Coverage Report

Created: 2025-06-10 07:06

/src/ghostpdl/psi/iscannum.c
Line
Count
Source (jump to first uncovered line)
1
/* Copyright (C) 2001-2024 Artifex Software, Inc.
2
   All Rights Reserved.
3
4
   This software is provided AS-IS with no warranty, either express or
5
   implied.
6
7
   This software is distributed under license and may not be copied,
8
   modified or distributed except as expressly authorized under the terms
9
   of the license contained in the file LICENSE in this distribution.
10
11
   Refer to licensing information at http://www.artifex.com or contact
12
   Artifex Software, Inc.,  39 Mesa Street, Suite 108A, San Francisco,
13
   CA 94129, USA, for further information.
14
*/
15
16
17
/* Number scanner for Ghostscript interpreter */
18
#include "math_.h"
19
#include "ghost.h"
20
#include "ierrors.h"
21
#include "scommon.h"
22
#include "iscan.h"
23
#include "iscannum.h"   /* defines interface */
24
#include "scanchar.h"
25
#include "store.h"
26
27
/*
28
 * Warning: this file has a "spaghetti" control structure.  But since this
29
 * code accounts for over 10% of the execution time of some PostScript
30
 * files, this is one of the few places we feel this is justified.
31
 */
32
33
/*
34
 * Scan a number.  If the number consumes the entire string, return 0;
35
 * if not, set *psp to the first character beyond the number and return 1.
36
 */
37
int
38
scan_number(const byte * str, const byte * end, int sign,
39
            ref * pref, const byte ** psp, int scanner_options)
40
249M
{
41
249M
    const byte *sp = str;
42
249M
#define GET_NEXT(cvar, sp, end_action)\
43
1.00G
  if (sp >= end) { end_action; } else cvar = *sp++
44
45
    /*
46
     * Powers of 10 up to 6 can be represented accurately as
47
     * a single-precision float.
48
     */
49
249M
#define NUM_POWERS_10 6
50
249M
    static const float powers_10[NUM_POWERS_10 + 1] = {
51
249M
        1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6
52
249M
    };
53
249M
    static const double neg_powers_10[NUM_POWERS_10 + 1] = {
54
249M
        1e0, 1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6
55
249M
    };
56
57
249M
    ps_int ival;
58
249M
    double dval;
59
249M
    int exp10;
60
249M
    int code = 0;
61
249M
    int c, d;
62
249M
    ps_int max_ps_int_scan, min_ps_int_scan;
63
249M
    const byte *const decoder = scan_char_decoder;
64
249M
#define IS_DIGIT(d, c)\
65
668M
  ((d = decoder[c]) < 10)
66
249M
#define WOULD_OVERFLOW(val, d, maxv)\
67
249M
  (val >= maxv / 10 && (val > maxv / 10 || d > (int64_t)(maxv % 10)))
68
69
249M
    GET_NEXT(c, sp, return_error(gs_error_syntaxerror));
70
249M
    if (!IS_DIGIT(d, c)) {
71
50.0M
        if (c != '.')
