|
| 1 | +#!/usr/bin/perl |
| 2 | + |
| 3 | +# Detect bare 'grep' used as a test assertion where 'test_grep' |
| 4 | +# should be used, and '! test_grep' where 'test_grep !' should |
| 5 | +# be used. |
| 6 | +# |
| 7 | +# The shared shell parser tokenizes test bodies so that 'grep' |
| 8 | +# inside heredocs, command substitutions like $(grep ...), and |
| 9 | +# quoted strings is collapsed into a single token and never seen |
| 10 | +# by our check. Without the parser, a regex approach would |
| 11 | +# false-positive on patterns like: |
| 12 | +# |
| 13 | +# write_script foo.sh <<-\EOF |
| 14 | +# grep pattern file # data, not an assertion |
| 15 | +# EOF |
| 16 | +# |
| 17 | +# cmd 2>&1 | |
| 18 | +# grep pattern # continuation of pipeline |
| 19 | +# |
| 20 | +# The parser handles both by construction. |
| 21 | + |
| 22 | +use strict; |
| 23 | +use warnings; |
| 24 | +use File::Basename; |
| 25 | +do(dirname($0) . "/lib-shell-parser.pl") |
| 26 | + or die "$0: failed to load lib-shell-parser.pl: $@$!\n"; |
| 27 | + |
| 28 | +my $exit_code = 0; |
| 29 | + |
| 30 | +# GrepLintParser inherits ScriptParser's ability to find |
| 31 | +# test_expect_success/failure blocks and call check_test() |
| 32 | +# on each body. We override check_test() to walk the token |
| 33 | +# stream looking for bare grep assertions. |
| 34 | +package GrepLintParser; |
| 35 | + |
| 36 | +our @ISA = ('ScriptParser'); |
| 37 | + |
| 38 | +# After these tokens, the next token is a command word. |
| 39 | +# For example, in 'echo foo && grep bar file', the 'grep' |
| 40 | +# after '&&' is at command position and should be flagged. |
| 41 | +my %cmd_start; |
| 42 | +@cmd_start{qw(&& || ; ;; do then else elif), "\n", '{', '('} = (); |
| 43 | + |
| 44 | +# Subset of cmd_start that also ends a pipeline. |
| 45 | +# 'echo | grep foo && grep bar' -- the second grep is NOT |
| 46 | +# in a pipeline because '&&' ended it. But 'echo | \n grep' |
| 47 | +# keeps the pipeline across the newline. |
| 48 | +my %pipe_end; |
| 49 | +@pipe_end{qw(&& || ; ;;)} = (); |
| 50 | + |
| 51 | +# If any of these appear after grep (before the next command |
| 52 | +# separator), grep's output is being piped or redirected, |
| 53 | +# so it is a filter, not an assertion. |
| 54 | +# For example: 'grep pattern file | wc -l' or |
| 55 | +# 'grep pattern file >output' |
| 56 | +my %filter_op; |
| 57 | +@filter_op{'|', '>', '>>', '<'} = (); |
| 58 | + |
| 59 | +# Check if the raw source line has a '# lint-ok' annotation |
| 60 | +# that suppresses the warning. |
| 61 | +sub lint_ok { |
| 62 | + my ($raw_lines, $ln) = @_; |
| 63 | + return 0 if $ln < 1 || $ln > @$raw_lines; |
| 64 | + return $raw_lines->[$ln - 1] =~ /lint-ok/; |
| 65 | +} |
| 66 | + |
| 67 | +# Scan forward from 'grep' to see if its output is piped or |
| 68 | +# redirected (making it a filter, not an assertion). Stops |
| 69 | +# at the next command separator. |
| 70 | +# |
| 71 | +# Correctly handles compound-command redirects like |
| 72 | +# '{ grep pattern file; } >output' because the scan passes |
| 73 | +# through '}' (not in %cmd_start) and finds '>'. |
| 74 | +sub is_filter { |
| 75 | + my ($tokens, $pos) = @_; |
| 76 | + for (my $j = $pos + 1; $j < @$tokens; $j++) { |
| 77 | + my $t = $tokens->[$j]->[0]; |
| 78 | + return 0 if exists $cmd_start{$t}; |
| 79 | + return 1 if exists $filter_op{$t}; |
| 80 | + } |
| 81 | + return 0; |
| 82 | +} |
| 83 | + |
| 84 | +# Called by ScriptParser for each test_expect_success/failure body. |
| 85 | +sub check_test { |
| 86 | + my $self = shift @_; |
| 87 | + my $title = ScriptParser::unwrap(shift @_); |
| 88 | + my $body_token = shift @_; |
| 89 | + my $lineno = $body_token->[3]; |
| 90 | + my $body = ScriptParser::unwrap($body_token); |
| 91 | + if ($body eq '-') { |
| 92 | + my $herebody = shift @_; |
| 93 | + if ($herebody) { |
| 94 | + $body = $herebody->{content}; |
| 95 | + $lineno = $herebody->{start_line}; |
| 96 | + } |
| 97 | + } |
| 98 | + return unless $body; |
| 99 | + |
| 100 | + # Double-quoted bodies can have line splices (\<newline>) |
| 101 | + # that the Lexer consumes without emitting into the body |
| 102 | + # text. For example: |
| 103 | + # |
| 104 | + # test_expect_success "title" " |
| 105 | + # x=\$(echo \ |
| 106 | + # hello) && |
| 107 | + # grep pattern file |
| 108 | + # " |
| 109 | + # |
