[Bug 1089] Are headers complete enough?

------- Additional Comments From [email protected]  2005-10-06 14:57 -------
To summarize:

The following headers are required by the source standard, but are not defined
in any underlying specification:

err.h
getopt.h
libintl.h
netinet/in.h
netinet/ip.h
pty.h
rpc/pmap_clnt.h
rpc/rpc.h
rpc/svc.h
security/pam_appl.h
sys/param.h
utmp.h
zlib.h


For some (most) of these, we have data definitons. Data defs don't include
prototypes (because func prototypes aren't data definitions).

However, the data defs do supply all the other relevant information needed for
the source spec header ... all of the typedefs and constants etc.

So, our choices seems to be:
1. Ignore this problem ... this is an ABI spec and you don't need to know about
 anyhting that isn't a data def.

2. Add a separate section to the spec detailing header files and their contents
(as POSIX does).

3. In every man page where we reference a header that isn't adequately described
elsewhere, add a description of what should be in the referenced header.

4. Add a note at the beginning of the (generated) set of pages that describe
interfaces that are defined by this spec saying "Where interface listed here
includes one or more system header files, in addition to the data definitions
listed in the previous section, a function prototype for the interface shall be
included."

5. Add the function prototypes to the data definitions. An argument could be
made that these are, in fact, data definitions, since it is possible to take the
address of a function, and what you get is a pointer to a function returning a
particular type, and taking particular arguments ...

Option 5 is the simplest to implement, followed by option 4.
Option 5 is the simplest for the reader.

Is option 5 acceptable???   
 
--    
Configure bugmail: http://bugs.linuxbase.org/userprefs.cgi?tab=email   
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.