72
7.49k
            return_error(gs_error_syntaxerror);
73
        /* Might be a number starting with '.'. */
74
50.0M
        GET_NEXT(c, sp, return_error(gs_error_syntaxerror));
75
50.0M
        if (!IS_DIGIT(d, c))
76
50.0M
            return_error(gs_error_syntaxerror);
77
1.00k
        ival = 0;
78
1.00k
        goto i2r;
79
50.0M
    }
80
    /* Accumulate an integer in ival. */
81
    /* Do up to 4 digits without a loop, */
82
    /* since we know this can't overflow and since */
83
    /* most numbers have 4 (integer) digits or fewer. */
84
199M
    ival = d;
85
199M
    if (end - sp >= 3) { /* just check once */
86
195M
        if (!IS_DIGIT(d, (c = *sp))) {
87
41.7M
            sp++;
88
41.7M
            goto ind;
89
41.7M
        }
90
153M
        ival = ival * 10 + d;
91
153M
        if (!IS_DIGIT(d, (c = sp[1]))) {
92
148M
            sp += 2;
93
148M
            goto ind;
94
148M
        }
95
4.84M
        ival = ival * 10 + d;
96
4.84M
        sp += 3;
97
4.84M
        if (!IS_DIGIT(d, (c = sp[-1])))
98
2.51M
            goto ind;
99
2.32M
        ival = ival * 10 + d;
100
2.32M
    }
101
102
6.91M
    max_ps_int_scan = scanner_options & SCAN_CPSI_MODE ? MAX_PS_INT32 : MAX_PS_INT;
103
6.91M
    min_ps_int_scan = scanner_options & SCAN_CPSI_MODE ? MIN_PS_INT32 : MIN_PS_INT;
104
105
10.1M
    for (;; ival = ival * 10 + d) {
106
10.1M
        GET_NEXT(c, sp, goto iret);
107
5.21M
        if (!IS_DIGIT(d, c))
108
2.01M
            break;
109
3.20M
        if (WOULD_OVERFLOW(((ps_uint)ival), d, max_ps_int_scan)) {
110
2.11k
            if (ival == max_ps_int_scan / 10 && d == (max_ps_int_scan % 10) + 1 && sign < 0) {
111
8
                GET_NEXT(c, sp, c = EOFC);
112
8
                dval = -(double)min_ps_int_scan;
113
8
                if (c == 'e' || c == 'E') {
114
0
                    exp10 = 0;
115
0
                    goto fs;
116
8
                } else if (c == '.') {
117
2
                    GET_NEXT(c, sp, c = EOFC);
118
2
                    exp10 = 0;
119
2
                    goto fd;
120
6
                } else if (!IS_DIGIT(d, c)) {
121
3
                    ival = min_ps_int_scan;
122
3
                    break;
123
3
                }
124
8
            } else
125
2.10k
                dval = (double)ival;
126
2.11k
            goto l2d;
127
2.11k
        }
128
3.20M
    }
129
194M
  ind:        /* We saw a non-digit while accumulating an integer in ival. */
130
194M
    switch (c) {
131
1.26M
        case '.':
132
1.26M
            GET_NEXT(c, sp, c = EOFC);
133
1.26M
            goto i2r;
134
48.0M
        default:
135
48.0M
            *psp = sp;
136
48.0M
            code = 1;
137
48.0M
            break;
138
1
        case EOFC:
139
1
            break;
140
2.06k
        case 'e':
141
2.73k
        case 'E':
142
2.73k
            if (sign < 0)
143
112
                ival = -ival;
144
2.73k
            dval = (double)ival;
145
2.73k
            exp10 = 0;
146
2.73k
            goto fe;
147
145M
        case '#':
148
145M
            {
149
145M
                const int radix = ival;
150
145M
                ps_int uval = 0, imax;
151
152
145M
                if (sign || radix < min_radix || radix > max_radix)
153
43
                    return_error(gs_error_syntaxerror);
154
                /* Avoid multiplies for power-of-2 radix. */
155
145M
                if (!(radix & (radix - 1))) {
156
145M
                    int shift;
157
158
145M
                    switch (radix) {
159
81
                        case 2:
160
81
                            shift = 1, imax = MAX_PS_UINT >> 1;
161
81
                            break;
162
11
                        case 4:
163
11
                            shift = 2, imax = MAX_PS_UINT >> 2;
164
11
                            break;
165
7
                        case 8:
166
7
                            shift = 3, imax = MAX_PS_UINT >> 3;
167
7
                            break;
168
145M
                        case 16:
