Zsh Mailing List Archive
Messages sorted by: Reverse Date, Date, Thread, Author

[PATCH] another fix for 'compset -q'



Another minor "improvement" of 'compset -q'
It just avoids the warning in the example below; you can't get
any useful completion either with or without the patch.

zsh% zsh -c 'for (( i = 0;<TAB>
5: compcore.c:1676: expecting 'x' at offset 6 of "i = 0"

The cursor 'x' is not in the current word "i = 0" but in the
next token. This indicates the function gotword() (lex.c)
resets lexflags to 0 too early. I guess the global variable
zlemetall (which is used in gotword()) need not count the 'x'
added at the cursor, because get_cmp_string() (zle_tricky.c)
also adds 'x' at the cursor by using addx() but does not
increment zlemetall.

It seems gotword() takes account of addedx by itself.

If zlemetall is decremented by 1 (= addedx), then the
wb and we returned by gotword() also decrease. So we
need the 2nd hunk.



diff --git a/Src/Zle/compcore.c b/Src/Zle/compcore.c
index f733e0ee5..fd415da89 100644
--- a/Src/Zle/compcore.c
+++ b/Src/Zle/compcore.c
@@ -1610,9 +1610,11 @@ set_comp_sep(void)
     inpush(dupstrspace(tmp), 0, NULL);
     zlemetaline = tmp;
     /*
-     * Length of temporary string, calculated above.
+     * tl is the length of temporary string, calculated above.
+     * It seems zlemetall need not include the 'x' added at the cursor.
+     * addedx is taken care of in function gotword() (lex.c).
      */
-    zlemetall = tl;
+    zlemetall = tl - addedx;
     strinbeg(0);
     noaliases = 1;
     do {
@@ -1668,8 +1670,8 @@ set_comp_sep(void)
 	    DPUTS(!p, "no current word in substr");
 	    got = 1;
 	    cur = countlinknodes(foo) - 1;  /* cur is 0 offset */
-	    swb = wb - 1 - dq - sq - dolq;
-	    swe = we - 1 - dq - sq - dolq;
+	    swb = wb - dq - sq - dolq;
+	    swe = we - dq - sq - dolq;
             sqq = lsq;
 	    soffs = zlemetacs - swb - css;
 	    DPUTS2(p[soffs] != 'x', "expecting 'x' at offset %d of \"%s\"",



Messages sorted by: Reverse Date, Date, Thread, Author