1#! /usr/bin/env perl 2# 3# Copyright 2020-2021 The OpenSSL Project Authors. All Rights Reserved. 4# Copyright Siemens AG 2019-2020 5# 6# Licensed under the Apache License 2.0 (the "License"). 7# You may not use this file except in compliance with the License. 8# You can obtain a copy in the file LICENSE in the source distribution 9# or at https://www.openssl.org/source/license.html 10# 11# check-format.pl 12# - check formatting of C source according to OpenSSL coding style 13# 14# usage: 15# check-format.pl [-l|--sloppy-len] [-l|--sloppy-bodylen] 16# [-s|--sloppy-space] [-c|--sloppy-comment] 17# [-m|--sloppy-macro] [-h|--sloppy-hang] 18# [-e|--eol-comment] [-1|--1-stmt] 19# <files> 20# 21# run self-tests: 22# util/check-format.pl util/check-format-test-positives.c 23# util/check-format.pl util/check-format-test-negatives.c 24# 25# checks adherence to the formatting rules of the OpenSSL coding guidelines 26# assuming that the input files contain syntactically correct C code. 27# This pragmatic tool is incomplete and yields some false positives. 28# Still it should be useful for detecting most typical glitches. 29# 30# options: 31# -l | --sloppy-len increase accepted max line length from 80 to 84 32# -l | --sloppy-bodylen do not report function body length > 200 33# -s | --sloppy-space do not report whitespace nits 34# -c | --sloppy-comment do not report indentation of comments 35# Otherwise for each multi-line comment the indentation of 36# its lines is checked for consistency. For each comment 37# that does not begin to the right of normal code its 38# indentation must be as for normal code, while in case it 39# also has no normal code to its right it is considered to 40# refer to the following line and may be indented equally. 41# -m | --sloppy-macro allow missing extra indentation of macro bodies 42# -h | --sloppy-hang when checking hanging indentation, do not report 43# * same indentation as on line before 44# * same indentation as non-hanging indent level 45# * indentation moved left (not beyond non-hanging indent) 46# just to fit contents within the line length limit 47# -e | --eol-comment report needless intermediate multiple consecutive spaces also before end-of-line comments 48# -1 | --1-stmt do more aggressive checks for { 1 stmt } - see below 49# 50# There are non-trivial false positives and negatives such as the following. 51# 52# * When a line contains several issues of the same kind only one is reported. 53# 54# * When a line contains more than one statement this is (correctly) reported 55# but in some situations the indentation checks for subsequent lines go wrong. 56# 57# * There is the special OpenSSL rule not to unnecessarily use braces around 58# single statements: 59# { 60# stmt; 61# } 62# except within if ... else constructs where some branch contains more than one 63# statement. Since the exception is hard to recognize when such branches occur 64# after the current position (such that false positives would be reported) 65# the tool by checks for this rule by defaul only for do/while/for bodies. 66# Yet with the --1-stmt option false positives are preferred over negatives. 67# False negatives occur if the braces are more than two non-empty lines apart. 68# 69# * The presence of multiple consecutive spaces is regarded a coding style nit 70# except when this is before end-of-line comments (unless the --eol-comment is given) and 71# except when done in order to align certain columns over multiple lines, e.g.: 72# # define AB 1 73# # define CDE 22 74# # define F 3333 75# This pattern is recognized - and consequently extra space not reported - 76# for a given line if in the nonempty line before or after (if existing) 77# for each occurrence of " \S" (where \S means non-space) in the given line 78# there is " \S" in the other line in the respective column position. 79# This may lead to both false negatives (in case of coincidental " \S") 80# and false positives (in case of more complex multi-column alignment). 