ASC 2.0 Compiler bug -- still biting us

[rskdbl]

My team is still impacted greatly by a lingering ASC 2.0 compiler bug.  This is a bug that unpredictably results in build artifacts that throw Type Coercion errors at runtime.  This issue affects both Flash Player and iOS targets.

Bug 3562040 was  opened almost three years ago on this, as well as the related thread here: Type Coercion failures with MovieClips in ASC 2.0  And this issue was reportedly a high priority issue 5 months ago, but I have yet to see any note about it in the AIR 19, 20, or 21-beta release notes since: What happened to this known issue in AIR 18? Was it fixed? (ASC 2.0, Type Coercion: 3578605)‌  Is there any progress to report, or an updated timeframe in which we can expect a fix?  Is there any way we can help the Flash/AIR team get to the bottom of this?

I don't think I can over-stress how impactful this issue is to our development cycle.  It affects our productivity, it affects team morale.  And I would challenge anyone to feel good about choosing a technology that is dependent on a closed-source SDK when critical regressions can be unpredictably introduced into your product by its compiler, independent from any change to source.  And especially when the issue goes unaddressed for years.  I don't see how new features in the runtime can be prioritized when issues like this linger.  Can we please get attention on this?

Thank you,

Russell

[chris.campbell]

Hi Russell,

Our team is still working on this bug, but we've recently discovered a potential workaround.  Could you try replacing the \AIRSDK\lib\mxmlc-cli.jar file with the one in the zip file linked below.  We'd be very interested in hearing the results.

Shared Files - Acrobat.com

Thanks,

Chris

[rskdbl]

Thanks chris.campbell‌.  We will try this out and let you know the results.

[rskdbl]

chris.campbell‌, early results are not good.  After patching the SDK, we have since had a runtime break occur in our AIR app, which manifests both on iPad (the .ipa target), as well as on Mac (the .air target, wrapping the same main swf artifact).  The bad swf actually stems from a particular swc library, which is an intermediary artifact in the build process.  Once that swc was rebuilt, the app no longer produces the runtime issue.

Thoughts/Questions:

1. Would your engineering team be interested in taking a look at the good and bad artifacts I've saved from the builds?
2. Is the patched mxmlc-cli.jar expected to work with AIR 18 SDK, and for both Mac/Win?
3. Perhaps we need a similar patch for the compc-cli.jar, since the issue manifested -- at aleast this time -- from a built swc, and not a swf.  Could you supply a compc substitute for us to test?
4. We're curious how the patched mxmlc-cli.jar is supposed to affect the compiler, since it is presumably just a stub for any paths to the old compiler, and it appears to just contain a manifest file.  Can you provide an explanation of how this works, or what to look for to know that it's working?

@

Thank you,

Russell

[jindal_nipun]

Hi Russell,

Thanks for pointing out. The jar shared won't impact as you mentioned.

I am sharing compiler.jar which should give you the expected results.

https://www.dropbox.com/s/vlvmcwnfycowse7/compiler.jar.zip?dl=0

Thanks for your time. Can you please give it a try and let us know.

Regards,

Nipun

Adobe Air Team

[rskdbl]

Thank you jindal.nipun‌ .  I will try this out and get back to you.

[rskdbl]

Early results are in and look very promising.  Repeated compc runs result in swc artifacts that are binary equivalent from build to build.  This is a very different result from what I see with the released (old) compiler.jar -- at least in the AIR 18 SDK we're still using -- where the swc artifacts differ a little bit from build to build, and occasionally are significantly larger in size.

We have not integrated this into a full build just yet.  I'll let you know if we see problems there.

[rskdbl]

jindal.nipun‌ and team,

We have put our app through extensive testing and are very pleased with the results from this patch.  We have found no irregularities between builds.

When should we expect to see this fix as part of an SDK release?

Thanks!

Russell