Re: SIGIO, siginfo

[yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)]

hi,

> On Jan 19, 12:57pm, yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi) wrote:
> -- Subject: Re: SIGIO, siginfo
> 
> | linux saves the descriptor number.
> | we can do the same if this interface is intended to be linux-compatibile.
> | it has a good side effect to remove an XXXSMP in kpsignal.
> 
> Ouch, this seems to me like asking for trouble. What are the semantics if
> a process does a dup2() to this file descriptor then, or closes it before
> a queued sigio is delivered?

iirc, linux just saves the descripter which was used for the fcntl
as a plain integer.  it won't be affected by dup or close.

> | i don't think SIGIO and its fcntl/ioctl are in POSIX.
> | SIGPOLL is in POSIX, but it's for STREAMS which we don't have.
> 
> This is historically true, but I think at this point most implementations
> consider them aliases. TOG's signal page does not associate it with streams
> but TOG's ioctl page does (since it include <stropts.h>).
> 
> | > http://en.wikipedia.org/wiki/SIGPOLL
> | 
> | which part of the page are you reffering to?
> | "On Linux, SIGIO is a synonym for SIGPOLL." ?
> 
> "On POSIX-compliant platforms, SIGPOLL is the signal thrown by computer
> programs when asynchronous I/O event occurs."

i guess that wikipedia isn't a right place to look. :)

anyway, i don't have a strong opinion about the SIGPOLL synonym.

> What should we do about the inconsistent direction?

it was due to simple bugs.  fixed.

YAMAMOTO Takashi

> 
> christos