| 110 | + # The '\' on line 2 splices lines 2-3 into one body line, |
| 111 | + # so the inner parser's line count falls behind the source. |
| 112 | + # Build a table mapping each inner-parser line to the number |
| 113 | + # of splices before it, so we can adjust the reported line. |
| 114 | + my $raw_lines = $self->{raw_lines}; |
| 115 | + my @splice_adj; |
| 116 | + if ($body_token->[0] =~ /^"/) { |
| 117 | + my $splice_count = 0; |
| 118 | + my $body_line = 0; |
| 119 | + for my $src_ln ($lineno .. ($body_token->[4] || $lineno)) { |
| 120 | + my $line = $raw_lines->[$src_ln - 1]; |
| 121 | + # Odd trailing backslashes = line splice (\<nl>). |
| 122 | + # Even = escaped backslashes (\\), not a splice. |
| 123 | + if (defined($line) && $line =~ /(\\*)\s*$/ && |
| 124 | + length($1) % 2 == 1) { |
| 125 | + $splice_count++; |
| 126 | + } else { |
| 127 | + $splice_adj[++$body_line] = $splice_count; |
| 128 | + } |
| 129 | + } |
| 130 | + } |
| 131 | + |
| 132 | + # Re-parse the body text. The Lexer has already collapsed |
| 133 | + # heredocs, $(...), and quoted strings into single tokens, |
| 134 | + # so 'grep' inside those never appears in this stream. |
| 135 | + my $parser = ShellParser->new(\$body); |
| 136 | + my @tokens = $parser->parse(); |
| 137 | + |
| 138 | + my $file = $self->{file}; |
| 139 | + my $cmd_pos = 1; |
| 140 | + my $in_pipe = 0; |
| 141 | + |
| 142 | + for (my $i = 0; $i < @tokens; $i++) { |
| 143 | + my $text = $tokens[$i]->[0]; |
| 144 | + |
| 145 | + # Map inner-parser line to file line, adjusting for |
| 146 | + # splices in double-quoted bodies. |
| 147 | + my $rel = $tokens[$i]->[3] || 0; |
| 148 | + my $ln = $rel + ($lineno || 1) - 1; |
| 149 | + $ln += $splice_adj[$rel] if @splice_adj && $rel <= $#splice_adj; |
| 150 | + |
| 151 | + # Command separators: next token is a command word. |
| 152 | + # 'echo foo && grep bar' -- grep is at cmd position. |
| 153 | + if (exists $cmd_start{$text}) { |
| 154 | + $cmd_pos = 1; |
| 155 | + $in_pipe = 0 if exists $pipe_end{$text}; |
| 156 | + next; |
| 157 | + } |
| 158 | + # Pipe: next token is a command word receiving input. |
| 159 | + # 'echo foo | grep bar' -- grep is a filter here. |
| 160 | + if ($text eq '|') { |
| 161 | + $cmd_pos = 1; |
| 162 | + $in_pipe = 1; |
| 163 | + next; |
| 164 | + } |
| 165 | + # Close brace/paren: not a command position. |
| 166 | + # '{ echo; } && grep' -- grep comes after &&, not }. |
| 167 | + if ($text eq '}' || $text eq ')') { |
| 168 | + $cmd_pos = 0; |
| 169 | + next; |
| 170 | + } |
| 171 | + |
| 172 | + # Not at command position -- skip arguments, patterns, |
| 173 | + # filenames. 'test_must_fail grep' -- grep is an |
| 174 | + # argument to test_must_fail, not a command. |
| 175 | + next unless $cmd_pos; |
| 176 | + |
| 177 | + # '!' is a negation prefix; the next token is still |
| 178 | + # a command word. Check for the specific anti-pattern |
| 179 | + # '! test_grep' which suppresses test_grep's diagnostic |
| 180 | + # output on failure. |
| 181 | + if ($text eq '!') { |
| 182 | + if ($i + 1 < @tokens && |
| 183 | + $tokens[$i + 1]->[0] eq 'test_grep' && |
| 184 | + !lint_ok($raw_lines, $ln)) { |
| 185 | + print "$file:$ln: error: ", |
| 186 | + 'use "test_grep !" instead of ', |
| 187 | + '"! test_grep"', "\n"; |
| 188 | + $exit_code = 1; |
| 189 | + } |
| 190 | + next; |
| 191 | + } |
| 192 | + |
| 193 | + # Flag bare grep at command position that is not |
| 194 | + # receiving piped input and not followed by a |
| 195 | + # redirect or pipe (i.e., it is an assertion). |
| 196 | + if ($text eq 'grep' && !$in_pipe && |
| 197 | + !is_filter(\@tokens, $i) && |
| 198 | + !lint_ok($raw_lines, $ln)) { |
| 199 | + print "$file:$ln: error: ", |
| 200 | + "bare grep outside pipeline ", |
| 201 | + "(use test_grep)\n"; |
| 202 | + $exit_code = 1; |
| 203 | + } |
| 204 | + |
| 205 | + $cmd_pos = 0; |
| 206 | + } |
| 207 | +} |
| 208 | + |
| 209 | +package main; |
| 210 | + |
| 211 | +for my $file (@ARGV) { |
| 212 | + open(my $fh, '<:unix:crlf', $file) or die "$0: $file: $!\n"; |
| 213 | + my @raw_lines = <$fh>; |
| 214 | + close $fh; |
| 215 | + my $s = join('', @raw_lines); |
| 216 | + my $parser = GrepLintParser->new(\$s); |
| 217 | + $parser->{file} = $file; |
| 218 | + $parser->{raw_lines} = \@raw_lines; |
| 219 | + $parser->parse(); |
| 220 | +} |
| 221 | +exit $exit_code; |
0 commit comments