sort glob() output for consistent results

[binary1230]

Since glob() returns an ordering that is os-dependent, we need to sort here in order to ensure that all OS'es load the non-prioritized plugins in the same order

This change does not fix any loading issues, BUT it will ensure that if there IS a load order issue, it will happen either everywhere or nowhere. This eliminates "why is it working on my machine but not on the server".

sidenote: Originally I wanted to do this change as a precaution, however, it turns out it's necessary for my computer and magstock staging server to load things in the same order.

before fix:

non-prioritized plugin load order on magfest staging server:

['bands', 'barcode', 'magstock', 'uber_analytics']

non-prioritized plugin load order on my box:

['magstock', 'barcode', 'uber_analytics', 'bands']

after fix:

both environments now load in the same order:

['bands', 'barcode', 'magstock', 'uber_analytics']

[binary1230]

fixes #15

[binary1230]

in this example btw, magstock needs to get loaded after bands plugin because it overrides some stuff there.

the RIGHT answer is to put the band plugin in priority_plugins (a separate change that will be coming)

but, it's really confusing without this change because on my box band plugin happens to load before magstock (correct) and on staging it just happens to load after magstock (broken). so this change just makes sure either both boxes work, or they both fail. This eliminates one source of possible inconsistencies between multiple environments.

[kitsuta]

I've been ambivalent about whether this change was really needed but it seems you do keep running into trouble. If nothing else, this will catch config errors more reliably.

That being said, it looks like the Travis build failed?

[binary1230]

Yea seems to be falling what looks like an unrelated unit test only for
python 2.7, weird
On May 28, 2016 3:47 PM, "Victoria Earl" notifications@github.com wrote:

I've been ambivalent about whether this change was really needed but it
seems you do keep running into trouble. If nothing else, this will catch
config errors more reliably.

That being said, it looks like the Travis build failed?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#16 (comment), or mute
the thread
https://github.com/notifications/unsubscribe/AFKYyEAe_BLro8XkqFVdoC8fImy1u9g-ks5qGJvfgaJpZM4IpEqN
.

[kitsuta]

@binary1230 Do you want to merge anyway?

[binary1230]

hey so I just re-ran the travis build and this time it worked, it looked like some kind of transient thing. The sideboard tests open websockets and stuff, so it's possible something got janky there.