*: add most of the required setup_common_bufsiz() calls
[oweals/busybox.git] / findutils / grep.c
index 030e62461d2fc14768857ad055d87c4cd06299d1..b072cd44126f7f4abb16dcc0fa3edde752d47226 100644 (file)
@@ -5,7 +5,7 @@
  * Copyright (C) 1999,2000,2001 by Lineo, inc. and Mark Whitley
  * Copyright (C) 1999,2000,2001 by Mark Whitley <markw@codepoet.org>
  *
- * Licensed under the GPL v2 or later, see the file LICENSE in this tarball.
+ * Licensed under GPLv2 or later, see file LICENSE in this source tree.
  */
 /* BB_AUDIT SUSv3 defects - unsupported option -x "match whole line only". */
 /* BB_AUDIT GNU defects - always acts as -a.  */
  * 2004,2006 (C) Vladimir Oleynik <dzo@simtreas.ru> -
  * correction "-e pattern1 -e pattern2" logic and more optimizations.
  * precompiled regex
- */
-/*
+ *
  * (C) 2006 Jac Goudsmit added -o option
  */
 
+//applet:IF_GREP(APPLET(grep, BB_DIR_BIN, BB_SUID_DROP))
+//applet:IF_FEATURE_GREP_EGREP_ALIAS(APPLET_ODDNAME(egrep, grep, BB_DIR_BIN, BB_SUID_DROP, egrep))
+//applet:IF_FEATURE_GREP_FGREP_ALIAS(APPLET_ODDNAME(fgrep, grep, BB_DIR_BIN, BB_SUID_DROP, fgrep))
+
+//kbuild:lib-$(CONFIG_GREP) += grep.o
+
+//config:config GREP
+//config:      bool "grep"
+//config:      default y
+//config:      help
+//config:        grep is used to search files for a specified pattern.
+//config:
+//config:config FEATURE_GREP_EGREP_ALIAS
+//config:      bool "Enable extended regular expressions (egrep & grep -E)"
+//config:      default y
+//config:      depends on GREP
+//config:      help
+//config:        Enabled support for extended regular expressions. Extended
+//config:        regular expressions allow for alternation (foo|bar), grouping,
+//config:        and various repetition operators.
+//config:
+//config:config FEATURE_GREP_FGREP_ALIAS
+//config:      bool "Alias fgrep to grep -F"
+//config:      default y
+//config:      depends on GREP
+//config:      help
+//config:        fgrep sees the search pattern as a normal string rather than
+//config:        regular expressions.
+//config:        grep -F always works, this just creates the fgrep alias.
+//config:
+//config:config FEATURE_GREP_CONTEXT
+//config:      bool "Enable before and after context flags (-A, -B and -C)"
+//config:      default y
+//config:      depends on GREP
+//config:      help
+//config:        Print the specified number of leading (-B) and/or trailing (-A)
+//config:        context surrounding our matching lines.
+//config:        Print the specified number of context lines (-C).
+
 #include "libbb.h"
+#include "common_bufsiz.h"
 #include "xregex.h"
 
