Should App.run_after use GCD?

[taybenlor]

I'm currently spicing up one of my RubyMotion apps with BubbleWrap. A little bit of code I have works fine with GCD but crashes using App.run_after. My knowledge of the Objective C threading model and Timers vs GCD is limited, so I'm not sure whether this would break a bunch of existing code.

Anyway, my example:

loc = @mapView.region.center

Dispatch::Queue.main.after 0.4 do 
  new_loc = @mapView.region.center
  if loc == new_loc
    self.updateLocation
  end
end

The change to BubbleWrap would be super quick, though I haven't made the change and tested it. Just wanted to check it was a sane thing to do.

[siuying]

If you replace the loc with @loc, the code below should work in run_after.

You need to use @, because it will create an instance variable in the
object, and thus keep it from being released by GC.

Ben Taylor 於 2012年12月24日星期一寫道：

I'm currently spicing up one of my RubyMotion apps with BubbleWrap. A
little bit of code I have works fine with GCD but crashes using
App.run_after. My knowledge of the Objective C threading model and Timers
vs GCD is limited, so I'm not sure whether this would break a bunch of
existing code.

Anyway, my example:

loc = @mapView.region.center

Dispatch::Queue.main.after 0.4 do
new_loc = @mapView.region.center
if loc == new_loc
self.updateLocation
end
end

The change to BubbleWrap would be super quick, though I haven't made the
change and tested it. Just wanted to check it was a sane thing to do.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180.

[taybenlor]

So why does this work fine with GCD, but not using App.run_after?

[taybenlor]

And, if it does work fine with GCD, but not with an NSTimer - doesn't that make a great cause for switching to GCD over the NSTimer?

[siuying]

This depends on the implementation of the Dispatch. In any case, the loc is
a local that is not retained, and it is not safe to use in block, even it
work now, it will cause you trouble some day.

Ben Taylor 於 2012年12月24日星期一寫道：

So why does this same example work fine with GCD, but not using
App.run_after?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180#issuecomment-11660457.

[taybenlor]

Is there something in the RubyMotion docs explaining this? It massively violates the way blocks work in Ruby.

• Ben

On Tuesday, 25 December 2012 at 12:19 AM, Francis Chong wrote:

This depends on the implementation of the Dispatch. In any case, the loc is
a local that is not retained, and it is not safe to use in block, even it
work now, it will cause you trouble some day.

Ben Taylor 於 2012年12月24日星期一寫道：

So why does this same example work fine with GCD, but not using
App.run_after?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180#issuecomment-11660457.

—
Reply to this email directly or view it on GitHub (#180 (comment)).

[siuying]

That is not the reason of choosing GCD vs NSTimer. Generally GCD is a lower
level implementation than that of foundation classes. Class like
NSOperation are implemented using GCD.

Ben Taylor 於 2012年12月24日星期一寫道：

And, if it does work fine with GCD, but not with an NSTimer - doesn't that
make a great cause for switching to GCD over the NSTimer?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180#issuecomment-11660480.

[taybenlor]

No that's fine. I'm more concerned about the fact that you're claiming blocks are no longer closures.

[siuying]

I don't sure are there any official documentation about this. But yes, the
GC of rubymotion is not like regular ruby. I hope it will be fixed someday.

Ben Taylor 於 2012年12月24日星期一寫道：

Is there something in the RubyMotion docs explaining this? It massively
violates the way blocks work in Ruby.

• Ben

On Tuesday, 25 December 2012 at 12:19 AM, Francis Chong wrote:

This depends on the implementation of the Dispatch. In any case, the loc
is
a local that is not retained, and it is not safe to use in block, even
it
work now, it will cause you trouble some day.

Ben Taylor 於 2012年12月24日星期一寫道：

So why does this same example work fine with GCD, but not using
App.run_after?

—
Reply to this email directly or view it on GitHub<
https://github.com/rubymotion/BubbleWrap/issues/180#issuecomment-11660457>.

—
Reply to this email directly or view it on GitHub (
#180 (comment)).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180#issuecomment-11660616.

[siuying]

Refer this blog.blazingcloud.net/2012/07/16/rubymotion-block-scope-bug/

It should have been fixed. I remember I've encountered some similar issue
in a different case and reported them.
Please report any issue if you find some unexpected behavior.!

Ben Taylor 於 2012年12月24日星期一寫道：

No that's fine. I'm more concerned about the fact that you're claiming
blocks are no longer closures.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180#issuecomment-11660681.

[siuying]

About this issue, would you please create a test case that reproduce it?

Ben Taylor 於 2012年12月24日星期一寫道：

No that's fine. I'm more concerned about the fact that you're claiming
blocks are no longer closures.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/180#issuecomment-11660681.