ARM64_32 Debug Mode #7012

lewurm · 2019-09-17T11:36:01Z

spouliot

set do-not-merge label since it depends on another PR, remove it once it's merged

rolfbjarne

You're targeting themono-2019-08 branch, but we've already merged that branch into master (so you can target master directly with your PR`).

rolfbjarne · 2019-09-17T11:47:38Z

tools/mtouch/mtouch.cs

@@ -372,7 +372,12 @@ public static string GetAotCompiler (Application app, bool is64bits)
 					return Path.Combine (cross_prefix, "bin", "arm-darwin-mono-sgen");
 				}
 			case ApplePlatform.WatchOS:
-				return Path.Combine (cross_prefix, "bin", "armv7k-unknown-darwin-mono-sgen");
+				/* Use arm64_32 cross only for Debug mode */
+				if (app.IsArchEnabled (Abi.ARM64_32) && !app.EnableLLVMOnlyBitCode) {


This is wrong: when building for both armv7k and arm64_32 this if condition will be true when building for either architecture, not only arm64_32.

The function doesn't have the information it needs, it needs to be passed the Abi as wells.

Ahh, thanks! Right, because MtouchArch can be set to ARMv7k, ARM64_32. However MtouchExtraArgs cannot be set separately; --interpreter will not work for ARMv7k.

In terms of user experience, any suggestion what we should do here? Maybe do not require the user to set --interpreter explicitly in MtouchExtraArgs, but force it in mtouch when ARM64_32 + Debug Mode is detected?

force it in mtouch when ARM64_32 + Debug Mode is detected

I think this would be the best option for customers, but I think we'd want to emit a warning in this case explaining what we're doing (I don't like silently changing behavior), and explain that it's possible to get the normal AOT compiler by disabling Debug mode in the Debug configuration (I'm assuming that actually works).

That said, this won't be a problem if customers have device-specific builds turned on (which iirc is on by default in new projects).

"normal" AOT aka. FullAOT for arm64_32 doesn't work currently, hence the interpreter. It would only be needed for Debug Mode (even when the managed debugger is disabled), but since we can use the interpreter there, I'm not sure if we will ever need to do the work to support FullAOT.

"normal" AOT aka.

Right now people can use Debug mode and it works, except that you can't debug. My concern is that if someone runs into a bug in the interpreter, they should be able to stop using the interpreter somehow so that the app runs correctly on the watch, even though debugging doesn't work (they could at least use Console.WriteLines).

No, FullAOT does not work at all for arm64_32 currently 😕

monojenkins · 2019-09-17T14:50:46Z

Build success
✅ Build succeeded
✅ API Diff (from stable)
✅ API Diff (from PR only) (no change)
✅ Generator Diff (no change)
✅ Test run succeeded

monojenkins · 2019-09-17T15:15:11Z

Build failure
❌ Build failed or was aborted

✅ Build succeeded
🔥 Failed to create API Diff 🔥

monojenkins · 2019-09-17T17:41:19Z

Build success
✅ Build succeeded
✅ API Diff (from stable)
✅ API Diff (from PR only) (no change)
✅ Generator Diff (no change)
✅ Test run succeeded

lewurm · 2019-09-17T20:13:34Z

tools/mtouch/mtouch.cs

+			} else {
+				if (app.Platform == ApplePlatform.WatchOS && app.IsArchEnabled (Abi.ARM64_32)) {
+					if (app.IsArchEnabled (Abi.ARMv7k)) {
+						throw ErrorHelper.CreateError (99, "Please use device builds on WatchOS.");


I didn't manage to disable device-specific builds, but I believe this would be the condition.

Also: To what does the error code map to? Is there some guideline?

mtouch error codes are listed and documented here: https://github.com/xamarin/xamarin-macios/blob/master/docs/website/mtouch-errors.md#mt0142-cannot-find-the-assembly-assembly-passed-as-an-argument-to---interpreter

it looks like the next available code would be 143.

Also you're not checking for debug mode here, so this will trigger for release builds as well.

lewurm · 2019-09-17T20:13:54Z

tools/mtouch/mtouch.cs

+					if (app.IsArchEnabled (Abi.ARMv7k)) {
+						throw ErrorHelper.CreateError (99, "Please use device builds on WatchOS.");
+					} else {
+						ErrorHelper.Warning (99, "ARM64_32 Debug mode requires --interpreter[=all], forcing it.");


To what does the warning code map to? Is there some guideline?

monojenkins · 2019-09-18T00:47:07Z

Build success
✅ Build succeeded
✅ API Diff (from stable)
✅ API Diff (from PR only) (no change)
✅ Generator Diff (no change)
✅ Test run succeeded

monojenkins · 2019-09-23T20:44:03Z

Build failure
✅ Build succeeded
✅ API Diff (from stable)
🔥 Failed to compare API and create generator diff 🔥
Failed to update apidiff references
Search for Comparing API & creating generator diff in the log to view the complete log.
🔥 Test run failed 🔥