+
 /* options */
+//usage:#define grep_trivial_usage
+//usage:       "[-HhnlLoqvsriw"
+//usage:       "F"
+//usage:       IF_FEATURE_GREP_EGREP_ALIAS("E")
+//usage:       IF_EXTRA_COMPAT("z")
+//usage:       "] [-m N] "
+//usage:       IF_FEATURE_GREP_CONTEXT("[-A/B/C N] ")
+//usage:       "PATTERN/-e PATTERN.../-f FILE [FILE]..."
+//usage:#define grep_full_usage "\n\n"
+//usage:       "Search for PATTERN in FILEs (or stdin)\n"
+//usage:     "\n       -H      Add 'filename:' prefix"
+//usage:     "\n       -h      Do not add 'filename:' prefix"
+//usage:     "\n       -n      Add 'line_no:' prefix"
+//usage:     "\n       -l      Show only names of files that match"
+//usage:     "\n       -L      Show only names of files that don't match"
+//usage:     "\n       -c      Show only count of matching lines"
+//usage:     "\n       -o      Show only the matching part of line"
+//usage:     "\n       -q      Quiet. Return 0 if PATTERN is found, 1 otherwise"
+//usage:     "\n       -v      Select non-matching lines"
+//usage:     "\n       -s      Suppress open and read errors"
+//usage:     "\n       -r      Recurse"
+//usage:     "\n       -i      Ignore case"
+//usage:     "\n       -w      Match whole words only"
+//usage:     "\n       -x      Match whole lines only"
+//usage:     "\n       -F      PATTERN is a literal (not regexp)"
+//usage:       IF_FEATURE_GREP_EGREP_ALIAS(
+//usage:     "\n       -E      PATTERN is an extended regexp"
+//usage:       )
+//usage:       IF_EXTRA_COMPAT(
+//usage:     "\n       -z      Input is NUL terminated"
+//usage:       )
+//usage:     "\n       -m N    Match up to N times per file"
+//usage:       IF_FEATURE_GREP_CONTEXT(
+//usage:     "\n       -A N    Print N lines of trailing context"
+//usage:     "\n       -B N    Print N lines of leading context"
+//usage:     "\n       -C N    Same as '-A N -B N'"
+//usage:       )
+//usage:     "\n       -e PTRN Pattern to match"
+//usage:     "\n       -f FILE Read pattern from file"
+//usage:
+//usage:#define grep_example_usage
+//usage:       "$ grep root /etc/passwd\n"
+//usage:       "root:x:0:0:root:/root:/bin/bash\n"
+//usage:       "$ grep ^[rR]oo. /etc/passwd\n"
+//usage:       "root:x:0:0:root:/root:/bin/bash\n"
+//usage:
+//usage:#define egrep_trivial_usage NOUSAGE_STR
+//usage:#define egrep_full_usage ""
+//usage:#define fgrep_trivial_usage NOUSAGE_STR
+//usage:#define fgrep_full_usage ""
+
 #define OPTSTR_GREP \
-       "lnqvscFiHhe:f:Lorm:" \
-       USE_FEATURE_GREP_CONTEXT("A:B:C:") \
-       USE_FEATURE_GREP_EGREP_ALIAS("E") \
-       USE_DESKTOP("w") \
+       "lnqvscFiHhe:f:Lorm:wx" \
+       IF_FEATURE_GREP_CONTEXT("A:B:C:") \
+       IF_FEATURE_GREP_EGREP_ALIAS("E") \
+       IF_EXTRA_COMPAT("z") \
        "aI"
 /* ignored: -a "assume all files to be text" */
 /* ignored: -I "assume binary files have no matches" */
-
 enum {
        OPTBIT_l, /* list matched file names only */
        OPTBIT_n, /* print line# */
@@ -49,11 +139,13 @@ enum {
        OPTBIT_o, /* show only matching parts of lines */
        OPTBIT_r, /* recurse dirs */
        OPTBIT_m, /* -m MAX_MATCHES */
-       USE_FEATURE_GREP_CONTEXT(    OPTBIT_A ,) /* -A NUM: after-match context */
-       USE_FEATURE_GREP_CONTEXT(    OPTBIT_B ,) /* -B NUM: before-match context */
-       USE_FEATURE_GREP_CONTEXT(    OPTBIT_C ,) /* -C NUM: -A and -B combined */
-       USE_FEATURE_GREP_EGREP_ALIAS(OPTBIT_E ,) /* extended regexp */
-       USE_DESKTOP(                 OPTBIT_w ,) /* whole word match */
+       OPTBIT_w, /* -w whole word match */
+       OPTBIT_x, /* -x whole line match */
+       IF_FEATURE_GREP_CONTEXT(    OPTBIT_A ,) /* -A NUM: after-match context */
+       IF_FEATURE_GREP_CONTEXT(    OPTBIT_B ,) /* -B NUM: before-match context */
+       IF_FEATURE_GREP_CONTEXT(    OPTBIT_C ,) /* -C NUM: -A and -B combined */
+       IF_FEATURE_GREP_EGREP_ALIAS(OPTBIT_E ,) /* extended regexp */
+       IF_EXTRA_COMPAT(            OPTBIT_z ,) /* input is NUL terminated */
        OPT_l = 1 << OPTBIT_l,
        OPT_n = 1 << OPTBIT_n,
        OPT_q = 1 << OPTBIT_q,
@@ -70,11 +162,13 @@ enum {
        OPT_o = 1 << OPTBIT_o,
        OPT_r = 1 << OPTBIT_r,
        OPT_m = 1 << OPTBIT_m,
-       OPT_A = USE_FEATURE_GREP_CONTEXT(    (1 << OPTBIT_A)) + 0,
-       OPT_B = USE_FEATURE_GREP_CONTEXT(    (1 << OPTBIT_B)) + 0,
-       OPT_C = USE_FEATURE_GREP_CONTEXT(    (1 << OPTBIT_C)) + 0,
-       OPT_E = USE_FEATURE_GREP_EGREP_ALIAS((1 << OPTBIT_E)) + 0,
-       OPT_w = USE_DESKTOP(                 (1 << OPTBIT_w)) + 0,
+       OPT_w = 1 << OPTBIT_w,
+       OPT_x = 1 << OPTBIT_x,
+       OPT_A = IF_FEATURE_GREP_CONTEXT(    (1 << OPTBIT_A)) + 0,
+       OPT_B = IF_FEATURE_GREP_CONTEXT(    (1 << OPTBIT_B)) + 0,
+       OPT_C = IF_FEATURE_GREP_CONTEXT(    (1 << OPTBIT_C)) + 0,
+       OPT_E = IF_FEATURE_GREP_EGREP_ALIAS((1 << OPTBIT_E)) + 0,
+       OPT_z = IF_EXTRA_COMPAT(            (1 << OPTBIT_z)) + 0,
 };
 
