| 1 | /* Output like sprintf to a buffer of specified size. |
| 2 | Also takes args differently: pass one pointer to the end |
| 3 | of the format string in addition to the format string itself. |
| 4 | Copyright (C) 1985, 2001-2014 Free Software Foundation, Inc. |
| 5 | |
| 6 | This file is part of GNU Emacs. |
| 7 | |
| 8 | GNU Emacs is free software: you can redistribute it and/or modify |
| 9 | it under the terms of the GNU General Public License as published by |
| 10 | the Free Software Foundation, either version 3 of the License, or |
| 11 | (at your option) any later version. |
| 12 | |
| 13 | GNU Emacs is distributed in the hope that it will be useful, |
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | GNU General Public License for more details. |
| 17 | |
| 18 | You should have received a copy of the GNU General Public License |
| 19 | along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */ |
| 20 | |
| 21 | /* If you think about replacing this with some similar standard C function of |
| 22 | the printf family (such as vsnprintf), please note that this function |
| 23 | supports the following Emacs-specific features: |
| 24 | |
| 25 | . For %c conversions, it produces a string with the multibyte representation |
| 26 | of the (`int') argument, suitable for display in an Emacs buffer. |
| 27 | |
| 28 | . For %s and %c, when field width is specified (e.g., %25s), it accounts for |
| 29 | the display width of each character, according to char-width-table. That |
| 30 | is, it does not assume that each character takes one column on display. |
| 31 | |
| 32 | . If the size of the buffer is not enough to produce the formatted string in |
| 33 | its entirety, it makes sure that truncation does not chop the last |
| 34 | character in the middle of its multibyte sequence, producing an invalid |
| 35 | sequence. |
| 36 | |
| 37 | . It accepts a pointer to the end of the format string, so the format string |
| 38 | could include embedded null characters. |
| 39 | |
| 40 | . It signals an error if the length of the formatted string is about to |
| 41 | overflow ptrdiff_t or size_t, to avoid producing strings longer than what |
| 42 | Emacs can handle. |
| 43 | |
| 44 | OTOH, this function supports only a small subset of the standard C formatted |
| 45 | output facilities. E.g., %u and %ll are not supported, and precision is |
| 46 | ignored %s and %c conversions. (See below for the detailed documentation of |
| 47 | what is supported.) However, this is okay, as this function is supposed to |
| 48 | be called from `error' and similar functions, and thus does not need to |
| 49 | support features beyond those in `Fformat', which is used by `error' on the |
| 50 | Lisp level. */ |
| 51 | |
| 52 | /* This function supports the following %-sequences in the `format' |
| 53 | argument: |
| 54 | |
| 55 | %s means print a string argument. |
| 56 | %S is silently treated as %s, for loose compatibility with `Fformat'. |
| 57 | %d means print a `signed int' argument in decimal. |
| 58 | %o means print an `unsigned int' argument in octal. |
| 59 | %x means print an `unsigned int' argument in hex. |
| 60 | %e means print a `double' argument in exponential notation. |
| 61 | %f means print a `double' argument in decimal-point notation. |
| 62 | %g means print a `double' argument in exponential notation |