81# 82# * When just part of control structures depend on #if(n)(def), which can be 83# considered bad programming style, indentation false positives occur, e.g.: 84# #if X 85# if (1) /* bad style */ 86# #else 87# if (2) /* bad style resulting in false positive */ 88# #endif 89# c; /* resulting further false positive */ 90 91use strict; 92# use List::Util qw[min max]; 93use POSIX; 94 95use constant INDENT_LEVEL => 4; 96use constant MAX_LINE_LENGTH => 80; 97use constant MAX_BODY_LENGTH => 200; 98 99# global variables @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 100 101# command-line options 102my $max_length = MAX_LINE_LENGTH; 103my $sloppy_bodylen = 0; 104my $sloppy_SPC = 0; 105my $sloppy_hang = 0; 106my $sloppy_cmt = 0; 107my $sloppy_macro = 0; 108my $eol_cmt = 0; 109my $extended_1_stmt = 0; 110 111while ($ARGV[0] =~ m/^-(\w|-[\w\-]+)$/) { 112 my $arg = $1; shift; 113 if ($arg =~ m/^(l|-sloppy-len)$/) { 114 $max_length += INDENT_LEVEL; 115 } elsif ($arg =~ m/^(b|-sloppy-bodylen)$/) { 116 $sloppy_bodylen = 1; 117 } elsif ($arg =~ m/^(s|-sloppy-space)$/) { 118 $sloppy_SPC= 1; 119 } elsif ($arg =~ m/^(c|-sloppy-comment)$/) { 120 $sloppy_cmt = 1; 121 } elsif ($arg =~ m/^(m|-sloppy-macro)$/) { 122 $sloppy_macro = 1; 123 } elsif ($arg =~ m/^(h|-sloppy-hang)$/) { 124 $sloppy_hang = 1; 125 } elsif ($arg =~ m/^(e|-eol-comment)$/) { 126 $eol_cmt = 1; 127 } elsif ($arg =~ m/^(1|-1-stmt)$/) { 128 $extended_1_stmt = 1; 129 } else { 130 die("unknown option: -$arg"); 131 } 132} 133 134# status variables 135my $self_test; # whether the current input file is regarded to contain (positive/negative) self-tests 136my $line; # current line number 137my $line_before; # number of previous not essentially empty line (containing at most whitespace and '\') 138my $line_before2; # number of not essentially empty line before previous not essentially empty line 139my $contents; # contents of current line (without blinding) 140# $_ # current line, where comments etc. get blinded 141my $contents_before; # contents of $line_before (without blinding), if $line_before > 0 142my $contents_before_; # contents of $line_before after blinding comments etc., if $line_before > 0 143my $contents_before2; # contents of $line_before2 (without blinding), if $line_before2 > 0 144my $contents_before_2; # contents of $line_before2 after blinding comments etc., if $line_before2 > 0 145my $in_multiline_string; # line starts within multi-line string literal 146my $count; # -1 or number of leading whitespace characters (except newline) in current line, 147 # which should be $block_indent + $hanging_offset + $local_offset or $expr_indent 148my $count_before; # number of leading whitespace characters (except line ending chars) in $contents_before 149my $has_label; # current line contains label 150my $local_offset; # current extra indent due to label, switch case/default, or leading closing brace(s) 151my $line_body_start; # number of line where last function body started, or 0 152my $line_function_start; # number of line where last function definition started, used if $line_body_start != 0 153my $last_function_header; # header containing name of last function defined, used if $line_function_start != 0 154my $line_opening_brace; # number of previous line with opening brace after do/while/for, optionally for if/else 155 156my $keyword_opening_brace; # name of previous keyword, used if $line_opening_brace != 0 157my $ifdef__cplusplus; # line before contained '#ifdef __cplusplus' (used in header files) 158my $block_indent; # currently required normal indentation at block/statement level 159my $hanging_offset; # extra indent, which may be nested, for just one hanging statement or expr or typedef 160my @in_do_hanging_offsets; # stack of hanging offsets for nested 'do' ... 'while' 161my @in_if_hanging_offsets; # stack of hanging offsets for nested 'if' (but not its potential 'else' branch) 162my $if_maybe_terminated; # 'if' ends and $hanging_offset should be reset unless the next line starts with 'else' 163my @nested_block_indents; # stack of indentations at block/statement level, needed due to hanging statements 164my @nested_hanging_offsets;# stack of nested $hanging_offset values, in parallel to @nested_block_indents 165my @nested_in_typedecl; # stack of nested $in_typedecl values, partly in parallel to @nested_block_indents 166my @nested_indents; # stack of hanging indents due to parentheses, braces, brackets, or conditionals 167my @nested_symbols; # stack of hanging symbols '(', '{', '[', or '?', in parallel to @nested_indents 168my @nested_conds_indents; # stack of hanging indents due to conditionals ('?' ... ':') 169my $expr_indent; # resulting hanging