 #define PRINT_FILES_WITH_MATCHES    (option_mask32 & OPT_l)
@@ -84,10 +178,15 @@ enum {
 #define PRINT_MATCH_COUNTS          (option_mask32 & OPT_c)
 #define FGREP_FLAG                  (option_mask32 & OPT_F)
 #define PRINT_FILES_WITHOUT_MATCHES (option_mask32 & OPT_L)
+#define NUL_DELIMITED               (option_mask32 & OPT_z)
 
 struct globals {
        int max_matches;
+#if !ENABLE_EXTRA_COMPAT
        int reflags;
+#else
+       RE_TRANSLATE_TYPE case_fold; /* RE_TRANSLATE_TYPE is [[un]signed] char* */
+#endif
        smalluint invert_search;
        smalluint print_filename;
        smalluint open_errors;
@@ -96,20 +195,33 @@ struct globals {
        int lines_before;
        int lines_after;
        char **before_buf;
+       IF_EXTRA_COMPAT(size_t *before_buf_size;)
        int last_line_printed;
 #endif
        /* globals used internally */
        llist_t *pattern_head;   /* growable list of patterns to match */
        const char *cur_file;    /* the current file we are reading */
-};
-#define G (*(struct globals*)&bb_common_bufsiz1)
+} FIX_ALIASING;
+#define G (*(struct globals*)bb_common_bufsiz1)
 #define INIT_G() do { \
-       struct G_sizecheck { \
-               char G_sizecheck[sizeof(G) > COMMON_BUFSIZE ? -1 : 1]; \
-       }; \
+       setup_common_bufsiz(); \
+       BUILD_BUG_ON(sizeof(G) > COMMON_BUFSIZE); \
 } while (0)
 #define max_matches       (G.max_matches         )
-#define reflags           (G.reflags             )
+#if !ENABLE_EXTRA_COMPAT
+# define reflags          (G.reflags             )
+#else
+# define case_fold        (G.case_fold           )
+/* http://www.delorie.com/gnu/docs/regex/regex_46.html */
+# define reflags           re_syntax_options
+# undef REG_NOSUB
+# undef REG_EXTENDED
+# undef REG_ICASE
+# define REG_NOSUB    bug:is:here /* should not be used */
+/* Just RE_SYNTAX_EGREP is not enough, need to enable {n[,[m]]} too */
+# define REG_EXTENDED (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES)
+# define REG_ICASE    bug:is:here /* should not be used */
+#endif
 #define invert_search     (G.invert_search       )
 #define print_filename    (G.print_filename      )
 #define open_errors       (G.open_errors         )
@@ -117,6 +229,7 @@ struct globals {
 #define lines_before      (G.lines_before        )
 #define lines_after       (G.lines_after         )
 #define before_buf        (G.before_buf          )
+#define before_buf_size   (G.before_buf_size     )
 #define last_line_printed (G.last_line_printed   )
 #define pattern_head      (G.pattern_head        )
 #define cur_file          (G.cur_file            )
@@ -124,14 +237,24 @@ struct globals {
 
