64-bit transition part 11: ksh93 nval (#964)

This is the eleventh of the thickfold patch series, which enables
ksh93 to operate within a 64-bit address space.

The parts of ksh93 affected by this commit are:
- Various pieces of code wrongly assuming char is signed have been
  fixed (follow-up to #962).
- The nval system, string.c, array handling and waitevent.c. This
  is probably one of the most consequential changes, the effect
  being that ksh is now capable of storing variables larger than
  UINT_MAX.

Conspicuous changes of note:
- With the expansion of np->nvsize to a size_t (which can be either
  64-bit or 32-bit), the void* padding variable was a necessary
  addition for alignment purposes, so as to silence UBSan errors
  and prevent untimely memory faults.
  - (This part by Martijn Dekker) The NV_MINSZ and nv_namptr macros
    in nval.h have been corrected to properly align the value data
    that follows a static Namval_t header (minus its nvlink member)
    in a portable manner, i.e., according to max_align_t's
    alignment requirement (re: 01abbc6). This fixes longstanding
    incorrect alignment logic that happened to work on most
    platforms but broke if the layout of the Namval struct
    (Namval_t) was changed, causing undefined behaviour mainly in
    nvtype.c. This change should finally make it safe to apply
    future edits to Namval_t (except that nvlink and nvname must
    remain unchanged).
- An -Wunreachable-code-return warning at the end of nv_create()
  has been fixed.
- nv_setsize() as exposed in the public nval API now requires an
  argument of exactly (size_t)-1 to function like the nv_size()
  macro.
- nv_create(): ported over an old fix from my expand-nvflags branch
  to allow usage of strlen without downcasting the result to an
  int. This fix adds a uint64_t nvflags variable for storing flags,
  which introduces some inconsequential warnings. Fixing/avoiding
  those is out of the scope of this project; that's better suited
  for a future revision of expand-nvflags.
- Removed the unused and inessential attsize and attval struct
  members (these are remnants of the A__z botch, re: f215b10).

Change in the number of warnings on Linux when compiling with clang
using -Wsign-compare -Wshorten-64-to-32 -Wsign-conversion
-Wimplicit-int-conversion: 802 => 481 => 37 (progression from part
10 => part 11 => part 13)

Progresses #592
Co-authored-by: Martijn Dekker <martijn@inlv.org>

Author: JohnoKing

src/cmd/ksh93/include/defs.h

@@ -112,7 +112,7 @@ extern void		*sh_arithcomp(char*);
 extern pid_t 		sh_fork(int,int*);
 extern pid_t		_sh_fork(pid_t, int ,int*);
 extern void		sh_invalidate_ifs(void);
-extern char 		*sh_mactrim(char*,int);
+extern char 		*sh_mactrim(char*,int8_t);
 extern int 		sh_macexpand(struct argnod*,struct argnod**,int);
 extern int		sh_macfun(const char*,ptrdiff_t);
 extern void 		sh_machere(Sfio_t*, Sfio_t*, char*);

src/cmd/ksh93/include/name.h

@@ -57,7 +57,7 @@ struct Namref
 	Dt_t		*root;
 	char		*sub;
 #if SHOPT_FIXEDARRAY
-	int		curi;
+	ssize_t		curi;
 	char		dim;
 #endif /* SHOPT_FIXEDARRAY */
 };
@@ -143,12 +143,12 @@ struct Ufunction
 #define array_elem(ap)	((ap)->nelem&ARRAY_MASK)
 #define array_assoc(ap)	((ap)->fun)
 
-extern int		array_maxindex(Namval_t*);
+extern ssize_t		array_maxindex(Namval_t*);
 extern char 		*nv_endsubscript(Namval_t*, char*, int);
 extern Namfun_t 	*nv_enforcedisc(Namval_t*);
 extern int		nv_arrayisset(Namval_t*, Namarr_t*);
 extern int		nv_arraysettype(Namval_t*, Namval_t*,const char*,int);
-extern int		nv_aimax(Namval_t*);
+extern ssize_t		nv_aimax(Namval_t*);
 extern int		nv_atypeindex(Namval_t*, const char*);
 extern void		nv_setlist(struct argnod*, int, Namval_t*);
 #if SHOPT_OPTIMIZE
@@ -162,7 +162,7 @@ extern void		nv_setlist(struct argnod*, int, Namval_t*);
 #   define nv_setoptimize(argaddr)	/* no-op */
 #   define nv_getoptimize()		NULL
 #endif /* SHOPT_OPTIMIZE */
-extern void		nv_outname(Sfio_t*,char*, int);
+extern void		nv_outname(Sfio_t*,char*, ptrdiff_t);
 extern void 		nv_unref(Namval_t*);
 extern int		nv_hasget(Namval_t*);
 extern void		clone_all_disc(Namval_t*, Namval_t*, int);
@@ -186,7 +186,7 @@ extern int		nv_istable(Namval_t*);
 extern size_t		nv_datasize(Namval_t*, size_t*);
 extern Namfun_t		*nv_mapchar(Namval_t*, const char*);
 #if SHOPT_FIXEDARRAY
-   extern int		nv_arrfixed(Namval_t*, Sfio_t*, int, char*);
+   extern ssize_t	nv_arrfixed(Namval_t*, Sfio_t*, int, char*);
 #endif /* SHOPT_FIXEDARRAY */
 