…documentation

monojenkins · 2019-09-24T06:35:30Z

Build failure
❌ Build failed or was aborted

🔥 Build failed 🔥

rolfbjarne · 2019-09-24T11:38:41Z

build

monojenkins · 2019-09-24T14:58:45Z

Build success
✅ Build succeeded
✅ API Diff (from stable)
✅ API Diff (from PR only) (no change)
✅ Generator Diff (no change)
✅ Test run succeeded

lewurm · 2019-09-24T15:28:07Z

Leaving a papertrail here:

Depends on mono/mono#16886

This is in xamarin-macios/master now, it went in with this bump: #7044

lewurm requested a review from rolfbjarne September 17, 2019 11:36

lewurm requested a review from spouliot as a code owner September 17, 2019 11:36

spouliot added the do-not-merge Do not merge this pull request label Sep 17, 2019

spouliot approved these changes Sep 17, 2019

View reviewed changes

rolfbjarne requested changes Sep 17, 2019

View reviewed changes

lewurm changed the base branch from mono-2019-08 to master September 17, 2019 15:08

lewurm force-pushed the watch4-debug-mode branch from ace9798 to d55dc20 Compare September 17, 2019 15:13

lewurm changed the title ~~[mono-2019-08] ARM64_32 Debug Mode~~ ARM64_32 Debug Mode Sep 17, 2019

dalexsoto approved these changes Sep 17, 2019

View reviewed changes

lewurm force-pushed the watch4-debug-mode branch from d55dc20 to f9dc6ea Compare September 17, 2019 20:07

lewurm commented Sep 17, 2019

View reviewed changes

lewurm force-pushed the watch4-debug-mode branch from f9dc6ea to b2f901d Compare September 17, 2019 20:14

lewurm requested a review from rolfbjarne September 23, 2019 19:57

lewurm added 3 commits September 24, 2019 07:55

[builds] add arm64_32 cross compiler

6d8fe12

[mtouch] force --interpreter on arm64_32 debug mode build

5630cb9

[mtouch] include debug check for arm64_32 and reflect error codes in …

c373abe

…documentation

lewurm force-pushed the watch4-debug-mode branch from 75eced9 to c373abe Compare September 24, 2019 05:57

rolfbjarne approved these changes Sep 24, 2019

View reviewed changes

lewurm removed the do-not-merge Do not merge this pull request label Sep 24, 2019

rolfbjarne merged commit 26ec282 into xamarin:master Sep 24, 2019

rolfbjarne mentioned this pull request Nov 12, 2019

Debugger doesn't connect to watchOS app deployed to real device (Apple Watch Series 5) #7276

Closed

rolfbjarne added the notes-mention Deserves a mention in release notes label Sep 14, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ARM64_32 Debug Mode #7012

ARM64_32 Debug Mode #7012

lewurm commented Sep 17, 2019

spouliot left a comment

rolfbjarne left a comment

rolfbjarne Sep 17, 2019

lewurm Sep 17, 2019

rolfbjarne Sep 17, 2019

lewurm Sep 17, 2019 •

edited

rolfbjarne Sep 18, 2019

lewurm Sep 19, 2019

monojenkins commented Sep 17, 2019

monojenkins commented Sep 17, 2019

monojenkins commented Sep 17, 2019

lewurm Sep 17, 2019 •

edited

rolfbjarne Sep 18, 2019

lewurm Sep 17, 2019

monojenkins commented Sep 18, 2019

monojenkins commented Sep 23, 2019

monojenkins commented Sep 24, 2019

rolfbjarne commented Sep 24, 2019

monojenkins commented Sep 24, 2019

lewurm commented Sep 24, 2019

ARM64_32 Debug Mode #7012

ARM64_32 Debug Mode #7012

Conversation

lewurm commented Sep 17, 2019

spouliot left a comment

Choose a reason for hiding this comment

rolfbjarne left a comment

Choose a reason for hiding this comment

rolfbjarne Sep 17, 2019

Choose a reason for hiding this comment

lewurm Sep 17, 2019

Choose a reason for hiding this comment

rolfbjarne Sep 17, 2019

Choose a reason for hiding this comment

lewurm Sep 17, 2019 • edited

Choose a reason for hiding this comment

rolfbjarne Sep 18, 2019

Choose a reason for hiding this comment

lewurm Sep 19, 2019

Choose a reason for hiding this comment

monojenkins commented Sep 17, 2019

monojenkins commented Sep 17, 2019

monojenkins commented Sep 17, 2019

lewurm Sep 17, 2019 • edited

Choose a reason for hiding this comment

rolfbjarne Sep 18, 2019

Choose a reason for hiding this comment

lewurm Sep 17, 2019

Choose a reason for hiding this comment

monojenkins commented Sep 18, 2019

monojenkins commented Sep 23, 2019

monojenkins commented Sep 24, 2019

rolfbjarne commented Sep 24, 2019

monojenkins commented Sep 24, 2019

lewurm commented Sep 24, 2019

lewurm Sep 17, 2019 •

edited

lewurm Sep 17, 2019 •

edited