Deadloop in Windows, `os.sleep(-1)`

[litlighilit]

Description

when defined(windows):
  import std/os
  sleep(-1)  # <- deadloop

BTW, on POSIX, it just returns immediately, as sleep calls nanosleep, which exits with errno set as EINVAL for negative values.

Nim Version

At least from 1.6.14 to 2.1.1 and devel

Current Output

No response

Expected Output

No response

Possible Solution

Two direction:

1. raises a exception on all platforms when milsecs is negative.
2. add check in Windows, returns immediately for negative values

Additional Information

In Python, time.sleep raises ValueError on negative value.

Maybe we can raises OSError as Nim's sleep is in std/os if the first direction is accepted.

On the other hand, golang's time.Sleep will return immediately for negative values.

[litlighilit]

BTW, It seems to deadloop as os.sleep calls Sleep on Windows, whose parameter type, is in fact DWORD, an uint32, but Nim wraps it as int32, so pass negative int just becomes a large unsigned int.

Still a DWORD relative issue...

[Araq]

What's wrong with -1 meaning "forever"?

[litlighilit]

on posix, it returns immediately. (and set errno to EINVAL)

[litlighilit]

What's wrong with -1 meaning "forever"?

And strictly speaking, not forever, but a very large uint amount of time.

e.g. Currently on Windows: sleep(-1) in fact does Sleep(4294967295)

a.k.a. while sleep(-1) means sleep forever is somewhat reasonable, it means sleep for 49 days on Windows and sleep for no time on POSIX.

[juancarlospaco]

sleep should take Natural.