Re: [bug-gawk] gawk character class bug

[Ed Morton]

Arnold - Thanks for the fix and for the quick turnaround!

 

    Ed.

---

From: "Aharon Robbins" <address@hidden>
To: address@hidden, address@hidden
Sent: Monday, March 11, 2013 3:44:20 PM
Subject: Re: [bug-gawk] gawk character class bug

Hi Ed. Re this:

> Date: Thu, 7 Mar 2013 18:00:01 +0000 (UTC)
> From: Ed Morton <address@hidden>
> To: address@hidden
> Subject: [bug-gawk] gawk character class bug
>
> There's a bug in gawk wrt issuing a warning for character classes being
> used in a bracket _expression_. I want to remove the first character
> in a string that's either a closing?? square bracket "]" or a space
> "[:space:]".In /usr/xpg4/bin/awk on Solaris I can do this:
>
> $ echo "a]b" | /usr/xpg4/bin/awk '{sub(/[][:space:]]/,"")}1'
> ab
>
> In gawk it works just fine too and does correctly understand that
> [:space:] is a character class BUT I get an unwarranted?warning
> message:
>
> $ echo "a]b" | gawk '{sub(/[][:space:]]/,"")}1' ?? ?? ?? ?? ?? ??
> gawk: cmd. line:1: warning: regexp component `[:space:]' should probably be `[[:space:]]'
> ab
>
> See the discussion at
> https://groups.google.com/forum/#!topic/comp.lang.awk/MGX4VKyuv0k ??for
> more info.
>
> Regards,
>
>  Ed Morton.

So, it is a (small) bug. The fix is below. This will is relative to the
gawk-4.0-stable branch. I will check in the fix and a new test in the
stable and master branches shortly.

Thanks,

Arnold
---------------------------------------------------
diff --git a/re.c b/re.c
index 711b53e..4c03177 100644
--- a/re.c
+++ b/re.c
@@ -564,8 +564,22 @@ again:
 
                 if (*sp == '[')
                         count++;
-                else if (*sp == ']')
-                        count--;
+                /*
+                 * ] as first char after open [ is skipped
+                 * \] is skipped
+                 * [^]] is skipped
+                 */
+                if (*sp == ']' && sp > sp2) {
+                         if (sp[-1] != '['
+                             && sp[-1] != '\\')
+                                 ;
+                         else if ((sp - sp2) >= 2
+                                  && sp[-1] == '^' && sp[-2] == '[')
+                                 ;
+                         else
+                                count--;
+                }
+
                 if (*sp == '-' && do_lint && ! range_warned && count == 1
                     && sp[-1] != '[' && sp[1] != ']'
                     && ! isdigit((unsigned char) sp[-1]) && ! isdigit((unsigned char) sp[1])