| 63 | or in decimal-point notation, whichever uses fewer characters. |
| 64 | %c means print a `signed int' argument as a single character. |
| 65 | %% means produce a literal % character. |
| 66 | |
| 67 | A %-sequence may contain optional flag, width, and precision specifiers, and |
| 68 | a length modifier, as follows: |
| 69 | |
| 70 | %<flags><width><precision><length>character |
| 71 | |
| 72 | where flags is [+ -0], width is [0-9]+, precision is .[0-9]+, and length |
| 73 | is empty or l or the value of the pD or pI or pMd (sans "d") macros. |
| 74 | Also, %% in a format stands for a single % in the output. A % that |
| 75 | does not introduce a valid %-sequence causes undefined behavior. |
| 76 | |
| 77 | The + flag character inserts a + before any positive number, while a space |
| 78 | inserts a space before any positive number; these flags only affect %d, %o, |
| 79 | %x, %e, %f, and %g sequences. The - and 0 flags affect the width specifier, |
| 80 | as described below. For signed numerical arguments only, the ` ' (space) |
| 81 | flag causes the result to be prefixed with a space character if it does not |
| 82 | start with a sign (+ or -). |
| 83 | |
| 84 | The l (lower-case letter ell) length modifier is a `long' data type |
| 85 | modifier: it is supported for %d, %o, and %x conversions of integral |
| 86 | arguments, must immediately precede the conversion specifier, and means that |
| 87 | the respective argument is to be treated as `long int' or `unsigned long |
| 88 | int'. Similarly, the value of the pD macro means to use ptrdiff_t, |
| 89 | the value of the pI macro means to use EMACS_INT or EMACS_UINT, the |
| 90 | value of the pMd etc. macros means to use intmax_t or uintmax_t, |
| 91 | and the empty length modifier means `int' or `unsigned int'. |
| 92 | |
| 93 | The width specifier supplies a lower limit for the length of the printed |
| 94 | representation. The padding, if any, normally goes on the left, but it goes |
| 95 | on the right if the - flag is present. The padding character is normally a |
| 96 | space, but (for numerical arguments only) it is 0 if the 0 flag is present. |
| 97 | The - flag takes precedence over the 0 flag. |
| 98 | |
| 99 | For %e, %f, and %g sequences, the number after the "." in the precision |
| 100 | specifier says how many decimal places to show; if zero, the decimal point |
| 101 | itself is omitted. For %s and %S, the precision specifier is ignored. */ |
| 102 | |
| 103 | #include <config.h> |
| 104 | #include <stdio.h> |
| 105 | #include <float.h> |
| 106 | #include <unistd.h> |
| 107 | #include <limits.h> |
| 108 | |
| 109 | #include "lisp.h" |
| 110 | |
| 111 | /* Since we use the macro CHAR_HEAD_P, we have to include this, but |
| 112 | don't have to include others because CHAR_HEAD_P does not contains |
| 113 | another macro. */ |
| 114 | #include "character.h" |
| 115 | |
| 116 | /* Generate output from a format-spec FORMAT, |
| 117 | terminated at position FORMAT_END. |
| 118 | (*FORMAT_END is not part of the format, but must exist and be readable.) |
| 119 | Output goes in BUFFER, which has room for BUFSIZE chars. |
| 120 | BUFSIZE must be positive. If the output does not fit, truncate it |
| 121 | to fit and return BUFSIZE - 1; if this truncates a multibyte |
| 122 | sequence, store '\0' into the sequence's first byte. |