169
145M
                            shift = 4, imax = MAX_PS_UINT >> 4;
170
145M
                            break;
171
57
                        case 32:
172
57
                            shift = 5, imax = MAX_PS_UINT >> 5;
173
57
                            break;
174
0
                        default:  /* can't happen */
175
0
                            return_error(gs_error_rangecheck);
176
145M
                    }
177
692M
                    for (;; uval = (uval << shift) + d) {
178
692M
                        GET_NEXT(c, sp, break);
179
619M
                        d = decoder[c];
180
619M
                        if (d >= radix) {
181
72.7M
                            *psp = sp;
182
72.7M
                            code = 1;
183
72.7M
                            break;
184
72.7M
                        }
185
546M
                        if (uval > imax)
186
2
                            return_error(gs_error_limitcheck);
187
546M
                    }
188
145M
                } else {
189
102
                    ps_int irem = MAX_PS_UINT % radix;
190
191
102
                    imax = MAX_PS_UINT / radix;
192
530
                    for (;; uval = uval * radix + d) {
193
530
                        GET_NEXT(c, sp, break);
194
479
                        d = decoder[c];
195
479
                        if (d >= radix) {
196
45
                            *psp = sp;
197
45
                            code = 1;
198
45
                            break;
199
45
                        }
200
434
                        if (uval >= imax &&
201
434
                            (uval > imax || d > irem)
202
434
                            )
203
6
                            return_error(gs_error_limitcheck);
204
434
                    }
205
102
                }
206
145M
                if (scanner_options & SCAN_CPSI_MODE) {
207
0
                    ps_uint32 int1 = 0;
208
0
                    int1 |= (uval & 0xffffffff);
209
0
                    make_int(pref, (ps_int)((ps_int32)int1));
210
0
                }
211
145M
                else
212
145M
                    make_int(pref, uval);
213
214
145M
                return code;
215
145M
            }
216
194M
    }
217
52.9M
iret:
218
52.9M
    if (scanner_options & SCAN_CPSI_MODE) {
219
0
        make_int(pref, (sign < 0 ? (ps_int32)-ival : (ps_int32)ival));
220
0
    }
221
52.9M
    else {
222
52.9M
        make_int(pref, (sign < 0 ? (ps_int)-ival : (ps_int)ival));
223
52.9M
    }
224
52.9M
    return code;
225
226
    /* Accumulate a double in dval. */
227
2.11k
l2d:
228
2.11k
    exp10 = 0;
229
13.5k
    for (;;) {
230
13.5k
        dval = dval * 10 + d;
231
13.5k
        GET_NEXT(c, sp, c = EOFC);
232
13.5k
        if (!IS_DIGIT(d, c))
233
2.11k
            break;
234
13.5k
    }
235
2.11k
    switch (c) {
236
53
        case '.':
237
53
            GET_NEXT(c, sp, c = EOFC);
238
53
            exp10 = 0;
239
53
            goto fd;
240
1.21k
        default:
241
1.21k
            *psp = sp;
242
1.21k
            code = 1;
243
            /* falls through */
244
2.01k
        case EOFC:
245
2.01k
            if (sign < 0)
246
532
                dval = -dval;
247
2.01k
            goto rret;
248
10
        case 'e':
249
37
        case 'E':
250
37
            exp10 = 0;
251
37
            goto fs;
252
9
        case '#':
253
9
            return_error(gs_error_syntaxerror);
254
2.11k
    }
255
256
    /* We saw a '.' while accumulating an integer in ival. */
257
1.26M
i2r:
258
1.26M
    exp10 = 0;
259
4.59M
    while (IS_DIGIT(d, c) || c == '-') {
260
        /*
261
         * PostScript gives an error on numbers with a '-' following a '.'
262
         * Adobe Acrobat Reader (PDF) apparently doesn't treat this as an
263
         * error. Experiments show that the numbers following the '-' are
264
         * ignored, so we swallow the fractional part. SCAN_PDF_INV_NUM
265
         *  enables this compatibility kloodge.