 typedef struct grep_list_data_t {
        char *pattern;
-       regex_t preg;
+/* for GNU regex, matched_range must be persistent across grep_file() calls */
+#if !ENABLE_EXTRA_COMPAT
+       regex_t compiled_regex;
+       regmatch_t matched_range;
+#else
+       struct re_pattern_buffer compiled_regex;
+       struct re_registers matched_range;
+#endif
 #define ALLOCATED 1
 #define COMPILED 2
        int flg_mem_alocated_compiled;
 } grep_list_data_t;
 
-
-static void print_line(const char *line, int linenum, char decoration)
+#if !ENABLE_EXTRA_COMPAT
+#define print_line(line, line_len, linenum, decoration) \
+       print_line(line, linenum, decoration)
+#endif
+static void print_line(const char *line, size_t line_len, int linenum, char decoration)
 {
 #if ENABLE_FEATURE_GREP_CONTEXT
        /* Happens when we go to next file, immediately hit match
@@ -139,8 +262,9 @@ static void print_line(const char *line, int linenum, char decoration)
        if (linenum < 1)
                return;
        /* possibly print the little '--' separator */
-       if ((lines_before || lines_after) && did_print_line &&
-                       last_line_printed != linenum - 1) {
+       if ((lines_before || lines_after) && did_print_line
+        && last_line_printed != linenum - 1
+       ) {
                puts("--");
        }
        /* guard against printing "--" before first line of first file */
@@ -152,26 +276,66 @@ static void print_line(const char *line, int linenum, char decoration)
        if (PRINT_LINE_NUM)
                printf("%i%c", linenum, decoration);
        /* Emulate weird GNU grep behavior with -ov */
-       if ((option_mask32 & (OPT_v|OPT_o)) != (OPT_v|OPT_o))
+       if ((option_mask32 & (OPT_v|OPT_o)) != (OPT_v|OPT_o)) {
+#if !ENABLE_EXTRA_COMPAT
                puts(line);
+#else
+               fwrite(line, 1, line_len, stdout);
+               putchar(NUL_DELIMITED ? '\0' : '\n');
+#endif
+       }
 }
 
-static int grep_file(FILE *file)
+#if ENABLE_EXTRA_COMPAT
+/* Unlike getline, this one removes trailing '\n' */
+static ssize_t FAST_FUNC bb_getline(char **line_ptr, size_t *line_alloc_len, FILE *file)
 {
+       ssize_t res_sz;
        char *line;
+       int delim = (NUL_DELIMITED ? '\0' : '\n');
+
+       res_sz = getdelim(line_ptr, line_alloc_len, delim, file);
+       line = *line_ptr;
+
+       if (res_sz > 0) {
+               if (line[res_sz - 1] == delim)
+                       line[--res_sz] = '\0';
+       } else {
+               free(line); /* uclibc allocates a buffer even on EOF. WTF? */
+       }
+       return res_sz;
+}
+#endif
+
+static int grep_file(FILE *file)
+{
        smalluint found;
        int linenum = 0;
        int nmatches = 0;
-       regmatch_t regmatch;
+#if !ENABLE_EXTRA_COMPAT
+       char *line;
+#else
+       char *line = NULL;
+       ssize_t line_len;
+       size_t line_alloc_len;
+# define rm_so start[0]
+# define rm_eo end[0]
+#endif
 #if ENABLE_FEATURE_GREP_CONTEXT
        int print_n_lines_after = 0;
        int curpos = 0; /* track where we are in the circular 'before' buffer */
        int idx = 0; /* used for iteration through the circular buffer */
 #else
        enum { print_n_lines_after = 0 };
-#endif /* ENABLE_FEATURE_GREP_CONTEXT */
+#endif
 
