fix(ios): check for selector availability

[vijaysingh-axway]

https://jira.appcelerator.org/browse/TIMOB-28148

[build]

	Fails
🚫	Tests have failed, see below for more information.

	Warnings
⚠️	tests/Resources/ti.ui.tableview.test.js#L1347 - tests/Resources/ti.ui.tableview.test.js line 1347 – 'value' is assigned a value but never used. (no-unused-vars)

	Messages
📖	💾 Here's the generated SDK zipfile.
📖	✊ The commits in this PR match our conventions! Feel free to Rebase and Merge this PR when ready.
📖	❌ 1 tests have failed There are 1 tests failing and 984 skipped out of 10904 total tests.

Tests:

Classname	Name	Time	Error
ios.iphone.Titanium.UI.iOS.CollisionBehavior.example	works (14.0)	15	Error: timeout of 15000ms exceeded file:///Users/build/Library/Developer/CoreSimulator/Devices/0BA89FD6-590B-4E1C-A92C-C7071914B3F8/data/Containers/Bundle/Application/12106AD2-A890-4995-A13B-4E54B7ECC13D/mocha.app/ti-mocha.js:4326:27

Generated by 🚫 dangerJS against 6d1ffc9

[hansemannn]

I'm pretty sure this does not fix the real issue. Primarily, because this used to work before and the error suggests that a wrong instance is bubbled up in the event

[janvennemann]

@vijaysingh-axway I think Hans is right here. I did a quick debug session and it appears that self.view is holding a wrong value. Xcode debugger says it's nil so it may have been accidentally released and now holds an invalid pointer. Can you please double check what's going on?

[janvennemann]

See comment above.

[sgtcoolguy]

Looking at diffs between 9.0.3 and 9.1.0, this change related to TIMOB-27935 looks suspicious: b15d184

I think this stems because now a row proxy returns the TiUITableViewCell whereas before it returned nil.

The code generally assumes the view returned is a TiUIView, but in this case, it is not.

[vijaysingh-axway]

To fix an issue (kind of parity) https://jira.appcelerator.org/browse/TIMOB-27935, as Chris has mentioned, I have returned the cell view associated with row proxy which is not TiUIView. This is causing the crash. Initially it was nil so there was not issue. So-
In current fix I have checked for availability of selector then only call it. I think there should not be any related issue after this fix.

[sgtcoolguy]

So, this may be a quick and dirty fix for this issue, but it exposes the fact that the earlier fix broke the [TiViewProxy view] return type contract of returning a TiUIView*. It no longer returns a TiUIView*, it returns a subclass of UIView*. The better long-term "fix" here would be if we could separate callers that needed a TiUIView* from those needing a UIView*. i.e. the code around calculating the rect property actually only needs a UIView*, but the size property needs a TiUIView* (which shows that the earlier fix only fixed the rect property, not size!) - and then have different methods to return either?

Or maybe we can extract some common protocol used by callers and ensure TiUITableViewCell and TiUIView adhere to it?

[vijaysingh-axway]

I am thinking to revert the changes of PR #11758 and reopen https://jira.appcelerator.org/browse/TIMOB-27935 to properly fix.

[vijaysingh-axway]

Reopened https://jira.appcelerator.org/browse/TIMOB-27935 to look in this again.

[ssjsamir]

FR Passed: Tested with the following test case:

var tableData = [];
   
var win = Ti.UI.createWindow({ backgroundColor: 'white' });
 
var table = Ti.UI.createTableView({ objName: 'table' });
 
for (var i = 0; i <= 20; i++) {
var row = Ti.UI.createTableViewRow({ title: 'click me' });
 
var view = Ti.UI.createView({
layout: 'vertical',
height: Ti.UI.SIZE,
touchEnabled: false,
width: "50%",
left: 0
});
row.add(view)
tableData.push(row);
}
 
table.setData(tableData);
 
table.addEventListener('click', function (e) {
if (e.source && e.source.objName !== 'table') {
e.source.add(Ti.UI.createLabel({
text: 'REQUIRED',
left: '60%',
color: 'blue',
horizontalWrap: true,
width: "40%",
height: Ti.UI.SIZE
}));
 
}
});
 
win.add(table);
win.open();

Test Environment

MacOS Big Sur: 11.0 Beta 9
Xcode: 12.2 beta 2
Java Version: 1.8.0_242
Android NDK: 21.3.6528147
Node.js: 12.18.1
""NPM":"5.0.0","CLI":"8.1.1""
iphone 11 (ios 14)

[sgtcoolguy]

I get that the rollback off the other fix "fixes" this issue. But it also re-opens a different crash that affected a customer. We need to address that as well - we cannot merge this in and be done with it, because we're just trading one crash for another

Spoke about what this re-opens on Teams

[sgtcoolguy]

OK, I was mistaken. The fix for TIMOB-27935 caused this bug and TIMOB-28001 to occur - not that the fix for TIMOB-27935 also fixed TIMOB-28001
So, rolling back really only re-opens https://jira.appcelerator.org/browse/TIMOB-27935 which is an issue found internally and does not seem to be as severe as this bug or TIMOB-28001 (both crashes).