indent within (multi-line) expressions including type exprs, else 0 170my $hanging_symbol; # character ('(', '{', '[', not: '?') responsible for $expr_indent, if $expr_indent != 0 171my $in_expr; # in expression after if/while/for/switch/return/enum/LHS of assignment 172my $in_paren_expr; # in parenthesized if/while/for condition and switch expression, if $expr_indent != 0 173my $in_typedecl; # nesting level of typedef/struct/union/enum 174my $in_directive; # number of lines so far within preprocessor directive, e.g., macro definition 175my $directive_nesting; # currently required indentation of preprocessor directive according to #if(n)(def) 176my $directive_offset; # indent offset within multi-line preprocessor directive, if $in_directive > 0 177my $in_macro_header; # number of open parentheses + 1 in (multi-line) header of #define, if $in_directive > 0 178my $in_comment; # number of lines so far within multi-line comment, or < 0 when end is on current line 179my $leading_comment; # multi-line comment has no code before its beginning delimiter 180my $formatted_comment; # multi-line comment beginning with "/*-", which indicates/allows special formatting 181my $comment_indent; # comment indent, if $in_comment != 0 182my $num_reports_line = 0; # number of issues found on current line 183my $num_reports = 0; # total number of issues found 184my $num_indent_reports = 0;# total number of indentation issues found 185my $num_nesting_issues = 0;# total number of directive nesting issues found 186my $num_syntax_issues = 0; # total number of syntax issues found during sanity checks 187my $num_SPC_reports = 0; # total number of whitespace issues found 188my $num_length_reports = 0;# total number of line length issues found 189 190sub reset_file_state { 191 $line = 0; 192 $line_before = 0; 193 $line_before2 = 0; 194 @nested_block_indents = (); 195 @nested_hanging_offsets = (); 196 @nested_in_typedecl = (); 197 @nested_symbols = (); 198 @nested_indents = (); 199 @nested_conds_indents = (); 200 $expr_indent = 0; 201 $in_paren_expr = 0; 202 $in_expr = 0; 203 $hanging_offset = 0; 204 @in_do_hanging_offsets = (); 205 @in_if_hanging_offsets = (); 206 $if_maybe_terminated = 0; 207 $block_indent = 0; 208 $ifdef__cplusplus = 0; 209 $in_multiline_string = 0; 210 $line_body_start = 0; 211 $line_opening_brace = 0; 212 $in_typedecl = 0; 213 $in_directive = 0; 214 $directive_nesting = 0; 215 $in_comment = 0; 216} 217 218# auxiliary submodules @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 219 220sub report_flexibly { 221 my $line = shift; 222 my $msg = shift; 223 my $contents = shift; 224 my $report_SPC = $msg =~ /space/; 225 return if $report_SPC && $sloppy_SPC; 226 227 print "$ARGV:$line:$msg:$contents" unless $self_test; 228 $num_reports_line++; 229 $num_reports++; 230 $num_indent_reports++ if $msg =~ m/indent/; 231 $num_nesting_issues++ if $msg =~ m/directive nesting/; 232 $num_syntax_issues++ if $msg =~ m/unclosed|unexpected/; 233 $num_SPC_reports++ if $report_SPC; 234 $num_length_reports++ if $msg =~ m/length/; 235} 236 237sub report { 238 my $msg = shift; 239 report_flexibly($line, $msg, $contents); 240} 241 242sub parens_balance { # count balance of opening parentheses - closing parentheses 243 my $str = shift; 244 return $str =~ tr/\(// - $str =~ tr/\)//; 245} 246 247sub blind_nonspace { # blind non-space text of comment as @, preserving length and spaces 248 # the @ character is used because it cannot occur in normal program code so there is no confusion 249 # comment text is not blinded to whitespace in order to be able to check extra SPC also in comments 250 my $comment_text = shift; 251 $comment_text =~ s/([\.\?\!])\s\s/$1. /g; # in extra SPC checks allow one extra SPC after period '.', '?', or '!' in comments 252 return $comment_text =~ tr/ /@/cr; 253} 254 255# submodule for indentation checking/reporting @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 256 257sub check_indent { # used for lines outside multi-line string literals 258 my $stmt_indent = $block_indent + $hanging_offset + $local_offset; 259 $stmt_indent = 0 if $stmt_indent < 0; # TODO maybe give warning/error 260 my $stmt_desc = $contents =~ 261 m/^\s*\/\*/ ? "intra-line comment" : 262 $has_label ? "label" : 263 ($hanging_offset != 0 ? "hanging " : ""). 