-       while ((line = xmalloc_fgetline(file)) != NULL) {
+       while (
+#if !ENABLE_EXTRA_COMPAT
+               (line = xmalloc_fgetline(file)) != NULL
+#else
+               (line_len = bb_getline(&line, &line_alloc_len, file)) >= 0
+#endif
+       ) {
                llist_t *pattern_ptr = pattern_head;
                grep_list_data_t *gl = gl; /* for gcc */
 
@@ -180,25 +344,108 @@ static int grep_file(FILE *file)
                while (pattern_ptr) {
                        gl = (grep_list_data_t *)pattern_ptr->data;
                        if (FGREP_FLAG) {
-                               found |= (strstr(line, gl->pattern) != NULL);
+                               char *match;
+                               char *str = line;
+ opt_f_again:
+                               match = ((option_mask32 & OPT_i)
+                                       ? strcasestr(str, gl->pattern)
+                                       : strstr(str, gl->pattern)
+                                       );
+                               if (match) {
+                                       if (option_mask32 & OPT_x) {
+                                               if (match != str)
+                                                       goto opt_f_not_found;
+                                               if (str[strlen(gl->pattern)] != '\0')
+                                                       goto opt_f_not_found;
+                                       } else
+                                       if (option_mask32 & OPT_w) {
+                                               char c = (match != str) ? match[-1] : ' ';
+                                               if (!isalnum(c) && c != '_') {
+                                                       c = match[strlen(gl->pattern)];
+                                                       if (!c || (!isalnum(c) && c != '_'))
+                                                               goto opt_f_found;
+                                               }
+                                               str = match + 1;
+                                               goto opt_f_again;
+                                       }
+ opt_f_found:
+                                       found = 1;
+ opt_f_not_found: ;
+                               }
                        } else {
+#if ENABLE_EXTRA_COMPAT
+                               unsigned start_pos;
+#else
+                               int match_flg;
+#endif
+                               char *match_at;
+
                                if (!(gl->flg_mem_alocated_compiled & COMPILED)) {
                                        gl->flg_mem_alocated_compiled |= COMPILED;
-                                       xregcomp(&(gl->preg), gl->pattern, reflags);
+#if !ENABLE_EXTRA_COMPAT
+                                       xregcomp(&gl->compiled_regex, gl->pattern, reflags);
+#else
+                                       memset(&gl->compiled_regex, 0, sizeof(gl->compiled_regex));
+                                       gl->compiled_regex.translate = case_fold; /* for -i */
+                                       if (re_compile_pattern(gl->pattern, strlen(gl->pattern), &gl->compiled_regex))
+                                               bb_error_msg_and_die("bad regex '%s'", gl->pattern);
+#endif
                                }
-                               regmatch.rm_so = 0;
-                               regmatch.rm_eo = 0;
-                               if (regexec(&(gl->preg), line, 1, &regmatch, 0) == 0) {
-                                       if (!(option_mask32 & OPT_w))
+#if !ENABLE_EXTRA_COMPAT
+                               gl->matched_range.rm_so = 0;
+                               gl->matched_range.rm_eo = 0;
+                               match_flg = 0;
+#else
+                               start_pos = 0;
+#endif
+                               match_at = line;
+ opt_w_again:
+//bb_error_msg("'%s' start_pos:%d line_len:%d", match_at, start_pos, line_len);
+                               if (
+#if !ENABLE_EXTRA_COMPAT
+                                       regexec(&gl->compiled_regex, match_at, 1, &gl->matched_range, match_flg) == 0
+#else
+                                       re_search(&gl->compiled_regex, match_at, line_len,
+                                                       start_pos, /*range:*/ line_len,
+                                                       &gl->matched_range) >= 0
+#endif
+                               ) {
+                                       if (option_mask32 & OPT_x) {
+                                               found = (gl->matched_range.rm_so == 0
+                                                        && match_at[gl->matched_range.rm_eo] == '\0');
+                                       } else
+                                       if (!(option_mask32 & OPT_w)) {
                                                found = 1;
-                                       else {
+                                       else {
                                                char c = ' ';
-                                               if (regmatch.rm_so)
-                                                       c = line[regmatch.rm_so - 1];
+                                               if (match_at > line || gl->matched_range.rm_so != 0) {
+                                                       c = match_at[gl->matched_range.rm_so - 1];
+                                               }
                                                if (!isalnum(c) && c != '_') {
-                                                       c = line[regmatch.rm_eo];
-                                                       if (!c || (!isalnum(c) && c != '_'))
-                                                               found = 1;
+                                                       c = match_at[gl->matched_range.rm_eo];
+                                               }
+                                               if (!isalnum(c) && c != '_') {
+                                                       found = 1;
+                                               } else {
+                       /*
+                        * Why check gl->matched_range.rm_eo?
+                        * Zero-length match makes -w skip the line:
+                        * "echo foo | grep ^" prints "foo",
+                        * "echo foo | grep -w ^" prints nothing.
+                        * Without such check, we can loop forever.
+                        */
+#if !ENABLE_EXTRA_COMPAT
+                                                       if (gl->matched_range.rm_eo != 0) {
+                                                               match_at += gl->matched_range.rm_eo;
+                                                               match_flg |= REG_NOTBOL;
+                                                               goto opt_w_again;
+                                                       }
+#else
+                                                       if (gl->matched_range.rm_eo > start_pos) {
+                                                               start_pos = gl->matched_range.rm_eo;
+                                                               goto opt_w_again;
+                                                       }
+#endif
                                                }
                                        }
                                }
@@ -261,7 +508,7 @@ static int grep_file(FILE *file)
 
