Re: [PATCH v2 01/10] sequencer.c: remove broken support for rfc2822 continuation in footer

From: Jonathan Nieder <jrnieder@gmail.com>
Date: Jan 22, 2013, 07:54 UTC
Message-ID: <20130122075413.GB6085@elie.Belkin>
In-Reply-To: <1358757627-16682-2-git-send-email-drafnel@gmail.com>

Hi,

Brandon Casey wrote:

> --- a/sequencer.c
> +++ b/sequencer.c

[...]

> @@ -1042,13 +1041,8 @@ static int ends_rfc2822_footer(struct strbuf *sb, int ignore_footer)

Git is checking if (sb->buf) ends with a "Signed-off-by:" style line. If it doesn't, it will need to add an extra blank line before adding a new sign-off.

First (snipped), it seeks back two newlines from the end and then forward to the next non-newline character, so (buf + i) is at the start of the last line of (the interesting part of) sb. Now:

> 	for (; i < len; i = k) {
> 		for (k = i; k < len && buf[k] != '\n'; k++)
>  			; /* do nothing */
>  		k++;

(buf + k) points to the end of this line.

> -		if ((buf[k] == ' ' || buf[k] == '\t') && !first)
> -			continue;

This is always the first line examined, so this "continue" never triggers.

> -
> -		first = 0;
> -
>  		for (j = 0; i + j < len; j++) {

If the line matches /^[[:alnum:]-]*:/, it passes and git moves on to the (nonexistent) next line. Otherwise, it fails.

Do I understand correctly? If so, this patch should be a no-op, which is good, I guess.

But in that case, couldn't this function be made much simpler? As far as I can tell, all the function needs to do is the following:

	1. Find the last line.
	2. Check if it is blank or matches /^[[:alnum:]-]*:/
	3. There is no step 3.  That's it.

In other words, something like:

	const char *eol, *p;

	/* End of line */
	eol = memrchr(sb->buf, '\n', sb->len - ignore_footer);
	if (!eol)
		eol = sb->buf;

	/* Start of line */
	p = memrchr(sb->buf, '\n', eol - sb->buf);
	if (p)
		p++;
	else
		p = sb->buf;

	if (p == eol)	/* Blank line? */
		return 1;

	/* "Signed-off-by"-style field */
	while ((isalnum(*p) || *p == '-') && p < eol)
		p++;
	return *p == ':';

where memrchr is defined roughly as follows[1]:

	#ifdef __GLIBC_PREREQ
	#if __GLIBC_PREREQ(2, 2)
	#define HAVE_MEMRCHR
	#endif
	#endif

	#ifndef HAVE_MEMRCHR
	#define memrchr gitmemrchr
	static inline void *gitmemrchr(const void *s, int c, size_t n)
	{
		const unsigned char *p = s;
		p += n;
		while (p != s)
			if (*--p == (unsigned char) c)
				return p;
		return NULL;
	}
	#endif

Does that look right?

Jonathan

[1] http://thread.gmane.org/gmane.comp.version-control.git/159081/focus=159121