weird assertion failure for __global shmap shared map[string]f64, when unrelated function is commented/uncommented

[spytheman]

V doctor:

V full version: V 0.4.8 8ffcc34.08c2019
OS: linux, Ubuntu 20.04.6 LTS
Processor: 4 cpus, 64bit, little endian, Intel(R) Core(TM) i3-3225 CPU @ 3.30GHz

getwd: /space/v/oo
vexe: /space/v/oo/v
vexe mtime: 2024-10-28 19:30:44

vroot: OK, value: /space/v/oo
VMODULES: OK, value: /home/delian/.vmodules
VTMP: OK, value: /tmp/v_1000

Git version: git version 2.46.0
Git vroot status: weekly.2024.44-4-gd79b8509-dirty
.git/config present: true

CC version: cc (Ubuntu 10.5.0-1ubuntu1~20.04) 10.5.0
thirdparty/tcc status: thirdparty-linux-amd64 0134e9b9

What did you do?
./v -g -o vdbg cmd/v && ./vdbg -enable-globals z_test.v

fn bug_when_uncommented() { numberfns['two'] = fn () int { return 2 } }

__global (
	shmap     shared map[string]f64
	numberfns map[string]fn () int
)

fn test_global_shared_map() {
	lock shmap {
		dump(shmap)
		shmap['one'] = 1.25
		shmap['two'] = -0.75
		shmap['three'] = 5
		dump(shmap)
		dump(shmap['one'])
		dump(shmap['two']) // changed by the top line to 0, should be -0.75
		dump(shmap['three']) // changed by the top line to 0, should be 5.0
	}
	x, y := rlock shmap {
		shmap['two'], shmap['one']
	}
	dump(x)
	dump(y)
	assert x == -0.75
	assert y == 1.25
}

What did you expect to see?

A test that runs without assertion failures

What did you see instead?

#0 21:30:28 ^ master /v/oo>v -enable-globals z_test.v
[z_test.v:10] shmap: &{}
[z_test.v:14] shmap: &{'one': 1.25}
[z_test.v:15] shmap['one']: 1.25
[z_test.v:16] shmap['two']: 0.0
[z_test.v:17] shmap['three']: 0.0
[z_test.v:22] x: 0.0
[z_test.v:23] y: 1.25
z_test.v:24: ✗ fn test_global_shared_map
   > assert x == f64(-0.75)
     Left value (len: 3): `0.0`
    Right value (len: 5): `-0.75`

#1 21:31:34 ^ master /v/oo>

Note also, that dump(shmap) only displays the first element, even though the map has several, which may be another bug.

Note

You can use the 👍 reaction to increase the issue's priority for developers.

Please note that only the 👍 reaction to the issue itself counts as a vote.
Other reactions and those to comments will not be taken into account.

_{Huly®: V_0.6-21153}

[spytheman]

If the first line is commented out, the test works with both tcc, clang and gcc .

The test is isolated from vlib/v/tests/init_global_test.v where it caused flaky behavior.

I think at some point, it did pass fine, and it may have regressed due to some recent change, but I have not yet bisected to see at what version.

[spytheman]

Here is a relevant diff between the good (commented line) and bad versions (uncommented line):

26579,26580c26593,26594
<               map_set(&shmap->val, &(string[]){_SLIT("two")}, &(f64[]) { -0.75 });
<               map_set(&shmap->val, &(string[]){_SLIT("three")}, &(f64[]) { 5 });
---
>               (*(f64*)map_get(ADDR(map, shmap->val), &(string[]){_SLIT("two")}, &(f64[]){ 0 })) = -0.75;
>               (*(f64*)map_get(ADDR(map, shmap->val), &(string[]){_SLIT("three")}, &(f64[]){ 0 })) = 5;
26596,26598c26610,26612