[pulseaudio-discuss] [PATCH] pa_poll(): Simplify detectin of invalid fds in select() emulation mode

[Daniel Mack]

On Thu, Apr 14, 2011 at 7:07 PM, Sean McNamara <smcnam at gmail.com> wrote:
> Hi,
>
> On Thu, Apr 14, 2011 at 9:57 AM, Daniel Mack <zonque at gmail.com> wrote:
>> On Thu, Apr 14, 2011 at 3:53 PM, Maarten Bosmans <mkbosmans at gmail.com> wrote:
>>> 2011/4/14 Daniel Mack <zonque at gmail.com>:
>>>> On Thu, Apr 14, 2011 at 12:46 AM, Maarten Bosmans <mkbosmans at gmail.com> wrote:
>>>>> return (fd >= 0) && (fcntl(fd, F_GETFL) != -1);
>>>>>
>>>>> For Windows, both fcntl and F_GETFL are not available.
>>>>
>>>> Sigh. Ok, thanks for the feedback. Attached is a new patch which keeps
>>>> things as they are for Windows but fixes the problem for OS X.
>>>
>>> That should do it, yes.
>>>
>>> +#else /* !OS_IS_WINDOWS */
>>> Minor: this is not exactly the symbol used in the #ifdef.
>>
>> Ok.
>>
>>> Anyway, isn't the #else only supposed to be used for OSX? In that case
>>> the comment should probably say so.
>>
>> No, it's actually really a workaround for Windows as we don't have
>> fcntl() there. (Just curious - the header file does exist, right? I
>> saw it being included by quite a lot of other source files in the PA
>> tree ...)
>
> The fcntl.h header *does* exist in MSVC (and in MingW...) but the core
> Microsoft C Runtime doesn't actually export the fcntl function. On
> Windows, fcntl header only exists to expose the O_* hexadecimal
> #defines for the flags for the open() call, which *is* supported.
>
> The only fcntl on Windows comes if you use an alternative C library,
> such as Cygwin's port of newlib or glibc. The native stack doesn't
> supply it.

Ok, thanks for expaining. Colin, can you pick the latest patch from
this thread then and apply? Or do you want me to push it and send a
pull request?


Thanks, everone.

Daniel