| 123 | Returns the number of bytes stored into BUFFER, excluding |
| 124 | the terminating null byte. Output is always null-terminated. |
| 125 | String arguments are passed as C strings. |
| 126 | Integers are passed as C integers. */ |
| 127 | |
| 128 | ptrdiff_t |
| 129 | doprnt (char *buffer, ptrdiff_t bufsize, const char *format, |
| 130 | const char *format_end, va_list ap) |
| 131 | { |
| 132 | const char *fmt = format; /* Pointer into format string. */ |
| 133 | char *bufptr = buffer; /* Pointer into output buffer. */ |
| 134 | |
| 135 | /* Use this for sprintf unless we need something really big. */ |
| 136 | char tembuf[DBL_MAX_10_EXP + 100]; |
| 137 | |
| 138 | /* Size of sprintf_buffer. */ |
| 139 | ptrdiff_t size_allocated = sizeof (tembuf); |
| 140 | |
| 141 | /* Buffer to use for sprintf. Either tembuf or same as BIG_BUFFER. */ |
| 142 | char *sprintf_buffer = tembuf; |
| 143 | |
| 144 | /* Buffer we have got with malloc. */ |
| 145 | char *big_buffer = NULL; |
| 146 | |
| 147 | ptrdiff_t tem = -1; |
| 148 | char *string; |
| 149 | char fixed_buffer[20]; /* Default buffer for small formatting. */ |
| 150 | char *fmtcpy; |
| 151 | int minlen; |
| 152 | char charbuf[MAX_MULTIBYTE_LENGTH + 1]; /* Used for %c. */ |
| 153 | USE_SAFE_ALLOCA; |
| 154 | |
| 155 | if (format_end == 0) |
| 156 | format_end = format + strlen (format); |
| 157 | |
| 158 | fmtcpy = (format_end - format < sizeof (fixed_buffer) - 1 |
| 159 | ? fixed_buffer |
| 160 | : SAFE_ALLOCA (format_end - format + 1)); |
| 161 | |
| 162 | bufsize--; |
| 163 | |
| 164 | /* Loop until end of format string or buffer full. */ |
| 165 | while (fmt < format_end && bufsize > 0) |
| 166 | { |
| 167 | if (*fmt == '%') /* Check for a '%' character */ |
| 168 | { |
| 169 | ptrdiff_t size_bound = 0; |
| 170 | ptrdiff_t width; /* Columns occupied by STRING on display. */ |
| 171 | enum { |
| 172 | pDlen = sizeof pD - 1, |
| 173 | pIlen = sizeof pI - 1, |
| 174 | pMlen = sizeof pMd - 2 |
| 175 | }; |
| 176 | enum { |
| 177 | no_modifier, long_modifier, pD_modifier, pI_modifier, pM_modifier |
| 178 | } length_modifier = no_modifier; |
| 179 | static char const modifier_len[] = { 0, 1, pDlen, pIlen, pMlen }; |
| 180 | int maxmlen = max (max (1, pDlen), max (pIlen, pMlen)); |
| 181 | int mlen; |
| 182 | |
| 183 | fmt++; |
| 184 | /* Copy this one %-spec into fmtcpy. */ |
| 185 | string = fmtcpy; |
| 186 | *string++ = '%'; |
| 187 | while (fmt < format_end) |
| 188 | { |
| 189 | *string++ = *fmt; |
| 190 | if ('0' <= *fmt && *fmt <= '9') |
| 191 | { |
| 192 | /* Get an idea of how much space we might need. |
| 193 | This might be a field width or a precision; e.g. |
| 194 | %1.1000f and %1000.1f both might need 1000+ bytes. |
| 195 | Parse the width or precision, checking for overflow. */ |
| 196 | ptrdiff_t n = *fmt - '0'; |
| 197 | while (fmt + 1 < format_end |
| 198 | && '0' <= fmt[1] && fmt[1] <= '9') |
| 199 | { |
| 200 | /* Avoid ptrdiff_t, size_t, and int overflow, as |
| 201 | many sprintfs mishandle widths greater than INT_MAX. |
| 202 | This test is simple but slightly conservative: e.g., |
| 203 | (INT_MAX - INT_MAX % 10) is reported as an overflow |
| 204 | even when it's not. */ |
| 205 | if (n >= min (INT_MAX, min (PTRDIFF_MAX, SIZE_MAX)) / 10) |
| 206 | error ("Format width or precision too large"); |
