jcoglan ( James Coglan )

pull request commentUKGovernmentBEIS/beis-report-official-development-assistance

(1066) Users can see a list of invalid activities in the index page

@mec My commit message explains why we were observing weird effects from Bullet in terms of what it's doing behind the scenes. I discussed this with @MariluzHerrera and my decision went as follows.

Using Bullet.add_whitelist in both the environment config and in individual tests is not advisable since it can lead to conflicts, where a test turns a particular whitelisted thing on and off, changing the base state of the app's config, and those are hard to debug, as we've seen here. The Bullet.{add,delete}_whitelist API does not seem designed for this type of use as it does not provide a reliable way of restoring whatever the prior state was. We need to pick one or the other.

Putting it in individual tests means people would need to add Bullet.{add,delete}_whitelist config to any test that hits a problematic controller action. This particular action is hit by a lot of tests, and it won't be obvious that one needs to add this config in order to silence the warnings Bullet produces, or even if that is a good idea. Getting warnings about code you didn't write is confusing.

Putting it in the environment config means we're globally ignoring a certain kind of optimisation problem -- you cannot. In general, I would rather be alerted to these, but I think ignoring an unused eager load is much less dangerous than ignoring an N+1 query, so I'm happy with the trade-off. In this case, the unused eager load comes from the fact that not all Activity objects use the parent association during validation -- fund activities don't require a parent. The fact that some but not all loaded activities use the parent association causes Bullet to raise an error whether you eager load the association or not -- it's impossible to make this code pass Bullet inspection.

I'm open to making the opposite decision and putting Bullet whitelist logic in individual tests, and not in the environment config, I just think we need to pick one or the other.

MariluzHerrera

comment created time in 6 days

issue commentfaye/faye

Unexpected error while processing request: uninitialized constant EventMachine::Deferrable::Rails

None of the lines in this backtrace refer to the constant Rails, but a lot of them are concerned with running callbacks that will yield to user-supplied code. Your Faye server setup code does not appear to reference anything from Rails, or load application code, and does not set up any event listeners or extensions:

require "faye"
faye_server = Faye::RackAdapter.new(:mount => '/faye', :timeout => 45)
run faye_server

The one problem I can see with this is that it sets the timeout to 45, which is greater than the Thin request timeout limit of 30. This may cause weird results.

What command are you using to start the server in faye.ru?

imi56

comment created time in 6 days

issue commentfaye/faye

uninitialized constant EventMachine::Rails (NameError)

Neither EventMachine nor Faye refers to Rails anywhere in their codebases. Have you bound any event listeners to Faye that depend on anything from Rails or your application?

imi56

comment created time in 6 days

issue commentfaye/faye

Faye requesting an insecure script <URL>

What's the code you're using to start the connection to the Faye server?

imi56

comment created time in 7 days

push eventjcoglan/dotfiles

James Coglan

commit sha ba910a185eb39cf23b4559c3eb5365c030ba8cab

Function for installing ruby with [email protected]

Changes in this PR

Screenshots of UI changes

Before

After

Next steps

Changes in this PR

Screenshots of UI changes

Before

After

Next steps

Changes in this PR

Next steps

Changes in this PR

Next steps

Screenshots of UI changes

Next steps

Changes in this PR

Screenshots of UI changes

Next steps

Screenshots of UI changes

Next steps