fix: incorrect pointer value comparison

[ltzmaxwell]

This is a fix to #1569 . Thanks to @thehowl for identifying this issue and providing a thorough and insightful analysis.

Here is an further analysis based on #1569 :

package main

func main() {
	{
		b := []byte("ABCDEFGHIJKL")
		a := b
		println(&a[0] == &a[0])
	}
}

this comparison would be false too.

The root cause for this is the way pointer values is obtained and compared, e.g.

package main

func main() {
	c := []byte{'A'}
	a := c
	println(&c[0])
	println(&c[0] == &a[0])
}

in this code snippet, the address of the c[0], (&c[0]) is obtained from this:

    ev := fillValueTV(store, &av.List[ii]) // by reference

that is a reference to the element of the underlying list of arrayValue,

in case like this:

package main

func main() {
	{
		b := []byte("ABCDEFGHIJKL")
		a := b
		println(&a[0] == &a[0])
	}
}

the address is obtained by

	bv := &TypedValue{ // heap alloc, so need to compare value rather than pointer in isEql(), line 482
		T: DataByteType,
		V: DataByteValue{
			Base:     av,
			Index:    ii,
			ElemType: et,
		},
	}

that is a new allocated *TV, which implies the value of it would not be same within the first and second & operation.
So we should actually compare the concrete value rather than the pointers for this case.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 45.84%. Comparing base (b619351) to head (1bffd37).
Report is 122 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1601      +/-   ##
==========================================
- Coverage   55.80%   45.84%   -9.96%     
==========================================
  Files         436      460      +24     
  Lines       66168    69636    +3468     
==========================================
- Hits        36922    31924    -4998     
- Misses      26356    35052    +8696     
+ Partials     2890     2660     -230     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[deelawn]

Looks pretty good. I think we need to guard against invalid type comparisons.

[zivkovicmilos]

@ltzmaxwell
Can you please check @deelawn's comments on this PR? 🙏

[ltzmaxwell]

@ltzmaxwell Can you please check @deelawn's comments on this PR? 🙏

Updated 🙏 .

[deelawn]

@ltzmaxwell Can you please check @deelawn's comments on this PR? 🙏

Updated 🙏 .

I don't think that fixed it -- see this comment #1601 (comment)