| 207 | n = n * 10 + fmt[1] - '0'; |
| 208 | *string++ = *++fmt; |
| 209 | } |
| 210 | |
| 211 | if (size_bound < n) |
| 212 | size_bound = n; |
| 213 | } |
| 214 | else if (! (*fmt == '-' || *fmt == ' ' || *fmt == '.' |
| 215 | || *fmt == '+')) |
| 216 | break; |
| 217 | fmt++; |
| 218 | } |
| 219 | |
| 220 | /* Check for the length modifiers in textual length order, so |
| 221 | that longer modifiers override shorter ones. */ |
| 222 | for (mlen = 1; mlen <= maxmlen; mlen++) |
| 223 | { |
| 224 | if (format_end - fmt < mlen) |
| 225 | break; |
| 226 | if (mlen == 1 && *fmt == 'l') |
| 227 | length_modifier = long_modifier; |
| 228 | if (mlen == pDlen && memcmp (fmt, pD, pDlen) == 0) |
| 229 | length_modifier = pD_modifier; |
| 230 | if (mlen == pIlen && memcmp (fmt, pI, pIlen) == 0) |
| 231 | length_modifier = pI_modifier; |
| 232 | if (mlen == pMlen && memcmp (fmt, pMd, pMlen) == 0) |
| 233 | length_modifier = pM_modifier; |
| 234 | } |
| 235 | |
| 236 | mlen = modifier_len[length_modifier]; |
| 237 | memcpy (string, fmt + 1, mlen); |
| 238 | string += mlen; |
| 239 | fmt += mlen; |
| 240 | *string = 0; |
| 241 | |
| 242 | /* Make the size bound large enough to handle floating point formats |
| 243 | with large numbers. */ |
| 244 | if (size_bound > min (PTRDIFF_MAX, SIZE_MAX) - DBL_MAX_10_EXP - 50) |
| 245 | error ("Format width or precision too large"); |
| 246 | size_bound += DBL_MAX_10_EXP + 50; |
| 247 | |
| 248 | /* Make sure we have that much. */ |
| 249 | if (size_bound > size_allocated) |
| 250 | { |
| 251 | if (big_buffer) |
| 252 | xfree (big_buffer); |
| 253 | big_buffer = xmalloc_atomic (size_bound); |
| 254 | sprintf_buffer = big_buffer; |
| 255 | size_allocated = size_bound; |
| 256 | } |
| 257 | minlen = 0; |
| 258 | switch (*fmt++) |
| 259 | { |
| 260 | default: |
| 261 | error ("Invalid format operation %s", fmtcpy); |
| 262 | |
| 263 | /* case 'b': */ |
| 264 | case 'l': |
| 265 | case 'd': |
| 266 | switch (length_modifier) |
| 267 | { |
| 268 | case no_modifier: |
| 269 | { |
| 270 | int v = va_arg (ap, int); |
| 271 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 272 | } |
| 273 | break; |
| 274 | case long_modifier: |
| 275 | { |
| 276 | long v = va_arg (ap, long); |
| 277 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 278 | } |
| 279 | break; |
| 280 | case pD_modifier: |
| 281 | signed_pD_modifier: |
| 282 | { |
| 283 | ptrdiff_t v = va_arg (ap, ptrdiff_t); |
| 284 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 285 | } |
| 286 | break; |
| 287 | case pI_modifier: |
| 288 | { |
| 289 | EMACS_INT v = va_arg (ap, EMACS_INT); |
| 290 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 291 | } |
| 292 | break; |
| 293 | case pM_modifier: |
| 294 | { |
| 295 | intmax_t v = va_arg (ap, intmax_t); |
| 296 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 297 | } |
| 298 | break; |
| 299 | } |
| 300 | /* Now copy into final output, truncating as necessary. */ |
| 301 | string = sprintf_buffer; |
| 302 | goto doit; |
| 303 | |
| 304 | case 'o': |
| 305 | case 'x': |
| 306 | switch (length_modifier) |
| 307 | { |
| 308 | case no_modifier: |
| 309 | { |
| 310 | unsigned v = va_arg (ap, unsigned); |
| 311 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 312 | } |
| 313 | break; |
| 314 | case long_modifier: |
| 315 | { |
