WONTFIX: select(2) in SUA 5.2 ignores timeout

With Windows Server 2003 R2, Microsoft incorporated Services for UNIX as a set of operating system components. The POSIX subsystem, Interix, is called the Subsystem for UNIX Applications (SUA) in Windows Server 2003 R2 and later.

Interix is the internal name of the Windows Posix Subsystem (PSXSS) that is based on OpenBSD and operates as an independent sister subsystem with the Windows subsystem (aka CSRSS or Client/Server Runtime Subsystem).

With the first version of SUA, aka Interix 5.2, Microsoft added a bunch of new UNIX APIs. Unfortunately the broke some things that were previously working in the previous edition which was called Interix 3.5 (aka Services for UNIX 3.5).

For example, select(2) is broken in SUA 5.2. It completely ignores the timeouts provided as arguments and returns immediately.

From the POSIX specification:

If the timeout parameter is not a null pointer, it specifies a maximum interval to wait for the selection to complete. If the specified time interval expires without any requested operation becoming ready, the function shall return. If the timeout parameter is a null pointer, then the call to pselect() or select() shall block indefinitely until at least one descriptor meets the specified criteria. To effect a poll, the timeout parameter should not be a null pointer, and should point to a zero-valued timespec structure.

Here is a little test program. What should happen is that select() should block for 10 seconds every time through the loop.

#include <stdio.h>;
#include <sys/time.h>

int main()
{
	printf("Testing select(2). Each pass through the loop should pause 10 seconds.\n\n");

	struct timeval time, pause;
	pause.tv_sec  = 10;
	pause.tv_usec = 0;
	int i;

	for( i=0; i&lt;10; i++ )
	{
		//insert a 10 second pause on every loop to test select().		
		gettimeofday(&amp;time, 0);
		printf(";Current time: %d\n", time.tv_sec);
		time.tv_sec += 10L;
		printf("Add 10 seconds: %d... And pause by calling select(2).\n", time.tv_sec);
		(void) select(0, 0, 0, 0, &pause);
	}
	return 0;
}

What actually happens is select() returns immediately.

% uname -svrm
Interix 5.2 SP-9.0.3790.4125 EM64T
% gcc selecttest.c -o selecttest 
% ./selecttest 
Testing select(2). Each pass through the loop should pause 10 seconds. 

Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2). 
Current time: 1142434664 
Add 10 seconds: 1142434674... And pause by calling select(2).
% 

The test run above should have taken 100 seconds but it actually completes in less than 1 second. This is a problem because many UNIX applications will use select() as a timing mechanism. Some will use select() as a timer even if they aren’t doing IO.

There is good news and bad news.

The bad news is that MSFT told me that the won’t fix this issue. Their official guidance is to use sleep(2) and usleep(2) to control timeouts in Interix 52.

The good news is that select(2) works properly on Interix 6.0 with Windows Vista and Windows Server 2008.