Re: Worrisome warnings after recent patches

Re: Worrisome warnings after recent patches

[Sven Wischnowsky]

Bart Schaefer wrote (in a mail directly to me):

> ../../../zsh-3.1.5/Src/Zle/zle_tricky.c: In function `join_strs':
> ../../../zsh-3.1.5/Src/Zle/zle_tricky.c:1803: warning: `mp' might be used uninitialized in this function
> ../../../zsh-3.1.5/Src/Zle/zle_tricky.c: In function `join_ends':
> ../../../zsh-3.1.5/Src/Zle/zle_tricky.c:1930: warning: `tn' might be used uninitialized in this function
>
> ... [ more of these warnings ]

The first one really was wrong, but in the other cases the variables
could not have been used uninitialized, partly due to the calling
context but in most cases due to the tests in the functions
themselves.

Bye
 Sven

diff -u os/Zle/zle_tricky.c Src/Zle/zle_tricky.c
--- os/Zle/zle_tricky.c	Wed Mar  3 17:22:42 1999
+++ Src/Zle/zle_tricky.c	Thu Mar  4 08:44:58 1999
@@ -1873,7 +1873,7 @@
 	    if (!t)
 		break;
 	} else {
-	    if (rr <= mp->llen) {
+	    if (rr <= 1) {
 		char *or = rs;
 
 		rs = realloc(rs, (rl += 20));
@@ -1927,7 +1927,7 @@
 	} else {
 	    /* Different anchors, see if we can find matching anchors
 	     * further down the lists. */
-	    Cline to, tn;
+	    Cline to, tn = NULL;
 	    int t = 0;
 
 	    /* But first build the common prefix. */
@@ -3414,6 +3414,7 @@
 	}
     }
     while (ll && lw) {
+	t = 0;
 	/* First try the matchers. */
 	for (ms = mstack; ms; ms = ms->next) {
 	    for (mp = ms->matcher; mp; mp = mp->next) {
@@ -3655,6 +3656,7 @@
 	}
     }
     while (ll && lw) {
+	t = 0;
 	for (ms = mstack; ms; ms = ms->next) {
 	    for (mp = ms->matcher; mp; mp = mp->next) {
 		if (lm == mp)
@@ -3984,11 +3986,12 @@
 	   char *suf, char *group, char *rems, char *remf, char *ign,
 	   int flags, int aflags, Cmatcher match, char *exp, char **argv)
 {
-    char *s, *t, *e, *me, *ms, *lipre = NULL, *lpre, *lsuf, **aign = NULL;
-    int lpl, lsl, i, pl, sl, test, bpl, bsl, llpl, llsl;
-    Aminfo ai;
+    char *s, *t, *e, *me, *ms, *lipre = NULL, *lpre = NULL, *lsuf = NULL;
+    char **aign = NULL;
+    int lpl, lsl, i, pl, sl, test, bpl, bsl, llpl = 0, llsl = 0;
+    Aminfo ai = NULL;
     Cline lc = NULL;
-    LinkList l;
+    LinkList l = NULL;
     Cmatch cm;
     struct cmlist mst;
     Cmlist oms = mstack;
@@ -4132,7 +4135,7 @@
 	    }
 	    /* Walk through the matches given. */
 	    for (; (s = dupstring(*argv)); argv++) {
-		sl = strlen(s);
+		sl = pl = strlen(s);
 		lc = NULL;
 		ms = NULL;
 		bpl = brpl;
@@ -5742,7 +5745,7 @@
     Compcond or, cc;
     Comp comp;
     int compadd, m = 0, d = 0, t, tt, i, j, a, b;
-    char *sc, *s, *ss;
+    char *sc = NULL, *s, *ss;
 
     /* This loops over the patterns separated by `-'s. */
     for (compc = occ->ext; compc; compc = compc->next) {
@@ -7871,7 +7874,7 @@
 	    }
 	}
 	else if (g->lcount) {
-	    int n = g->lcount, nl = (n + ncols - 1) / ncols, nc = nl, i, j, a;
+	    int n = g->lcount, nl = (n + ncols - 1) / ncols, nc = nl, i, j, a = 0;
 	    Cmatch *q;
 
 	    if (n && pnl) {

--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de

0 vs. NULL (RE: Worrisome warnings after recent patches)

[Andrej Borsenkow]

> +	    Cline to, tn = NULL;

What is the point of using NULL to initialize null pointer. The only
portable and official way is to use `0'(zero), that is garanteed to be
converted to whatever representation null pointer has on a given system.

I mention it, because I just recently have seen quite a number of bug
reports, that rsulted exactly from using something like NULL for this.

Just a reminder ...

/andrej

Re: 0 vs. NULL (RE: Worrisome warnings after recent patches)

[Bernd Eggink]

Andrej Borsenkow wrote:
> 
> > +         Cline to, tn = NULL;
> 
> What is the point of using NULL to initialize null pointer. The only
> portable and official way is to use `0'(zero), that is garanteed to be
> converted to whatever representation null pointer has on a given system.

No, this applies to C++ only, not to C. In C you should use the NULL
macro or (void*)0. 

Bernd

--
Bernd Eggink
Regionales Rechenzentrum der Uni Hamburg
eggink@rrz.uni-hamburg.de
http://www.rrz.uni-hamburg.de/eggink/BEggink.html

Re: 0 vs. NULL (RE: Worrisome warnings after recent patches)

[Bruce Stephens]

Bernd Eggink <eggink@uni-hamburg.de> writes:

> Andrej Borsenkow wrote:

> > What is the point of using NULL to initialize null pointer. The only
> > portable and official way is to use `0'(zero), that is garanteed to be
> > converted to whatever representation null pointer has on a given system.
> 
> No, this applies to C++ only, not to C. In C you should use the NULL
> macro or (void*)0.

There's an issue with arguments to functions which don't have
prototypes, but apart from that, the literal 0 as a pointer should be
fine in C.

Re: 0 vs. NULL (RE: Worrisome warnings after recent patches)

[Sven Wischnowsky]

Bruce Stephens wrote:

> Bernd Eggink <eggink@uni-hamburg.de> writes:
> 
> > Andrej Borsenkow wrote:
> 
> > > What is the point of using NULL to initialize null pointer. The only
> > > portable and official way is to use `0'(zero), that is garanteed to be
> > > converted to whatever representation null pointer has on a given system.
> > 
> > No, this applies to C++ only, not to C. In C you should use the NULL
> > macro or (void*)0.
> 
> There's an issue with arguments to functions which don't have
> prototypes, but apart from that, the literal 0 as a pointer should be
> fine in C.

Since the original message was a reaction to my patch: I know that `0' 
should be fine in C (at least with modern compilers) and personally I
prefer it. But using `NULL' is the convention used throughout the zsh
code, so...

Bye
 Sven


--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de