Test harness for crackaddr from sendmail 8.12.7

sendmail_crackaddr.c

// Test harness for crackaddr from sendmail 8.12.7 by @mechanicalnull
// sendmail license: https://web.archive.org/web/20160322142305/https://www.sendmail.com/pdfs/open_source/sendmail_license.pdf


#include <string.h>
#include <ctype.h>

#define MAXNAME 256
typedef int bool;
#define false 0
#define true 1
#define ColonOkInAddr 1
#define MACROEXPAND	((unsigned char)0201)	/* macro expansion */
char *MustQuoteChars = "@,;:\\()[].'";

/*
**  CRACKADDR -- parse an address and turn it into a macro
**
**	This doesn't actually parse the address -- it just extracts
**	it and replaces it with "$g".  The parse is totally ad hoc
**	and isn't even guaranteed to leave something syntactically
**	identical to what it started with.  However, it does leave
**	something semantically identical.
**
**	This algorithm has been cleaned up to handle a wider range
**	of cases -- notably quoted and backslash escaped strings.
**	This modification makes it substantially better at preserving
**	the original syntax.
**
**	Parameters:
**		addr -- the address to be cracked.
**
**	Returns:
**		a pointer to the new version.
**
**	Side Effects:
**		none.
**
**	Warning:
**		The return value is saved in local storage and should
**		be copied if it is to be reused.
*/

char *
crackaddr(addr)
	register char *addr;
{
	register char *p;
	register char c;
	int cmtlev;
	int realcmtlev;
	int anglelev, realanglelev;
	int copylev;
	int bracklev;
	bool qmode;
	bool realqmode;
	bool skipping;
	bool putgmac = false;
	bool quoteit = false;
	bool gotangle = false;
	bool gotcolon = false;
	register char *bp;
	char *buflim;
	char *bufhead;
	char *addrhead;
	static char buf[MAXNAME + 1];

    // Omit debug prints for testing
    /*
	if (tTd(33, 1))
		sm_dprintf("crackaddr(%s)\n", addr);
    */

	/* strip leading spaces */
	while (*addr != '\0' && isascii(*addr) && isspace(*addr))
		addr++;

	/*
	**  Start by assuming we have no angle brackets.  This will be
	**  adjusted later if we find them.
	*/

	bp = bufhead = buf;
	buflim = &buf[sizeof buf - 7];
	p = addrhead = addr;
	copylev = anglelev = realanglelev = cmtlev = realcmtlev = 0;
	bracklev = 0;
	qmode = realqmode = false;

	while ((c = *p++) != '\0')
	{
		/*
		**  If the buffer is overful, go into a special "skipping"
		**  mode that tries to keep legal syntax but doesn't actually
		**  output things.
		*/

		skipping = bp >= buflim;

		if (copylev > 0 && !skipping)
			*bp++ = c;

		/* check for backslash escapes */
		if (c == '\\')
		{
			/* arrange to quote the address */
			if (cmtlev <= 0 && !qmode)
				quoteit = true;

			if ((c = *p++) == '\0')
			{
				/* too far */
				p--;
				goto putg;
			}
			if (copylev > 0 && !skipping)
				*bp++ = c;
			goto putg;
		}

		/* check for quoted strings */
		if (c == '"' && cmtlev <= 0)
		{
			qmode = !qmode;
			if (copylev > 0 && !skipping)
				realqmode = !realqmode;
			continue;
		}
		if (qmode)
			goto putg;

		/* check for comments */
		if (c == '(')
		{
			cmtlev++;

			/* allow space for closing paren */
			if (!skipping)
			{
				buflim--;
				realcmtlev++;
				if (copylev++ <= 0)
				{
					if (bp != bufhead)
						*bp++ = ' ';
					*bp++ = c;
				}
			}
		}
		if (cmtlev > 0)
		{
			if (c == ')')
			{
				cmtlev--;
				copylev--;
				if (!skipping)
				{
					realcmtlev--;
					buflim++;
				}
			}
			continue;
		}
		else if (c == ')')
		{
			/* syntax error: unmatched ) */
			if (copylev > 0 && !skipping)
				bp--;
		}

		/* count nesting on [ ... ] (for IPv6 domain literals) */
		if (c == '[')
			bracklev++;
		else if (c == ']')
			bracklev--;

		/* check for group: list; syntax */
		if (c == ':' && anglelev <= 0 && bracklev <= 0 &&
		    !gotcolon && !ColonOkInAddr)
		{
			register char *q;

