Rake features and cucumber feature/name not producing same result


#1

Hi all.
If I run rake features or cucumber features/* I get one failing FIT
scenario
in one of my features.

If I then run that feature that contains the FIT table with the failing
scenario manually, it passes.

I run it again with rake features or cucumber features/* it fails, run
individually, it passes.

Obviously some state is carrying over between the features.

Any idea where i would start looking to debug the problem?

Are features transactional? Should I be investigating inter feature
state
contamination?

Mikel


#2

On 5 Nov 2008, at 03:58, Mikel L. wrote:

Obviously some state is carrying over between the features.

Any idea where i would start looking to debug the problem?

Are features transactional? Should I be investigating inter feature
state contamination?

Mikel

What is the failure message you’re seeing?

Could it be this?
http://rspec.lighthouseapp.com/projects/16211/tickets/60-patch-next_column_index-not-resetting-after-large-tables

cheers,
Matt


#3

On Wed, Nov 5, 2008 at 7:10 PM, Matt W. removed_email_address@domain.invalid wrote:

Are features transactional? Should I be investigating inter feature state
contamination?

What is the failure message you’re seeing?

Could it be this?

http://rspec.lighthouseapp.com/projects/16211/tickets/60-patch-next_column_index-not-resetting-after-large-tables

Alas no.

I’m getting expected: “1”
got: “0”

on a simple should change(Model, :count).by(1)

It follows the specific model around as well. If I change order of the
fit
table or not (the fit table has 20 rows of different models all being
exercised the same way, they are part of a replication tool that copies
a
database). The only time it passes is when I run the feature
independently.

I guess I’ll have to look at what the difference is between ( rake
features
|| cucumber features/* ) and cucumber features/specific.feature and try and
hunt down what beginning state I could be missing.

Any other ideas?

Mikel


#4

I guess I’ll have to look at what the difference is between ( rake
features || cucumber features/* ) and cucumber
features/specific.feature and try and hunt down what beginning state I
could be missing.

Any other ideas?

Mikel

Are you running using Webrat or Selenium/Watir?


Joseph W.
http://www.joesniff.co.uk


#5

On Thu, Nov 6, 2008 at 11:28 PM, Joseph W.
removed_email_address@domain.invalidwrote:

If you are using cucumber with rails and used the rails generator you
should find in your generated features/steps/env.rb file (In the latest
cucumber version 0.1.9 this is in features/support/env.rb):

Yes, I was already using the transactional fixtures. This wasn’t the
problem.

The final problem was actually two code bugs. One on a trigger in the
database and the other was a namespace confliict in my code.

Those two combined basically had me totally confused :slight_smile:

Mikel


#6

On Thu, Nov 6, 2008 at 2:53 AM, Joseph W.
removed_email_address@domain.invalidwrote:

I’m getting expected: “1”

                    got: "0"

on a simple should change(Model, :count).by(1)

It follows the specific model around as well. If I change order of the
fit table or not (the fit table has 20 rows of different models all being
exercised the same way, they are part of a replication tool that copies a
database). The only time it passes is when I run the feature independently.

Are you running using Webrat or Selenium/Watir?

Yes! Running webrat, latest trunk version (as of yesterday anyway).
Though
this feature doesn’t use it specifically, webrat is loaded with a
require
line in the features/steps/env.rb file.