Re: adding getdelim(3) and getline(3) to libc

[Alan Barrett <apb%cequrux.com@localhost>]

On Sun, 12 Jul 2009, Roy Marples wrote:
> The only issue is that getline collides with quite a few userland bits  
> such as ctags and nawk. I propose we change their internal definitons to  
> get_line, unless the really implement their own getline(3) and we can  
> just remove it.

A couple of months ago, in response to a PR or a mailing list message
about inability to build NetBSD on a host system that ahs POSIX
getline(), I started experimenting with a semantic patch that was
intended to rename getline to mygetline in cases where the usage was
incompatible with POSIX.  It almost works but has some issues whose
details I forget.

I am unlikely to continue this work, but I append a copy of the spatch
file in case anybody else wants to pick it up.  "spatch" is part of the
devel/coccinelle package.  Usage is something like "spatch -sp_file
getline.spatch -dir /usr/src".

--apb (Alan Barrett)

// convert getline() to mygetline() in cases where the argument or return
// types do not match the glibc or POSIX 2008 definition of
// 
//      ssize_t getline(char ** lineptr, size_t *n, FILE * stream);
// 

@ good_decl @
position p;
parameter P1, P2, P3;
@@
ssize_t getline@p(P1, P2, P3);

@ bad_decl @
position p != good_decl.p;
type RETURNTYPE;
parameter list ARGS;
@@
  RETURNTYPE
- getline@p
+ mygetline
  (ARGS);

@ good_defn @
position p;
parameter P1, P2, P3;
@@
ssize_t getline@p(P1, P2, P3) { ... }

@ bad_defn @
position p != good_defn.p;
type RETURNTYPE;
parameter list ARGS;
@@
  RETURNTYPE
- getline@p
+ mygetline
  (ARGS) { ... }

@ good_use expression @
position p;
expression E1, E2, E3;
@@
getline@p(E1, E2, E3)

@ bad_use expression @
position p != good_use.p;
expression list ARGS;
@@
- getline@p
+ mygetline
  (ARGS)