                                        /* now print each line in the buffer, clearing them as we go */
                                        while (before_buf[idx] != NULL) {
-                                               print_line(before_buf[idx], first_buf_entry_line_num, '-');
+                                               print_line(before_buf[idx], before_buf_size[idx], first_buf_entry_line_num, '-');
                                                free(before_buf[idx]);
                                                before_buf[idx] = NULL;
                                                idx = (idx + 1) % lines_before;
@@ -277,13 +524,36 @@ static int grep_file(FILE *file)
                                                /* -Fo just prints the pattern
                                                 * (unless -v: -Fov doesnt print anything at all) */
                                                if (found)
-                                                       print_line(gl->pattern, linenum, ':');
-                                       } else {
-                                               line[regmatch.rm_eo] = '\0';
-                                               print_line(line + regmatch.rm_so, linenum, ':');
+                                                       print_line(gl->pattern, strlen(gl->pattern), linenum, ':');
+                                       } else while (1) {
+                                               unsigned start = gl->matched_range.rm_so;
+                                               unsigned end = gl->matched_range.rm_eo;
+                                               unsigned len = end - start;
+                                               char old = line[end];
+                                               line[end] = '\0';
+                                               /* Empty match is not printed: try "echo test | grep -o ''" */
+                                               if (len != 0)
+                                                       print_line(line + start, len, linenum, ':');
+                                               if (old == '\0')
+                                                       break;
+                                               line[end] = old;
+                                               if (len == 0)
+                                                       end++;
+#if !ENABLE_EXTRA_COMPAT
+                                               if (regexec(&gl->compiled_regex, line + end,
+                                                               1, &gl->matched_range, REG_NOTBOL) != 0)
+                                                       break;
+                                               gl->matched_range.rm_so += end;
+                                               gl->matched_range.rm_eo += end;
+#else
+                                               if (re_search(&gl->compiled_regex, line, line_len,
+                                                               end, line_len - end,
+                                                               &gl->matched_range) < 0)
+                                                       break;
+#endif
                                        }
                                } else {
-                                       print_line(line, linenum, ':');
+                                       print_line(line, line_len, linenum, ':');
                                }
                        }
                }
@@ -291,12 +561,13 @@ static int grep_file(FILE *file)
                else { /* no match */
                        /* if we need to print some context lines after the last match, do so */
                        if (print_n_lines_after) {
-                               print_line(line, linenum, '-');
+                               print_line(line, strlen(line), linenum, '-');
                                print_n_lines_after--;
                        } else if (lines_before) {
                                /* Add the line to the circular 'before' buffer */
                                free(before_buf[curpos]);
                                before_buf[curpos] = line;
+                               IF_EXTRA_COMPAT(before_buf_size[curpos] = line_len;)
                                curpos = (curpos + 1) % lines_before;
                                /* avoid free(line) - we took the line */
                                line = NULL;
@@ -304,13 +575,17 @@ static int grep_file(FILE *file)
                }
 
