Message-ID: <20260203173939.3c5939c7@pumpkin>
Date: Tue, 3 Feb 2026 17:39:39 +0000
From: David Laight <david.laight.linux@...il.com>
To: Willy Tarreau <w@....eu>, Thomas Weißschuh
 <linux@...ssschuh.net>, linux-kernel@...r.kernel.org, Cheng Li
 <lechain@...il.com>
Subject: Re: [PATCH next 00/12] tools/nolibc: Enhance printf()

On Tue,  3 Feb 2026 10:29:48 +0000
david.laight.linux@...il.com wrote:

> From: David Laight <david.laight.linux@...il.com>
> 
> Update printf() so that it handles almost all the non-fp formats.
> In particular:
> - Left alignment.
> - Zero padding.
> - Field precision.
> - Variable field width and precision.
> - Width modifiers q, L, t and z.
> - Conversion specifiers i and X (X generates lower case).
> About the only thing that is missing is octal.
> 
> The tests are updated to match.
> 
> There is a slight increase in code size, but it is minimalised
> by the the heavy use of bit-pattern matches.

I failed to add 'next' to the last two patches, but I don't expect anything
to actually happen until after the next merge window.
If the patch to change all the number->ascii function is committed I'll
add the trivial patch to support octal.

For reference this is the final version:

/* printf(). Supports most of the normal integer and string formats.
 *  - %[#0-+ ][width|*[.precision|*]][{l,t,z,ll,L,j,q}]{d,i,u,c,x,X,p,s,m,%}
 *  - %% generates a single %
 *  - %m outputs strerror(errno).
 *  - # only affects %x and prepends 0x to non-zero values.
 *  - %o (octal) isn't supported.
 *  - %X outputs a..f the same as %x.
 *  - No support for wide characters.
 *  - invalid formats are copied to the output buffer.
 */
typedef int (*__nolibc_printf_cb)(void *state, const char *buf, size_t size);

#define __PF_FLAG(c) (1u << ((c) & 0x1f))
static __attribute__((unused, format(printf, 3, 0)))
int __nolibc_printf(__nolibc_printf_cb cb, void *state, const char *fmt, va_list args)
{
	char c;
	int len, written, width, precision;
	unsigned int flags, c_flag;
	char tmpbuf[32 + 24];
	const char *outstr;

	written = 0;
	while (1) {
		outstr = fmt;
		c = *fmt++;
		if (!c)
			break;

		width = 0;
		flags = 0;
		if (c != '%') {
			while (*fmt && *fmt != '%')
				fmt++;
			len = fmt - outstr;
		} else {
			/* we're in a format sequence */

			c = *fmt++;

			/* Flag characters */
			for (; c >= 0x20 && c <= 0x3f; c = *fmt++) {
				if ((__PF_FLAG(c) & (__PF_FLAG('-') | __PF_FLAG(' ') | __PF_FLAG('+') |
						     __PF_FLAG('#') | __PF_FLAG('0'))) == 0)
					break;
				flags |= __PF_FLAG(c);
			}

			/* width and precision */
			for (;; c = *fmt++) {
				if (c == '*') {
					precision = va_arg(args, unsigned int);
					c = *fmt++;
				} else {
					for (precision = 0; c >= '0' && c <= '9'; c = *fmt++)
						precision = precision * 10 + (c - '0');
				}
				if (flags & __PF_FLAG('.'))
					break;
				width = precision;
				if (c != '.') {
					/* Default precision for strings */
					precision = INT_MAX;
					break;
				}
				flags |= __PF_FLAG('.');
			}

			/* Length modifiers are lower case except 'L' which is the same a 'q' */
			if ((c >= 'a' && c <= 'z') || (c == 'L' && (c = 'q'))) {
				if (__PF_FLAG(c) & (__PF_FLAG('l') | __PF_FLAG('t') | __PF_FLAG('z') |
						    __PF_FLAG('j') | __PF_FLAG('q'))) {
					if (c == 'l' && fmt[0] == 'l') {
						fmt++;
						c = 'q';
					}
					/* These all miss "# -0+" */
					flags |= __PF_FLAG(c);
					c = *fmt++;
				}
			}