| 316 | unsigned long v = va_arg (ap, unsigned long); |
| 317 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 318 | } |
| 319 | break; |
| 320 | case pD_modifier: |
| 321 | goto signed_pD_modifier; |
| 322 | case pI_modifier: |
| 323 | { |
| 324 | EMACS_UINT v = va_arg (ap, EMACS_UINT); |
| 325 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 326 | } |
| 327 | break; |
| 328 | case pM_modifier: |
| 329 | { |
| 330 | uintmax_t v = va_arg (ap, uintmax_t); |
| 331 | tem = sprintf (sprintf_buffer, fmtcpy, v); |
| 332 | } |
| 333 | break; |
| 334 | } |
| 335 | /* Now copy into final output, truncating as necessary. */ |
| 336 | string = sprintf_buffer; |
| 337 | goto doit; |
| 338 | |
| 339 | case 'f': |
| 340 | case 'e': |
| 341 | case 'g': |
| 342 | { |
| 343 | double d = va_arg (ap, double); |
| 344 | tem = sprintf (sprintf_buffer, fmtcpy, d); |
| 345 | /* Now copy into final output, truncating as necessary. */ |
| 346 | string = sprintf_buffer; |
| 347 | goto doit; |
| 348 | } |
| 349 | |
| 350 | case 'S': |
| 351 | string[-1] = 's'; |
| 352 | case 's': |
| 353 | if (fmtcpy[1] != 's') |
| 354 | minlen = atoi (&fmtcpy[1]); |
| 355 | string = va_arg (ap, char *); |
| 356 | tem = strlen (string); |
| 357 | if (STRING_BYTES_BOUND < tem) |
| 358 | error ("String for %%s or %%S format is too long"); |
| 359 | width = strwidth (string, tem); |
| 360 | goto doit1; |
| 361 | |
| 362 | /* Copy string into final output, truncating if no room. */ |
| 363 | doit: |
| 364 | eassert (0 <= tem); |
| 365 | /* Coming here means STRING contains ASCII only. */ |
| 366 | if (STRING_BYTES_BOUND < tem) |
| 367 | error ("Format width or precision too large"); |
| 368 | width = tem; |
| 369 | doit1: |
| 370 | /* We have already calculated: |
| 371 | TEM -- length of STRING, |
| 372 | WIDTH -- columns occupied by STRING when displayed, and |
| 373 | MINLEN -- minimum columns of the output. */ |
| 374 | if (minlen > 0) |
| 375 | { |
| 376 | while (minlen > width && bufsize > 0) |
| 377 | { |
| 378 | *bufptr++ = ' '; |
| 379 | bufsize--; |
| 380 | minlen--; |
| 381 | } |
| 382 | minlen = 0; |
| 383 | } |
| 384 | if (tem > bufsize) |
| 385 | { |
| 386 | /* Truncate the string at character boundary. */ |
| 387 | tem = bufsize; |
| 388 | do |
| 389 | { |
| 390 | tem--; |
| 391 | if (CHAR_HEAD_P (string[tem])) |
| 392 | { |
| 393 | if (BYTES_BY_CHAR_HEAD (string[tem]) <= bufsize - tem) |
| 394 | tem = bufsize; |
| 395 | break; |
| 396 | } |
| 397 | } |
| 398 | while (tem != 0); |
| 399 | |
| 400 | memcpy (bufptr, string, tem); |
| 401 | bufptr[tem] = 0; |
| 402 | /* Trigger exit from the loop, but make sure we |
| 403 | return to the caller a value which will indicate |
| 404 | that the buffer was too small. */ |
| 405 | bufptr += bufsize; |
| 406 | bufsize = 0; |
| 407 | continue; |
| 408 | } |
| 409 | memcpy (bufptr, string, tem); |
| 410 | bufptr += tem; |
| 411 | bufsize -= tem; |
| 412 | if (minlen < 0) |
| 413 | { |
| 414 | while (minlen < - width && bufsize > 0) |
| 415 | { |
| 416 | *bufptr++ = ' '; |
| 417 | bufsize--; |
| 418 | minlen++; |
| 419 | } |
| 420 | minlen = 0; |
| 421 | } |
| 422 | continue; |
| 423 | |
| 424 | case 'c': |
| 425 | { |
| 426 | int chr = va_arg (ap, int); |
| 427 | tem = CHAR_STRING (chr, (unsigned char *) charbuf); |
| 428 | string = charbuf; |
| 429 | string[tem] = 0; |
| 430 | width = strwidth (string, tem); |
