Re: fencepost error in encoding processing

[Ken Raeburn <raeburn@raeburn.org>]

On Nov 16, 2009, at 08:03, Ludovic Courtès wrote:
> As far as encoding names are concerned, Bruno Haible pointed me to
> http://www.iana.org/assignments/character-sets and I added a link to  
> it
> in the manual a couple of days ago.

Between your link and Mike's, it looks to me like we should add  
several more characters.

The GNU libc code adds ":" and "," to the list.  The comment in  
iconv_open doesn't list the comma, but the function it calls does  
permit it.  There's also some special handling of "/".

The IANA list shows names using "+" and parens ("ebcdic-us-37+euro",  
"NF_Z_62-010_(1973)"), as well as colons.

I've skimmed the ICU page Mike pointed to, and it includes names like  
"UTF-16BE,version=1" and "ibm-1149_P100-197,swaplfnl" as well as "+"  
and ":" names, when showing "all aliases".  If we only try to support,  
say, IANA and MIME, then "+" and ":" are used but not "=".

Since we're scanning an Emacs-style coding specification, as long as  
whitespace and semicolon aren't on the list, I think we can be  
expansive, so let's go ahead and include all of ":,+=/()" to the  
allowed set.  The results will still be constrained by whatever the OS  
supports; we just don't want Guile to impose additional constraints.

Should we allow punctuation in general by calling ispunct (and  
explicitly checking for semicolon) instead?  (Note that isalnum and  
ispunct will also check for locale-specific characters... of course,  
the new encoding spec hasn't come into effect yet....)

Ken

     Allow more characters in coding system names in Emacs-style  
declarations.

     * libguile/read.c (scm_i_scan_for_encoding): Allow more punctuation
       symbols in coding system names.

diff --git a/libguile/read.c b/libguile/read.c
index 775612a..657e101 100644
--- a/libguile/read.c
+++ b/libguile/read.c
@@ -1506,8 +1506,7 @@ scm_i_scan_for_encoding (SCM port)
    i = 0;
    while (pos + i - header <= SCM_ENCODING_SEARCH_SIZE
           && pos + i - header < bytes_read
-	 && (isalnum((int) pos[i]) || pos[i] == '_' || pos[i] == '-'
-             || pos[i] == '.'))
+	 && (isalnum((int) pos[i]) || strchr("_-.:/,+=()", pos[i]) != NULL))
      i++;

    if (i == 0)