go: runtime: time is not correctly updated under wine

OMG, are you serious? You do allow proprietary binary crap in the executable and can not accept several lines diff which fixes real bug basically rendering system unusable in the most popular crossplatform environment on the world without shitty bureaucracy, which involves registering new google account, which virtually refuses to verify new user (probably because this phone number is already used in a completely irrelevant setup)?

Guys, are you aware of the 21 century auth technologies? OAuth maybe, did you hear this word? Do you know about github? Maybe words ‘signed-off-by’ highlight something - those beardy guys in linux-kernel use it. Or maybe code of conduct, facebook requires to sign a special page to accept patches…

What’s wrong with you? You create a good language for its purpose, and you add so much crap around it.

Someone, please copy this patch into review system or rewrite it the way you like, I’ve shown the basic idea, implementation fixes time bug in wine environment and works without overhead for common windows users.

Interrupt time precision is about 10 updates per microsecond, so calling something like code below at every start will not ever slow down any go program, getting that windows starting code already calls things like LoadLibrary()

It could be even better than what I wrote if using a systime()-like loop directly instead of calling systime() which synchronizes timer update with timer interrupt. There should be a new term for this kind of optimization though - a nanooptimization!

I would not mind calling this with the loop duration specified via some environment variable or anything else, which will be zero by default and only wine users would specify it. Although imo it looks a bit ugly like hell, it is up to the maintainers who believes what is best for the project.

// 1 usec at worst
func time_check() bool {
t1 := systime()
for i := 0; i<10; i++ {
    t2 := systime()
    if t2 != t1 {
        return true
    }
}
return false
}

osinit() {
...
if time_check() != true {
    set new time callback
}
}

But I believe that detecting that interrupt time is not updated rather than application runs under wine is a better idea - one problem, one local fix. Who knows, maybe wine will eventually port this memory feature, although so far there are only workarounds which use dedicated thread which updates this page once in 15.6 ms to mimic only system time.