| 431 | if (fmtcpy[1] != 'c') |
| 432 | minlen = atoi (&fmtcpy[1]); |
| 433 | goto doit1; |
| 434 | } |
| 435 | |
| 436 | case '%': |
| 437 | fmt--; /* Drop thru and this % will be treated as normal */ |
| 438 | } |
| 439 | } |
| 440 | |
| 441 | { |
| 442 | /* Just some character; Copy it if the whole multi-byte form |
| 443 | fit in the buffer. */ |
| 444 | char *save_bufptr = bufptr; |
| 445 | |
| 446 | do { *bufptr++ = *fmt++; } |
| 447 | while (fmt < format_end && --bufsize > 0 && !CHAR_HEAD_P (*fmt)); |
| 448 | if (!CHAR_HEAD_P (*fmt)) |
| 449 | { |
| 450 | /* Truncate, but return value that will signal to caller |
| 451 | that the buffer was too small. */ |
| 452 | *save_bufptr = 0; |
| 453 | break; |
| 454 | } |
| 455 | } |
| 456 | }; |
| 457 | |
| 458 | /* If we had to malloc something, free it. */ |
| 459 | xfree (big_buffer); |
| 460 | |
| 461 | *bufptr = 0; /* Make sure our string ends with a '\0' */ |
| 462 | |
| 463 | SAFE_FREE (); |
| 464 | return bufptr - buffer; |
| 465 | } |
| 466 | |
| 467 | /* Format to an unbounded buffer BUF. This is like sprintf, except it |
| 468 | is not limited to returning an 'int' so it doesn't have a silly 2 |
| 469 | GiB limit on typical 64-bit hosts. However, it is limited to the |
| 470 | Emacs-style formats that doprnt supports. |
| 471 | |
| 472 | Return the number of bytes put into BUF, excluding the terminating |
| 473 | '\0'. */ |
| 474 | ptrdiff_t |
| 475 | esprintf (char *buf, char const *format, ...) |
| 476 | { |
| 477 | ptrdiff_t nbytes; |
| 478 | va_list ap; |
| 479 | va_start (ap, format); |
| 480 | nbytes = doprnt (buf, TYPE_MAXIMUM (ptrdiff_t), format, 0, ap); |
| 481 | va_end (ap); |
| 482 | return nbytes; |
| 483 | } |
| 484 | |
| 485 | #if defined HAVE_X_WINDOWS && defined USE_X_TOOLKIT |
| 486 | |
| 487 | /* Format to buffer *BUF of positive size *BUFSIZE, reallocating *BUF |
| 488 | and updating *BUFSIZE if the buffer is too small, and otherwise |
| 489 | behaving line esprintf. When reallocating, free *BUF unless it is |
| 490 | equal to NONHEAPBUF, and if BUFSIZE_MAX is nonnegative then signal |
| 491 | memory exhaustion instead of growing the buffer size past |
| 492 | BUFSIZE_MAX. */ |
| 493 | ptrdiff_t |
| 494 | exprintf (char **buf, ptrdiff_t *bufsize, |
| 495 | char const *nonheapbuf, ptrdiff_t bufsize_max, |
| 496 | char const *format, ...) |
| 497 | { |
| 498 | ptrdiff_t nbytes; |
| 499 | va_list ap; |
| 500 | va_start (ap, format); |
| 501 | nbytes = evxprintf (buf, bufsize, nonheapbuf, bufsize_max, format, ap); |
| 502 | va_end (ap); |
| 503 | return nbytes; |
| 504 | } |
| 505 | |
| 506 | #endif |
| 507 | |
| 508 | /* Act like exprintf, except take a va_list. */ |
| 509 | ptrdiff_t |
| 510 | evxprintf (char **buf, ptrdiff_t *bufsize, |
| 511 | char const *nonheapbuf, ptrdiff_t bufsize_max, |
| 512 | char const *format, va_list ap) |
| 513 | { |
| 514 | for (;;) |
| 515 | { |
| 516 | ptrdiff_t nbytes; |
| 517 | va_list ap_copy; |
| 518 | va_copy (ap_copy, ap); |
| 519 | nbytes = doprnt (*buf, *bufsize, format, 0, ap_copy); |
| 520 | va_end (ap_copy); |
| 521 | if (nbytes < *bufsize - 1) |
| 522 | return nbytes; |
| 523 | if (*buf != nonheapbuf) |
| 524 | { |
| 525 | xfree (*buf); |
| 526 | *buf = NULL; |
| 527 | } |
| 528 | *buf = xpalloc (NULL, bufsize, 1, bufsize_max, 1); |
| 529 | } |
| 530 | } |