 extern const Namdisc_t	RESTRICTED_disc;

src/cmd/ksh93/include/national.h

@@ -32,7 +32,7 @@
 #   endif
 #endif /* SHOPT_MULTIBYTE */
 
-extern int sh_strchr(const char*,const char*);
-extern int sh_strwidth(const char*);
+extern ptrdiff_t sh_strchr(const char*,const char*);
+extern size_t sh_strwidth(const char*);
 
 #endif /* !_NATIONAL_H */

src/cmd/ksh93/include/nval.h

@@ -67,7 +67,7 @@ struct Namfun
 	const Namdisc_t	*disc;
 	char		nofree;
 	unsigned int	subshell;
-	uint32_t	dsize;
+	size_t		dsize;
 	Namfun_t	*next;
 	char		*last;
 	Namval_t	*type;
@@ -76,7 +76,7 @@ struct Namfun
 struct Nambfun
 {
 	Namfun_t        fun;
-	int		num;
+	ssize_t		num;
 	const char	**bnames;
 	Namval_t	*bltins[1];
 };
@@ -103,33 +103,35 @@ struct Namdecl
 /* attributes of name-value node attribute flags */
 
 #define NV_DEFAULT 0
-/* This defines the attributes for an attributed name-value pair node */
+
+/*
+ * This defines the attributes for an attributed name-value pair node.
+ *
+ * NOTE: For NV_MINSZ and nv_namptr (below) to work correctly, nvlink
+ * and nvname must be the first two members, in that order.
+*/
 struct Namval
 {
 	Dtlink_t	nvlink;		/* space for cdt links */
 	char		*nvname;	/* pointer to name of the node */
-#if _ast_sizeof_pointer == 8
-#   if _ast_intswap > 0
-	unsigned short	nvflag; 	/* attributes */
-	unsigned short	pad1;
-#   else
-	unsigned short	pad1;
-	unsigned short	nvflag; 	/* attributes */
-#   endif
-	uint32_t  	nvsize;		/* size or base */
-#else
-	unsigned short	nvflag; 	/* attributes */
-	unsigned short 	nvsize;		/* size or base */
-#endif
+	uint16_t	nvflag; 	/* attributes */
+	size_t  	nvsize;		/* size or base */
 	Namfun_t	*nvfun;		/* pointer to trap functions */
 	void		*nvalue;	/* pointer to any kind of value */
 	void		*nvmeta;	/* pointer to any of various kinds of type-dependent data */
 };
 
 #define NV_CLASS	".sh.type"
 #define NV_DATA		"_"	/* special class or instance variable */
-#define NV_MINSZ	(sizeof(struct Namval)-sizeof(Dtlink_t)-sizeof(char*))
-#define nv_namptr(p,n)	((Namval_t*)((char*)(p)+(n)*NV_MINSZ-sizeof(Dtlink_t)))
+
+/* For aligning and addressing Namval_t data headers or array members with the nvlink member chopped off */
+struct _Namval_align_
+{
+	char		dummy;
+	max_align_t	probe;
+};
+#define NV_MINSZ	roundof(sizeof(struct Namval) - offsetof(struct Namval, nvname), offsetof(struct _Namval_align_, probe))
+#define nv_namptr(p,n)	((Namval_t*)((char*)(p) + (n) * NV_MINSZ - offsetof(struct Namval, nvname)))
 