266
         */
267
3.33M
        if (c == '-') {
268
13
            if ((SCAN_PDF_INV_NUM & scanner_options) == 0)
269
13
                break;
270
0
            do {
271
0
                GET_NEXT(c, sp, c = EOFC);
272
0
            } while (IS_DIGIT(d, c));
273
0
            break;
274
13
        }
275
3.33M
        if (WOULD_OVERFLOW(ival, d, max_int)) {
276
5.02k
            dval = (double)ival;
277
5.02k
            goto fd;
278
5.02k
        }
279
3.33M
        ival = ival * 10 + d;
280
3.33M
        exp10--;
281
3.33M
        GET_NEXT(c, sp, c = EOFC);
282
3.33M
    }
283
1.25M
    if (sign < 0)
284
118k
        ival = -ival;
285
    /* Take a shortcut for the common case */
286
1.25M
    if (!(c == 'e' || c == 'E' || exp10 < -NUM_POWERS_10)) { /* Check for trailing garbage */
287
1.25M
        if (c != EOFC)
288
1.06M
            *psp = sp, code = 1;
289
1.25M
        make_real(pref, ival * neg_powers_10[-exp10]);
290
1.25M
        return code;
291
1.25M
    }
292
1.81k
    dval = (double)ival;
293
1.81k
    goto fe;
294
295
    /* Now we are accumulating a double in dval. */
296
5.08k
fd:
297
27.1k
    while (IS_DIGIT(d, c)) {
298
22.0k
        dval = dval * 10 + d;
299
22.0k
        exp10--;
300
22.0k
        GET_NEXT(c, sp, c = EOFC);
301
22.0k
    }
302
5.11k
fs:
303
5.11k
    if (sign < 0)
304
1.10k
        dval = -dval;
305
9.66k
fe:
306
    /* Now dval contains the value, negated if necessary. */
307
9.66k
    switch (c) {
308
3.26k
        case 'e':
309
4.27k
        case 'E':
310
4.27k
            {     /* Check for a following exponent. */
311
4.27k
                int esign = 0;
312
4.27k
                int iexp;
313
314
4.27k
                GET_NEXT(c, sp, return_error(gs_error_syntaxerror));
315
4.17k
                switch (c) {
316
625
                    case '-':
317
625
                        esign = 1;
318
                        /* fall through */
319
670
                    case '+':
320
670
                        GET_NEXT(c, sp, return_error(gs_error_syntaxerror));
321
4.17k
                }
322
                /* Scan the exponent.  We limit it arbitrarily to 999. */
323
4.16k
                if (!IS_DIGIT(d, c))
324
651
                    return_error(gs_error_syntaxerror);
325
3.51k
                iexp = d;
326
4.60k
                for (;; iexp = iexp * 10 + d) {
327
4.60k
                    GET_NEXT(c, sp, break);
328
3.93k
                    if (!IS_DIGIT(d, c)) {
329
2.84k
                        *psp = sp;
330
2.84k
                        code = 1;
331
2.84k
                        break;
332
2.84k
                    }
333
1.09k
                    if (iexp > 99)
334
4
                        return_error(gs_error_limitcheck);
335
1.09k
                }
336
3.50k
                if (esign)
337
607
                    exp10 -= iexp;
338
2.89k
                else
339
2.89k
                    exp10 += iexp;
340
3.50k
                break;
341
3.51k
            }
342
5.14k
        default:
343
5.14k
            *psp = sp;
344
5.14k
            code = 1;
345
5.39k
        case EOFC:
346
5.39k
            ;
347
9.66k
    }
348
    /* Compute dval * 10^exp10. */
349
8.89k
    if (exp10 > 0) {
350
3.90k
        while (exp10 > NUM_POWERS_10)
351
1.46k
            dval *= powers_10[NUM_POWERS_10],
352
1.46k
                exp10 -= NUM_POWERS_10;
353
2.44k
        dval *= powers_10[exp10];
354
6.45k
    } else if (exp10 < 0) {
355
31.2k
        while (exp10 < -NUM_POWERS_10)
356
25.3k
            dval /= powers_10[NUM_POWERS_10],
357
25.3k
                exp10 += NUM_POWERS_10;
358
5.97k
        dval /= powers_10[-exp10];
359
5.97k
    }
360
    /*
361
     * Check for an out-of-range result.  Currently we don't check for
362
     * absurdly large numbers of digits in the accumulation loops,
363
     * but we should.
364
     */
365
8.89k
    if (dval >= 0) {
366
7.61k
        if (dval > MAX_FLOAT)
367
12
            return_error(gs_error_limitcheck);
368
7.61k
    } else {
369
1.28k
        if (dval < -MAX_FLOAT)
370
7
            return_error(gs_error_limitcheck);
371
1.28k
    }
372
10.8k
rret:
373
10.8k
    make_real(pref, dval);
374
10.8k
    return code;
375
8.89k
}