			/*
			**  Check for DECnet phase IV ``::'' (host::user)
			**  or **  DECnet phase V ``:.'' syntaxes.  The latter
			**  covers ``user@DEC:.tay.myhost'' and
			**  ``DEC:.tay.myhost::user'' syntaxes (bletch).
			*/

			if (*p == ':' || *p == '.')
			{
				if (cmtlev <= 0 && !qmode)
					quoteit = true;
				if (copylev > 0 && !skipping)
				{
					*bp++ = c;
					*bp++ = *p;
				}
				p++;
				goto putg;
			}

			gotcolon = true;

			bp = bufhead;
			if (quoteit)
			{
				*bp++ = '"';

				/* back up over the ':' and any spaces */
				--p;
				while (isascii(*--p) && isspace(*p))
					continue;
				p++;
			}
			for (q = addrhead; q < p; )
			{
				c = *q++;
				if (bp < buflim)
				{
					if (quoteit && c == '"')
						*bp++ = '\\';
					*bp++ = c;
				}
			}
			if (quoteit)
			{
				if (bp == &bufhead[1])
					bp--;
				else
					*bp++ = '"';
				while ((c = *p++) != ':')
				{
					if (bp < buflim)
						*bp++ = c;
				}
				*bp++ = c;
			}

			/* any trailing white space is part of group: */
			while (isascii(*p) && isspace(*p) && bp < buflim)
				*bp++ = *p++;
			copylev = 0;
			putgmac = quoteit = false;
			bufhead = bp;
			addrhead = p;
			continue;
		}

		if (c == ';' && copylev <= 0 && !ColonOkInAddr)
		{
			if (bp < buflim)
				*bp++ = c;
		}

		/* check for characters that may have to be quoted */
		if (strchr(MustQuoteChars, c) != NULL)
		{
			/*
			**  If these occur as the phrase part of a <>
			**  construct, but are not inside of () or already
			**  quoted, they will have to be quoted.  Note that
			**  now (but don't actually do the quoting).
			*/

			if (cmtlev <= 0 && !qmode)
				quoteit = true;
		}

		/* check for angle brackets */
		if (c == '<')
		{
			register char *q;

			/* assume first of two angles is bogus */
			if (gotangle)
				quoteit = true;
			gotangle = true;

			/* oops -- have to change our mind */
			anglelev = 1;
			if (!skipping)
				realanglelev = 1;

			bp = bufhead;
			if (quoteit)
			{
				*bp++ = '"';

				/* back up over the '<' and any spaces */
				--p;
				while (isascii(*--p) && isspace(*p))
					continue;
				p++;
			}
			for (q = addrhead; q < p; )
			{
				c = *q++;
				if (bp < buflim)
				{
					if (quoteit && c == '"')
						*bp++ = '\\';
					*bp++ = c;
				}
			}
			if (quoteit)
			{
				if (bp == &buf[1])
					bp--;
				else
					*bp++ = '"';
				while ((c = *p++) != '<')
				{
					if (bp < buflim)
						*bp++ = c;
				}
				*bp++ = c;
			}
			copylev = 0;
			putgmac = quoteit = false;
			continue;
		}

		if (c == '>')
		{
			if (anglelev > 0)
			{
				anglelev--;
				if (!skipping)
				{
					realanglelev--;
					buflim++;
				}
			}
			else if (!skipping)
			{
				/* syntax error: unmatched > */
				if (copylev > 0)
					bp--;
				quoteit = true;
				continue;
			}
			if (copylev++ <= 0)
				*bp++ = c;
			continue;
		}

		/* must be a real address character */
	putg:
		if (copylev <= 0 && !putgmac)
		{
			if (bp > bufhead && bp[-1] == ')')
				*bp++ = ' ';
			*bp++ = MACROEXPAND;
			*bp++ = 'g';
			putgmac = true;
		}
	}

	/* repair any syntactic damage */
	if (realqmode)
		*bp++ = '"';
	while (realcmtlev-- > 0)
		*bp++ = ')';
	while (realanglelev-- > 0)
		*bp++ = '>';
	*bp++ = '\0';

    // Omitting debug prints for testing
    /*
	if (tTd(33, 1))
	{
		sm_dprintf("crackaddr=>`");
		xputs(buf);
		sm_dprintf("'\n");
	}
    */

	return buf;
}



#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>


int main( int argc, char **argv ){
        ssize_t num_bytes_read;
        char input_buf[1003] = {0};

        if( argc == 1 ) {
                num_bytes_read = read(STDIN_FILENO, input_buf, sizeof(input_buf)-1);
                if (num_bytes_read > 0)
                        crackaddr( input_buf );
        }
        else
                printf("Please supply input via stdin.\n");
}

Note this is trivial to turn into a persistent-mode or Libfuzzer-style harness. I used a standard main function for a more apples-to-apples comparison with the CGC version.