			/* Conversion specifiers are lower case except 'X' treated as 'x' */
			if (!((c >= 'a' && c <= 'z') || (c == 'X' && (c = 'x'))))
				goto bad_conversion_specifier;

			/* Numeric and pointer conversion specifiers.
			 * We need to check for "%p" or "%#x" later, merging here gives better code.
			 * But '#' collides with 'c' so shift right.
			 */
			c_flag = __PF_FLAG(c) | (flags & __PF_FLAG('#')) >> 1;
			if (c_flag & (__PF_FLAG('c') | __PF_FLAG('d') | __PF_FLAG('i') | __PF_FLAG('u') |
					    __PF_FLAG('x') | __PF_FLAG('p') | __PF_FLAG('s'))) {
				unsigned long long v;
				long long signed_v;
				char *out = tmpbuf + 32;
				int sign = 0;

				/* Annoying 'p' === '0' so mask from flags */
				if ((c_flag | (flags & ~__PF_FLAG('p'))) &
				    (__PF_FLAG('p') | __PF_FLAG('s') | __PF_FLAG('l') | __PF_FLAG('t') | __PF_FLAG('z'))) {
					v = va_arg(args, unsigned long);
					signed_v = (long)v;
				} else if (flags & (__PF_FLAG('j') | __PF_FLAG('q'))) {
					v = va_arg(args, unsigned long long);
					signed_v = v;
				} else {
					v = va_arg(args, unsigned int);
					signed_v = (int)v;
				}

				if (c_flag & __PF_FLAG('c')) {
					tmpbuf[0] = v;
					len = 1;
					outstr = tmpbuf;
					goto do_output;
				}

				if (c_flag & __PF_FLAG('s')) {
					if (!v) {
						outstr = "(null)";
						/* Match glibc, nothing output if precision too small */
						len = precision >= 6 ? 6 : 0;
						goto do_output;
					}
					outstr = (void *)v;
do_strnlen_output:
					len = strnlen(outstr, precision);
					goto do_output;
				}

				if (c_flag & (__PF_FLAG('d') | __PF_FLAG('i'))) {
					if (signed_v < 0) {
						sign = '-';
						v = -(signed_v + 1);
						v++;
					} else if (flags & __PF_FLAG('+')) {
						sign = '+';
					} else if (flags & __PF_FLAG(' ')) {
						sign = ' ';
					}
				}

				if (v == 0) {
					/* There are special rules for zero. */
					if (c_flag & __PF_FLAG('p')) {
						/* match glibc, precision is ignored */
						outstr = "(nil)";
						len = 5;
						goto do_output;
					}
					if (!precision) {
						/* Explicit %nn.0d, no digits output */
						len = 0;
						goto prepend_sign;
					}
					/* "#x" should output "0" not "0x0" */
					*out = '0';
					len = 1;
				} else {
					if (c_flag & (__PF_FLAG('d') | __PF_FLAG('i') | __PF_FLAG('u'))) {
						len = u64toa_r(v, out);
					} else {
						len = u64toh_r(v, out);
						if (c_flag & (__PF_FLAG('p') | __PF_FLAG('#' - 1)))
							sign = 'x' | '0' << 8;
					}
				}

