
Not at all. This was intended to fix that. Also, was that not happening for you before? Let me see if I can reproduce that. |
|
I'm testing it on https://github.com/jkutner/get_back. My Guardfile is this: It probably was happening with earlier versions and I just never noticed. |
|
It's not happening on my 1.8 rails app. I'll try get_back. |
|
Ok, I see what happens. When I kick off the tests by pressing |
|
I think we should merge this as is, and address the |
|
Ah, interesting. I never run the full test suite in guard since it takes about an hour ;-) This might be a quick fix. I'll take a look before you merge. |
|
ok |
|
That should do the trick. Now I unload the examples before each run in |
|
I didn't update the readme documenting the Rails reloading support. One of us should do that before the release, but I won't have time today. Also this should probably be at least a minor version bump (0.2.0?). |
|
I'll handle the readme. And I agree that this should be a 0.2.0 release. Awesome work! thanks for the code. |
|
I'm getting some pretty odd results when running the specs with your lastest commit. Are you seeing this? |
|
FWIW, the code works on a real project. And running the specs individually works too |
|
No, that's pretty strange seeing a NullPointerException since this has no |
|
May be the JVM version. I've found a workaround though. Do you see any problems with do this in |
|
I can't think of any other way that specs would run, so I think that should |
|
Merged! Good stuff. |
|
I'm getting this error now with the latest release. Investigating further. |
|
Are you seeing that with the 0.2.0 gem and not the code in master? |
|
I am seeing this in master. It seems to stem from the controller watch in the Guardfile: As the error goes awayf for me if I just map that watch to one of those values in the array. Will try to dig deeper in a bit. |
|
Opening a new issue #34 |
This PR contains a few different changes. Please review carefully.
Use Rails 3.2+ class reloading when available
Additional custom_reloaders can also be passed in via options to support
other frameworks like Rails 2.x or custom class reloaders.
I wrote support in
Reloadersfor prepending, but I didn't hook it up to jruby-rspec. Probably before this is merged, we should decide to either drop prepend support or support a way to prepend reloaders. We might also consider supporting a way to replace the default reloaders. I wanted to get your opinion on this wholeReloadersthing before I fleshed it all out.This closes #29
Unload previous examples so that they do not accumulate
This fixes #30
Syntax errors now stop the specs from running
Before it would log an error and keep on going anyway, sometimes scrolling an important error off of the screen. The Containment#protect block now throws the symbol that instructs guard that there was an error and to stop testing.