-prod affects WaitingGroup

[radare]

V version: V 0.1.22 19b4cf6
OS: macOS 10.15

What did you do?

$ cat match.v
import sync
import time

fn sub(wg mut sync.WaitGroup) {
	time.sleep(1)
	wg.done()
}

fn main() {
	mut wg := sync.new_waitgroup()
	wg.add(3)
	go sub(&wg)
	go sub(&wg)
	go sub(&wg)
	wg.wait()
}

What did you expect to see?

$ v -prod match.v
$ time ./match
./match  0.99s user 0.01s system 99% cpu 1.014 total
$

What did you see instead?

when using -prod i get undefined behaviour (sometimes ends quickly, sometimes never ends) related to WaitingGroup. it seems to be more consistent without release builds

$ time ./match
^C
./match  0.00s user 0.00s system 0% cpu 14.333 total
$
$ v -prod match.v
$ time ./match
./match  0.00s user 0.00s system 56% cpu 0.005 total
$ time ./match
./match  0.00s user 0.00s system 0% cpu 8.208 total
$ v match.v
$ time ./match
./match  1.00s user 0.00s system 73% cpu 1.369 total
$

[spytheman]

It happens on linux too.

[spytheman]

Passing -O1 (or -O2 or -O3) to the C backend induces this bug. Not sure what the cause is yet.

[radare]

Found the bug.. the compiler is optimizing this empty loop. imho V should forbid empty loops.

pub fn (wg mut WaitGroup) wait() {
println('wait0')
        for wg.active > 0 {
println('wait1')
                // time.sleep(1)
                // waiting
        }
println('wait2')
}

if you uncomment this time.sleep(1) line then the bug appears actually you can add any other statement inside the for loop and the bug disapears.

[radare]

not to mention this infinite loop should be using a mutex

[dumblob]

@medvednikov wanted to completely reimplement go routines (see e.g. #1868 ), so I wouldn't expect the current implementation to behave nicely (but just correctly).

[medvednikov]

Fixed.

[radare]

thank you <3

…

On 3 Dec 2019, at 14:42, Alexander Medvednikov ***@***.***> wrote: Fixed. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2874?email_source=notifications&email_token=AAG75FVCYKQ6IRWKB2YEGWTQWZO4ZA5CNFSM4JQ76CSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEFZNFCA#issuecomment-561173128>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG75FRA74CHLP2BECYWYKDQWZO4ZANCNFSM4JQ76CSA>.