In-Game: array_copy() when "src_index" is invalid will silently crash the game, should show an error

[gm-bug-reporter]

Description

it should either crash with an error message or have missing entries filled with zeros, like it is the case for using an out of bounds index for dest_index.

Steps To Reproduce

var src_array = [ 1, 2, 3 ]

var dest_array = [ 4, 5, 6 ]
array_copy(dest_array, 4, src_array, 1, 1)
show_debug_message($"dest_ind out of bounds, filled with 0's {dest_array}")

var dest_array = [ 4, 5, 6 ]
array_copy(dest_array, 1, src_array, 4, 1)
show_debug_message($"src_index out of bounds, will not get printed {dest_array}")

Which version of GameMaker are you reporting this issue for?

IDE v2024.8.1.171 Runtime v2024.8.1.218

Which operating system(s) are you seeing the problem on?

Windows 10.0.19045.0

Which platform(s) are you seeing the problem on?

Windows

2cfe6c82-021f-460d-af68-629cc27de693

[DiasFranciscoA]

The behavior you're observing is expected. When the dest_index is out of bounds, the destination array is expanded, and the new elements are filled with zeros. This happens because we're writing to indices that don't exist, which triggers the array to grow. While this behavior might seem unusual, it is consistent with legacy design.

However, the scenario you're describing involves an out-of-bounds src_index—meaning you’re trying to read from an invalid index in the source array. In this case, we don’t expand the array or fill it with zeros because accessing a non-existent index should not alter the source array. Instead, this scenario should result in an error or be safely ignored. Expanding the source array to accommodate an invalid read would not be the correct behavior.

[DiasFranciscoA]

They were indeed silent crashes are these are NOT intended nor should happen. Those were fixed and to mitigate bad usage this extra information should be add to the manual explaining those edge cases:

// array_copy(dest, destIdx, src, srcIdx, length) → Undefined
// 
// Imagine the copy process as the following:
//	* srcIndex - the index where the copy should start (the copy cursor)
//	* abs(length) - is the amount to copy (the amount the cursor will move) 
//	* sign(length) - is the copy direction (the direction in which the cursor moves)
// 
// NOTES (regarding length):
//	* length always includes the srcIndex (if length != 0, the element at this index will always be copied)
//	  independently of the direction of the copy.
// 
// NOTES (regarding destIdx):
//	* negative offsets count from the last element this mean [-1] will NOT append to the array but rather replace the last element.
//	* negative offsets can be out-of-range will just become ZERO by default. 
//	* positive offsets can be out-of-range will expand the destination array (with zeros).
//

@YYBartT @gurpreetsinghmatharoo

test project for QA:

how_array_copy_should_work.zip

[YYBartT]

Clarified the limitations on the source and destination array ranges on the array_copy() manual page and added a third code example. Also updated array_equals().

[cameron-home]

@DiasFranciscoA So what was the fix? Do we have error messages returned when the user does this edgecase instead of a silent crash? Not sure what I'm expecting to see when I check your QA project + the original one.

[DiasFranciscoA]

there shouldn't be silent crashes... the edge case that the user explains if he is reading out of bounds won't produce an error since values will always be clampped to the available size... if writing out of bounds this is acceptable and the array will expand in size to accomodate the values... the other silent crashing situations have been fixed.
In general the array_copy function was silent crashing if arrays were read and write out of bounds which was incorrect, whatever the user does nothing should trigger a silent crash.

[cameron-home]

Reopening due to this failure in IDE v2024.1100.0.680 Runtime v2024.1100.0.702:

[DiasFranciscoA]

This is now fixed in the Nov24 branch with PR:
https://github.com/YoYoGames/GMS2-Runner-Main/pull/242

[cameron-home]

Verified fixed in IDE v2024.1100.0.683 Runtime v2024.1100.0.705, as no more crashes are occurring

[YYBartT]

Added a note about offset on the Array Functions page to mention reads on the source array cannot go out of bounds.