Re: OpenBSD: signal handling and strange behaviour.

> On 30 Jan 2018, at 20:14, Otto Moerbeek <otto@drijf.net> wrote:
>
> On Tue, Jan 30, 2018 at 07:57:14PM +0000, Al Poole wrote:
>
>> Hi,
>>
>> I've been having some issues with select/sigprocmask and signals on
>> OpenBSD 6.2 (amd64).
>>
>> Since 6.0, I've been working on EFL (enlightenment.org) on FreeBSD and
>> OpenBSD. OpenBSD broke sometime after the 6.0 release (in the mainloop
>> SIGCHLD were getting lost and applications would hang).
>>
>> I'm hoping this test-case might also explain the issues with EFL.
>>
>> The test code URL:
>>
>> http://haxlab.org/stuff/sig.c.txt
>>
>> EXPECTED OUTPUT
>>
>> # identical on Linux, FreeBSD, MacOS and Windows POSIX emulation.
>>
>> $ ./a.out
>> Timeout
>> Timeout
>> Timeout
>> buf: Yo
>>
>> Timeout
>> Timeout
>> Timeout
>> buf: Yo
>>
>> Timeout
>> Timeout
>> Timeout
>> buf: Yo
>>
>> Child exits!
>> Exiting!
>>
>> BELOW THE OUTPUT ON OPENBSD 6.2 (amd64):
>>
>> $ ./a.out
>> ^Cbuf: Yo
>>
>>
>> ^Cbuf: Yo
>>
>> ^CExiting!
>>
>> The code should handle SIGINT (CTRL^C) gracefully, but I noticed
>> running the testcase on OpenBSD without interrupting (SIGINT) the
>> behaviour is so wildly different from testing on all the other major
>> UNIX and Unix-like systems.
>>
>> The OpenBSD handling of SIGINT doesn't work either in this case.
>>
>> I wonder if anyone could help here? Whether I am doing something
>> wrong, or maybe even finding the OpenBSD EFL main loop signal issue?
>>
>> Thanks very much,
>>
>> Al Poole
>
> This is what I see, tested on armv7 and amd64, without pressing ^C. It
> is not clear from your description if you are doing that during the
> runs on the "working" systems.
>
> $ cc -O2 x.c
> $ ./a.out
> Timeout
> Timeout
> Timeout
> buf: Yo
>
> Timeout
> Timeout
> Timeout
> buf: Yo
>
> Timeout
> Timeout
> Timeout
> buf: Yo
>
> Child exits!
> Exiting!
>
> So the expected output. How do you compile/run your program
> *exactly*?
>
> -Otto

Hi Otto,

Thanks for the reply. Yes sorry that was slightly ambiguous.

I'm glad to see the above.

Running on 6.2 GENERIC#2 amd64

Compiling -O0 and -O2 same behaviours.

I'm testing on a single-core virtualised machine on linode's platform.
Without interrupting (CTRL^C) the output is:

$ ./a.out
Timeout
Timeout
Timeout
buf: Yo
…5 minutes into execution.
^Cbuf: Yo

^Cbuf: Yo
Exiting!
Child exits!