Re: change ci/rlog, etc to accept '@' as part of symbol

[Sam Detweiler]

The issue is whether my email address, 
address@hidden, or  address@hidden is a valid 'author' string.

I claim it is. 

I agree that the '@' when preceded by a white space, 
or first on a line is a treated as a comment character.

I want to allow it IN a symbol. 

Here is my checkidentifier

checkidentifier(id, delimiter, dotok,AtOK)
        register char *id;
        int delimiter;
        register int dotok;
        register int AtOK;
/*   Function:  check whether the string starting at id is an   */
/*              identifier and return a pointer to the delimiter*/
/*              after the identifier.  White space, delim and 0 */
/*              are legal delimiters.  Aborts the program if not*/
/*              a legal identifier. Useful for checking commands*/
/*              If !delim, the only delimiter is 0.             */
/*              Allow '.' in identifier only if DOTOK is set.   */

        register char    *temp;
        register char c;
        register char delim = delimiter;
        int isid = false;

        temp = id;
        for (;;  id++) {
                switch (ctab[(unsigned char)(c = *id)]) {
                        case IDCHAR:
                        case LETTER:
                        case Letter:
                                isid = true;
                                continue;

                        case DIGIT:
                                continue;

                        case PERIOD:
                                if (dotok)
                                        continue;
                                break;
                        case SBEGIN:
                                if(AtOK)
                                  {
                                  /* AT sign can be embedded, but not start */
                                  if(isid)
                                    {
                                    continue;
                                    } /* end if */
                                  else
                                    {
                                     break;
                                    } /* end else */
                                  } /* end if */
                                else
                                  {
                                  /* fall thru into default behavior */
                                  } /* end if */
                        default:
                                break;
                }
                break;
        }
        if (     ! isid
            ||   (c  &&  (!delim || (c!=delim && c!=' ' && c!='\t' && c!='\n')))
        ) {
                /* append \0 to end of id before error message */
                while ((c = *id) && c!=' ' && c!='\t' && c!='\n' && c!=delim)
                    id++;
                *id = '\0';
                faterror("invalid %s `%s'",
                        dotok ? "identifier" : "symbol", temp
                );
        }
        return id;


On Thu, 12 May 2005 16:41:50 -0400 (EDT), Aaron S. Hawley wrote:

>allowing numeric and other "special" characters as the login value of a
>revision is a common conflict for users of RCS.  here's a debian gnu/linux
>bug report on the issue.

>http://bugs.debian.org/247494

>i'm sure there is similar archived discussions of this issue in other
>places.  they will likely accompany reasons why the "status quo" behavior
>exists.  though the status quo could be just backwards compatibility.

>if by "system" you mean some front-end to RCS, i'm sure there's some
>workaround you could do that could provide a workaround of patching RCS
>source files, making your "system" more portable (not requiring a patched
>version of RCS).


Sam