'const' fix and support for specifying retina devices for subliminal-test

[nzhuk]

No description provided.

[wearhere]

You're right that a multi-word SIM_DEVICE value must be single-quoted for it to ultimately be accepted (by defaults write, in the set_simulator_device script)--but it looks like defaults write actually requires both double- and single-quotes (like `"'iPhone (Retina 4-inch)'"), which is a little weird; I feel like the expectation is that you can usually get away with simply using one pair of quotes.

So, I feel like a better fix might be to have set_simulator_device single-quote the argument to defaults write:

defaults write com.apple.iphonesimulator SimulateDevice "'$DEVICE'"

Thoughts?

[nzhuk]

Yes, that could be a cleaner solution. The only thing I'm a bit worried about is that with the proposed solution set_simulator_device will produce different results than what you get when you change device type from iOS Simulator app's Device menu.

When you select "iPad" from the iOS Simulator -> Device menu, SimulateDevice value in com.apple.iphonesimulator prefs plist is set to iPad. If you run

defaults write com.apple.iphonesimulator SimulateDevice "'iPad'"

then SimulateDevice will be set to 'iPad'

However, running:

defaults write com.apple.iphonesimulator SimulateDevice "'iPad (Retina)'"

strips single quotes and sets SimulateDevice to iPad (Retina)

That said, iOS Simulator seems to handle quoted SimulateDevice values just fine, so this discrepancy doesn't seem to cause actual problems.

[wearhere]

Yeah if it can handle single-quoted arguments fine then I think that's the way to go, rather than putting the burden on the user to remember when to use single-quotes. Please fixup that change into the first commit and force push to update this request, then.

[wearhere]

Question for you on the first commit, @nzhuk. Second commit looks good. Btw I don't think that this request broke the build, it looks like Travis just hiccuped.

[nzhuk]

Ok, I'm a hg guy so I probably screwed up this git rebase/fixup thing in my fork, but the end result of all those commits seems to be what is needed.

[wearhere]

Thanks for updating the request, @nzhuk. I'm a bit puzzled by the sequence of the commits though: it looks like you have two changes (adding single quotes to the device model name, and adding the const qualifier) made multiple times? I'd think you'd only have two commits when you finished, one for each of those changes. Those look to be (roughly) 2130b10, and 6ba67b0. 2130b10 contains the code that undoes e03508d--you can get rid of that by squashing 2130b10 into e03508d, while keeping 2130b10's commit message.

[nzhuk]

Ok, I've created a new pull request with just those two commits ( #45 ), please ignore this one.