 #endif /* ENABLE_FEATURE_GREP_CONTEXT */
+#if !ENABLE_EXTRA_COMPAT
                free(line);
-
+#endif
                /* Did we print all context after last requested match? */
                if ((option_mask32 & OPT_m)
-                && !print_n_lines_after && nmatches == max_matches)
+                && !print_n_lines_after
+                && nmatches == max_matches
+               ) {
                        break;
-       }
+               }
+       } /* while (read line) */
 
        /* special-case file post-processing for options where we don't print line
         * matches, just filenames and possibly match counts */
@@ -353,20 +628,20 @@ static char *add_grep_list_data(char *pattern)
 
 static void load_regexes_from_file(llist_t *fopt)
 {
-       char *line;
-       FILE *f;
-
        while (fopt) {
+               char *line;
+               FILE *fp;
                llist_t *cur = fopt;
                char *ffile = cur->data;
 
                fopt = cur->link;
                free(cur);
-               f = xfopen_stdin(ffile);
-               while ((line = xmalloc_fgetline(f)) != NULL) {
+               fp = xfopen_stdin(ffile);
+               while ((line = xmalloc_fgetline(fp)) != NULL) {
                        llist_add_to(&pattern_head,
                                new_grep_list_data(line, ALLOCATED));
                }
+               fclose_if_not_stdin(fp);
        }
 }
 
@@ -393,7 +668,7 @@ static int grep_dir(const char *dir)
        int matched = 0;
        recursive_action(dir,
                /* recurse=yes */ ACTION_RECURSE |
-               /* followLinks=no */
+               /* followLinks=command line only */ ACTION_FOLLOWLINKS_L0 |
                /* depthFirst=yes */ ACTION_DEPTHFIRST,
                /* fileAction= */ file_action_grep,
                /* dirAction= */ NULL,
@@ -403,40 +678,48 @@ static int grep_dir(const char *dir)
 }
 
 int grep_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
-int grep_main(int argc, char **argv)
+int grep_main(int argc UNUSED_PARAM, char **argv)
 {
        FILE *file;
        int matched;
        llist_t *fopt = NULL;
+#if ENABLE_FEATURE_GREP_CONTEXT
+       int Copt, opts;
+#endif
+
+       /* For grep, exitcode of 1 is "not found". Other errors are 2: */
+       xfunc_error_retval = 2;
 
        /* do normal option parsing */
 #if ENABLE_FEATURE_GREP_CONTEXT
-       int Copt;
-
        /* -H unsets -h; -C unsets -A,-B; -e,-f are lists;
         * -m,-A,-B,-C have numeric param */
        opt_complementary = "H-h:C-AB:e::f::m+:A+:B+:C+";
-       getopt32(argv,
+       opts = getopt32(argv,
                OPTSTR_GREP,
                &pattern_head, &fopt, &max_matches,
                &lines_after, &lines_before, &Copt);
 
-       if (option_mask32 & OPT_C) {
+       if (opts & OPT_C) {
                /* -C unsets prev -A and -B, but following -A or -B
-                  may override it */
-               if (!(option_mask32 & OPT_A)) /* not overridden */
+                * may override it */
+               if (!(opts & OPT_A)) /* not overridden */
                        lines_after = Copt;
-               if (!(option_mask32 & OPT_B)) /* not overridden */
+               if (!(opts & OPT_B)) /* not overridden */
                        lines_before = Copt;
-               //option_mask32 |= OPT_A|OPT_B; /* for parser */
        }
        /* sanity checks */
-       if (option_mask32 & (OPT_c|OPT_q|OPT_l|OPT_L)) {
+       if (opts & (OPT_c|OPT_q|OPT_l|OPT_L)) {
                option_mask32 &= ~OPT_n;
                lines_before = 0;
                lines_after = 0;
-       } else if (lines_before > 0)
-               before_buf = xzalloc(lines_before * sizeof(char *));
+       } else if (lines_before > 0) {
+               if (lines_before > INT_MAX / sizeof(long long))
+                       lines_before = INT_MAX / sizeof(long long);
+               /* overflow in (lines_before * sizeof(x)) is prevented (above) */
+               before_buf = xzalloc(lines_before * sizeof(before_buf[0]));
+               IF_EXTRA_COMPAT(before_buf_size = xzalloc(lines_before * sizeof(before_buf_size[0]));)
+       }
 #else
        /* with auto sanity checks */
        /* -H unsets -h; -c,-q or -l unset -n; -e,-f are lists; -m N */
@@ -446,33 +729,53 @@ int grep_main(int argc, char **argv)
 #endif
        invert_search = ((option_mask32 & OPT_v) != 0); /* 0 | 1 */
 
