Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!gatech!purdue!decwrl!shelby!unix!hplabs!hpfcdc!rml
From: rml@hpfcdc.HP.COM (Bob Lenk)
Newsgroups: comp.sys.hp
Subject: Re: HP-UX type declaration problem with signal(2)
Message-ID: <5570228@hpfcdc.HP.COM>
Date: 29 Jun 89 02:38:59 GMT
References: <218@bnrunix.UUCP>
Organization: HP Ft. Collins, Co.
Lines: 34

> 6.2---
> extern  (*signal())();
> 6.5---
> extern  void (*signal())();
> 
> Should not 6.5 have a statement more like 6.2?????

The only difference is between the declarations is whether the signal
handler returns an integer or returns no value.  Throughout UN*X
history, signal handlers have never actually returned a value (or if
they did it was discarded).  However, the original C language did not
support declarations of functions that did not return a value, but more
recent versions support the type void for this purpose.  Thus the newer
declaration is technically more correct.  The only change this requires
in user is changing from declarations from (possibly implicit) int to
void.  Depending on the specifics, failures to do this may generate
compiler warnings, compiler errors, or neither.  Any code that compiles
will run without problem (as will existing binaries).

The reason that HP-UX changed the declaration of signal() in 6.5 is to
conform to the SVID (System V Interface Definition) Base System
Definition Addendum, published in 1986 as Volume III of Issue 2 of the
SVID.  The previous SVID, to which 6.2 conformed, specified the int
declaration.

As a previous response alludes, the (proposed) ANSI C Standard specifies
the void type, and specifies signal() with this declaration.  Although,
6.5 does not conform to this standard, these changes are consistent with
moving toward it.

This change should probably have been mentioned in the Update_info file,
but was not.

		Bob Lenk (rml@hpfcla)