 /* The following attributes are for internal use */
 #define NV_NOFREE	0x200	/* don't free the space when releasing value */
@@ -267,7 +269,7 @@ extern int		nv_isnull(Namval_t*);
 extern Namfun_t		*nv_isvtree(Namval_t*);
 extern Namval_t		*nv_lastdict(void);
 extern Namval_t		*nv_mkinttype(char*, size_t, int, const char*, Namdisc_t*);
-extern void 		nv_newattr(Namval_t*,unsigned,int);
+extern void 		nv_newattr(Namval_t*,unsigned,ssize_t);
 extern void 		nv_newtype(Namval_t*);
 extern Namval_t		*nv_open(const char*,Dt_t*,int);
 extern void 		nv_putval(Namval_t*,const char*,int);
@@ -280,7 +282,7 @@ extern void		nv_setref(Namval_t*, Dt_t*,int);
 extern int		nv_settype(Namval_t*, Namval_t*, int);
 extern void 		nv_setvec(Namval_t*,int,int,char*[]);
 extern void		nv_setvtree(Namval_t*);
-extern int 		nv_setsize(Namval_t*,int);
+extern size_t 		nv_setsize(Namval_t*,size_t);
 extern Namfun_t		*nv_disc(Namval_t*,Namfun_t*,int);
 extern void 		nv_unset(Namval_t*,int);
 extern Namval_t		*nv_search(const char *, Dt_t*, int);
@@ -289,7 +291,7 @@ extern Namval_t		*nv_type(Namval_t*);
 extern void		nv_addtype(Namval_t*,const char*, Optdisc_t*, size_t);
 extern const Namdisc_t	*nv_discfun(int);
 
-#define nv_size(np)		nv_setsize((np),-1)
+#define nv_size(np)		nv_setsize((np),(size_t)-1)
 #define nv_stack(np,nf)		nv_disc(np,nf,0)
 
 #endif /* !_NVAL_H */

src/cmd/ksh93/include/version.h

@@ -21,7 +21,7 @@
 #include <ast_release.h>
 #include "git.h"
 
-#define SH_RELEASE_DATE	"2026-04-02"	/* must be in this format for $((.sh.version)) */
+#define SH_RELEASE_DATE	"2026-04-05"	/* must be in this format for $((.sh.version)) */
 /*
  * This comment keeps SH_RELEASE_DATE a few lines away from SH_RELEASE_SVER to avoid
  * merge conflicts when cherry-picking dev branch commits onto a release branch.

src/cmd/ksh93/nval.3

@@ -25,7 +25,7 @@ Namdisc_t;
 .ft 5
 Namval_t 	*nv_open(const char *\fIname\fP, Dt_t *\fIdict\fP, int \fIflags\fP);
 Namval_t	*nv_create(const char *\fIname\fP,  Dt_t *\fIdict\fP, int \fIflags\fP, Namfun_t *\fIfp\fP);
-Namval_t	*nv_namptr(void *\fIptr\fP, int \fIindx\fP);
+Namval_t	*nv_namptr(void *\fIptr\fP, size_t \fIindx\fP);
 void		nv_delete(Namval_t *\fInp\fP, Dt_t *\fIdict\fP, int \fInofree\fP);
 .ft R
 .fi
@@ -44,13 +44,13 @@ int		nv_clone(Namval_t *\fIsrc\fP, Namval_t *\fIdest\fP, int \fIflags\fP);
 .nf
 .ft 5
 int		nv_isnull(Namval_t *\fInp\fP);
-int		nv_setsize(Namval_t *\fInp\fP, int \fIsize\fP);
-int		nv_size(Namval_t *\fInp\fP);
+size_t		nv_setsize(Namval_t *\fInp\fP, size_t \fIsize\fP);
+size_t		nv_size(Namval_t *\fInp\fP);
 unsigned	nv_isattr(Namval_t *\fInp\fP, unsigned \fIflags\fP);
 Namfun_t	*nv_isvtree(Namval_t *\fInp\fP);
 unsigned	nv_onattr(Namval_t *\fInp\fP, unsigned \fIflags\fP);
 unsigned	nv_offattr(Namval_t *\fInp\fP, unsigned \fIflags\fP);
-void		nv_newattr(Namval_t *\fInp\fP, unsigned \fIflags\fP, int \fIsize\fP);
+void		nv_newattr(Namval_t *\fInp\fP, unsigned \fIflags\fP, ssize_t \fIsize\fP);
 .ft R
 .fi
 
@@ -206,8 +206,8 @@ justified variables, the arithmetic base for integer variables,
 and the precision or number of places after the decimal point
 for floating point variables.  If \fIsize\fP is greater than or
 equal to zero, the current size is changed to this value.
-The \f3nv_size()\fP function is equivalent to \f3nv_setsize()\fP
-with the second argument negative.
+The \f3nv_setsize()\fP function is equivalent to \f3nv_size()\fP
+when the second argument is equal to -1.
 .PP
 The \f3nv_getval()\fP function returns the value of the given
 name-value pair as a string.  A \f3NULL\fP return value indicates