0.4.5 changed timezone comparison behavior

[jjb]

this spec:

Timecop.freeze(@t = Time.now)
expect{
  # code...
}.to change(@object, :attribute).from(nil).to(@t)

used to pass with 0.4.4, but now fails with 0.4.5, with the message:

attribute should have been changed to 2012-08-10 23:10:57 -0400, but is now Sat, 11 Aug 2012 03:10:57 UTC +00:00

If I use pry to set a breakpoint with 0.4.4 and 0.4.5, I observe:

• @t.to_f and @object.attribute.to_f are indeed identical as they should be
• the timezones for each object are the same with both versions.

So, my guess: pry is doing its timefreezing job, and the Time objects are the same, but for some reason the comparison is now taking the timezone into consideration where before it was not. Is this expected/possible?

[travisjeffery]

I can't say for sure without seeing that code block or a reproduced failing test.

[ches]

From the test's inspect output, it looks like you're comparing a Time (from Time.now) with an ActiveSupport::TimeWithZone, and your development system's zone is different from the app's configured default. Likely you just need to use Time.zone.now.

[travisjeffery]

Agreed @ches, the only change in 0.4.5 was to make time instance type consistent for comparisons so if you're having a problem now it's likely because you were comparing instances inconsistently before.

[jjb]

I just discovered that my use of Timecop was incorrect, and that is probably what was causing this problem. I didn't know that I had to do Timecop.return, for some reason I assumed that it did its own cleanup at the end of an rspec example.

The top example in the Readme is good and clear, so I'm not sure where I got the idea. But I'll submit a tweak to the other example in the readme just in case others run into the same problem.

Thanks for Timecop!!

[jjb]

ack, i take it back, that wasn't what caused my issue. I'm still having the issue, and Time.zone.now didn't help. it did however make the output more consistent. comparing time Time objects i get:

foo_at should have been changed to Mon, 08 Oct 2012 16:31:55 UTC +00:00, but is now Mon, 08 Oct 2012 16:31:55 UTC +00:00

(they are identical)

suspecting that the difference is sub-second, I switched the comparison to to_f of the Time objects in order to get more meaningful failure output. however the result was that the specs now pass. the same with to_i.

/cc @ches