				/* Add zero padding */
				if (flags & (__PF_FLAG('0') | __PF_FLAG('.'))) {
					if (!(flags & __PF_FLAG('.'))) {
						if (flags & __PF_FLAG('-'))
							/* Left justify overrides zero pad */
							goto prepend_sign;
						/* Zero pad to field width less sign */
						precision = width;
						if (sign) {
							precision--;
							if (sign >= 256)
								precision--;
						}
					}
					if (precision > 30)
						/* Don't run off the start of tmpbuf[] */
						precision = 30;
					for (; len < precision; len++) {
						/* Stop gcc generating horrid code and memset().
						 * This is OPTIMIZER_HIDE_VAR() from compiler.h.
						 */
						__asm__ volatile("" : "=r"(len) : "0"(len));
						*--out = '0';
					}
				}

prepend_sign:
				for (; sign; sign >>= 8) {
					len++;
					*--out = sign;
				}
				outstr = out;
			}
			else if (c == 'm') {
#ifdef NOLIBC_IGNORE_ERRNO
				outstr = "unknown error";
				len = __builtin_strlen(outstr);
#else
				outstr = strerror(errno);
				goto do_strnlen_output;
#endif /* NOLIBC_IGNORE_ERRNO */
			} else {
bad_conversion_specifier:
				if (c != '%')
					/* Invalid format, output the format string */
					fmt = outstr + 1;
				/* %% is documented as a 'conversion specifier'.
				 * Any flags, precision or length modifier are ignored.
				 */
				outstr = fmt - 1;
				len = 1;
				width = 0;
			}
		}

do_output:
		written += len;

                /* An OPTIMIZER_HIDE_VAR() seems to stop gcc back-merging this
                 * code into one of the conditionals above.
		 */
                __asm__ volatile("" : "=r"(len) : "0"(len));

		/* Output 'left pad', 'value' then 'right pad'. */
		flags &= __PF_FLAG('-');
		width -= len;
		if (flags && cb(state, outstr, len) != 0)
			return -1;
		while (width > 0) {
			int pad_len = ((width - 1) & 15) + 1;
			width -= pad_len;
			written += pad_len;
			if (cb(state, "                ", pad_len) != 0)
				return -1;
		}
		if (!flags && cb(state, outstr, len) != 0)
			return -1;
	}

	if (cb(state, NULL, 0) != 0)
		return -1;

	return written;
}
#undef _PF_FLAG

struct __nolibc_fprintf_cb_state {
	FILE *stream;
	unsigned int buf_offset;
	char buf[128];
};

static int __nolibc_fprintf_cb(void *v_state, const char *buf, size_t size)
{
	struct __nolibc_fprintf_cb_state *state = v_state;
	unsigned int off = state->buf_offset;

	if (off + size > sizeof(state->buf) || buf == NULL) {
		state->buf_offset = 0;
		if (off && _fwrite(state->buf, off, state->stream))
			return -1;
		if (size > sizeof(state->buf))
			return _fwrite(buf, size, state->stream);
		off = 0;
	}

	if (size) {
		state->buf_offset = off + size;
		memcpy(state->buf + off, buf, size);
	}
	return 0;
}

static __attribute__((unused, format(printf, 2, 0)))
int vfprintf(FILE *stream, const char *fmt, va_list args)
{
	struct __nolibc_fprintf_cb_state state;

	state.stream = stream;
	state.buf_offset = 0;
	return __nolibc_printf(__nolibc_fprintf_cb, &state, fmt, args);
}

struct __nolibc_sprintf_cb_state {
	char *buf;
	size_t size;
};

static int __nolibc_sprintf_cb(void *v_state, const char *buf, size_t size)
{
	struct __nolibc_sprintf_cb_state *state = v_state;
	char *tgt;

	if (size >= state->size) {
		if (state->size <= 1)
			return 0;
		size = state->size - 1;
	}
	tgt = state->buf;
	if (size) {
		state->size -= size;
		state->buf = tgt + size;
		memcpy(tgt, buf, size);
	} else {
		*tgt = '\0';
	}
	return 0;
}

static __attribute__((unused, format(printf, 3, 0)))
int vsnprintf(char *buf, size_t size, const char *fmt, va_list args)
{
	struct __nolibc_sprintf_cb_state state = { .buf = buf, .size = size };

	return __nolibc_printf(__nolibc_sprintf_cb, &state, fmt, args);
}

Powered by blists - more mailing lists