-       if (pattern_head != NULL) {
-               /* convert char **argv to grep_list_data_t */
+       {       /* convert char **argv to grep_list_data_t */
                llist_t *cur;
-
                for (cur = pattern_head; cur; cur = cur->link)
                        cur->data = new_grep_list_data(cur->data, 0);
        }
-       if (option_mask32 & OPT_f)
+       if (option_mask32 & OPT_f) {
                load_regexes_from_file(fopt);
+               if (!pattern_head) { /* -f EMPTY_FILE? */
+                       /* GNU grep treats it as "nothing matches" */
+                       llist_add_to(&pattern_head, new_grep_list_data((char*) "", 0));
+                       invert_search ^= 1;
+               }
+       }
 
        if (ENABLE_FEATURE_GREP_FGREP_ALIAS && applet_name[0] == 'f')
                option_mask32 |= OPT_F;
 
-       if (!(option_mask32 & (OPT_o | OPT_w)))
+#if !ENABLE_EXTRA_COMPAT
+       if (!(option_mask32 & (OPT_o | OPT_w | OPT_x)))
                reflags = REG_NOSUB;
+#endif
 
        if (ENABLE_FEATURE_GREP_EGREP_ALIAS
         && (applet_name[0] == 'e' || (option_mask32 & OPT_E))
        ) {
                reflags |= REG_EXTENDED;
        }
+#if ENABLE_EXTRA_COMPAT
+       else {
+               reflags = RE_SYNTAX_GREP;
+       }
+#endif
 
-       if (option_mask32 & OPT_i)
+       if (option_mask32 & OPT_i) {
+#if !ENABLE_EXTRA_COMPAT
                reflags |= REG_ICASE;
+#else
+               int i;
+               case_fold = xmalloc(256);
+               for (i = 0; i < 256; i++)
+                       case_fold[i] = (unsigned char)i;
+               for (i = 'a'; i <= 'z'; i++)
+                       case_fold[i] = (unsigned char)(i - ('a' - 'A'));
+#endif
+       }
 
        argv += optind;
-       argc -= optind;
 
        /* if we didn't get a pattern from -e and no command file was specified,
         * first parameter should be the pattern. no pattern, no worky */
@@ -482,12 +785,11 @@ int grep_main(int argc, char **argv)
                        bb_show_usage();
                pattern = new_grep_list_data(*argv++, 0);
                llist_add_to(&pattern_head, pattern);
-               argc--;
        }
 
        /* argv[0..(argc-1)] should be names of file to grep through. If
         * there is more than one file to grep, we will print the filenames. */
-       if (argc > 1)
+       if (argv[0] && argv[1])
                print_filename = 1;
        /* -H / -h of course override */
        if (option_mask32 & OPT_H)
@@ -499,7 +801,7 @@ int grep_main(int argc, char **argv)
         * stdin. Otherwise, we grep through all the files specified. */
        matched = 0;
        do {
-               cur_file = *argv++;
+               cur_file = *argv;
                file = stdin;
                if (!cur_file || LONE_DASH(cur_file)) {
                        cur_file = "(standard input)";
@@ -525,7 +827,7 @@ int grep_main(int argc, char **argv)
                matched += grep_file(file);
                fclose_if_not_stdin(file);
  grep_done: ;
-       } while (--argc > 0);
+       } while (*argv && *++argv);
 
        /* destroy all the elments in the pattern list */
        if (ENABLE_FEATURE_CLEAN_UP) {
@@ -537,7 +839,7 @@ int grep_main(int argc, char **argv)
                        if (gl->flg_mem_alocated_compiled & ALLOCATED)
                                free(gl->pattern);
                        if (gl->flg_mem_alocated_compiled & COMPILED)
-                               regfree(&(gl->preg));
+                               regfree(&gl->compiled_regex);
                        free(gl);
                        free(pattern_head_ptr);
                }