Notice (2018-05-24): bugzilla.xamarin.com is now in
Please join us on
Visual Studio Developer Community and in the
Mono organizations on
GitHub to continue tracking issues. Bugzilla will remain
available for reference in read-only mode. We will continue to work
on open Bugzilla bugs, copy them to the new locations
as needed for follow-up, and add the new items under Related
Our sincere thanks to everyone who has contributed on this bug
tracker over the years. Thanks also for your understanding as we
make these adjustments and improvements for the future.
Please create a new report on
Developer Community or GitHub with
your current version information, steps to reproduce, and relevant error
messages or log files if you are hitting an issue that looks similar to
this resolved bug and you do not yet see a matching new report.
Created attachment 14021 [details]
Sample Xamarin.iOS project demonstrating bug
In the attached sample, inside a method ("PopViewControllerAnimated(...)") attributed with [MonoPInvokeCallback(...)], the following line of code returns an __NSMallocBlock__ instead of a UIViewController, but only when "Enable-device specific builds" is unchecked, and only on 64-bit physical devices (tested on iPhone 6s). On 32-bit devices (tested on iPod Touch 5th Gen), the code returns the expected UINavigationController instance:
To trigger this behavior in the sample, tap the button on the first screen, which navigates to a second screen; then hit the back button to navigate back to the first screen, and the exception will occur. Bug described in comments in Hook.cs, too.
There is a bug in BlockLiteral.SetupBlock that causes iOS to think it's a block with a stret signature, effectively causing imp_implementationWithBlock to create a wrapper method that puts the arguments in the wrong place for all architectures but ARM64 (since ARM64 doesn't have any stret signatures).
This has been fixed (maccore/master: 20b932ed52f61b73a872951fd20df7fede91b532).
In any case your implementation is incorrect, if you're using imp_implementationWithBlock the first argument to the method is the block, and then the rest comes after (and this is how it will behave once we've released the fix).
Fortunately you don't have to create a block and use imp_implementationWithBlock, you can just create a method with the proper signature for the direct call instead, and that works just fine: https://gist.github.com/rolfbjarne/1115b3e1aa14ee7acfd2
Thanks for info, I was using the SetupBlock approach from stack overflow article indicating bugs with the non-block approach in your re-work of the example. I prefer your approach. Maybe a bug in this area causing issues in the past has been addressed already.
Which stackoverflow post is that?
I based my original approach from this article
@Mark, thanks, that stackoverflow post looks right (except that it doesn't currently work due to this bug).
I am able to reproduce this issue with 32 bit device. Here is the screencast for the same: http://www.screencast.com/t/is1KkOodKBB
To verify this issue I have checked this issue with the master build on 32 bit device: monotouch-220.127.116.117_20b932ed52f61b73a872951fd20df7fede91b532. Now this issue is working fine. Here is the screencast for the same: http://www.screencast.com/t/MAP3GxQOMUT9
I have checked this issue with the latest C6SR1 build on 32 bit device:
monotouch-18.104.22.168_f146d47db0476e3519dc6e8229af74a4e867f386. Now this issue is working fine. Here is the screencast for the same: http://www.screencast.com/t/9LgNjCU9JUp