264 ($hanging_offset != 0 ? "stmt/expr" : "stmt/decl"); # $in_typedecl is not fully to the point here 265 my ($ref_desc, $ref_indent) = $expr_indent == 0 ? ($stmt_desc, $stmt_indent) 266 : ("hanging '$hanging_symbol'", $expr_indent); 267 my ($alt_desc, $alt_indent) = ("", $ref_indent); 268 269 # allow indent 1 for labels - this cannot happen for leading ':' 270 ($alt_desc, $alt_indent) = ("outermost position", 1) if $expr_indent == 0 && $has_label; 271 272 if (@nested_conds_indents != 0 && substr($_, $count, 1) eq ":") { 273 # leading ':' within stmt/expr/decl - this cannot happen for labels, leading '&&', or leading '||' 274 # allow special indent at level of corresponding "?" 275 ($alt_desc, $alt_indent) = ("leading ':'", @nested_conds_indents[-1]); 276 } 277 # allow extra indent offset leading '&&' or '||' - this cannot happen for leading ":" 278 ($alt_desc, $alt_indent) = ("leading '$1'", $ref_indent + INDENT_LEVEL) if $contents =~ m/^[\s@]*(\&\&|\|\|)/; 279 280 if ($expr_indent < 0) { # implies @nested_symbols != 0 && @nested_symbols[0] eq "{" && @nested_indents[-1] < 0 281 # allow normal stmt indentation level for hanging initializer/enum expressions after trailing '{' 282 # this cannot happen for labels and overrides special treatment of ':', '&&' and '||' for this line 283 ($alt_desc, $alt_indent) = ("lines after '{'", $stmt_indent); 284 # decide depending on current actual indentation, preventing forth and back 285 @nested_indents[-1] = $count == $stmt_indent ? $stmt_indent : -@nested_indents[-1]; # allow $stmt_indent 286 $ref_indent = $expr_indent = @nested_indents[-1]; 287 } 288 289 # check consistency of indentation within multi-line comment (i.e., between its first, inner, and last lines) 290 if ($in_comment != 0 && $in_comment != 1) { # in multi-line comment but not on its first line 291 if (!$sloppy_cmt) { 292 if ($in_comment > 0) { # not at its end 293 report("indent = $count != $comment_indent within multi-line comment") 294 if $count != $comment_indent; 295 } else { 296 my $tweak = $in_comment == -2 ? 1 : 0; 297 report("indent = ".($count + $tweak)." != $comment_indent at end of multi-line comment") 298 if $count + $tweak != $comment_indent; 299 } 300 } 301 # do not check indentation of last line of non-leading multi-line comment 302 if ($in_comment < 0 && !$leading_comment) { 303 s/^(\s*)@/$1*/; # blind first '@' as '*' to prevent below delayed check for the line before 304 return; 305 } 306 return if $in_comment > 0; # not on its last line 307 # $comment_indent will be checked by the below checks for end of multi-line comment 308 } 309 310 # else check indentation of entire-line comment or entire-line end of multi-line comment 311 # ... w.r.t. indent of the following line by delayed check for the line before 312 if (($in_comment == 0 || $in_comment == 1) # no comment, intra-line comment, or begin of multi-line comment 313 && $line_before > 0 # there is a line before 314 && $contents_before_ =~ m/^(\s*)@[\s@]*$/) { # line before begins with '@', no code follows (except '\') 315 report_flexibly($line_before, "entire-line comment indent = $count_before != $count (of following line)", 316 $contents_before) if !$sloppy_cmt && $count_before != $count; 317 } 318 # ... but allow normal indentation for the current line, else above check will be done for the line before 319 if (($in_comment == 0 || $in_comment < 0) # (no commment,) intra-line comment or end of multi-line comment 320 && m/^(\s*)@[\s@]*$/) { # line begins with '@', no code follows (except '\') 321 if ($count == $ref_indent) { # indentation is like for (normal) code in this line 322 s/^(\s*)@/$1*/; # blind first '@' as '*' to prevent above delayed check for the line before 323 return; 324 } 325 return if !eof; # defer check of entire-line comment to next line 326 } 327 328 # else check indentation of leading intra-line comment or end of multi-line comment 329 if (m/^(\s*)@/) { # line begins with '@', i.e., any (remaining type of) comment 330 if (!$sloppy_cmt && $count != $ref_indent) { 331 report("intra-line comment indent = $count != $ref_indent") if $in_comment == 0; 332 report("multi-line comment indent = $count != $ref_indent") if $in_comment < 0; 333 } 334 return; 335 } 336 337 if ($sloppy_hang && ($hanging_offset != 0 || $expr_indent != 0)) { 338 # do not report same indentation as on the line before (potentially due to same violations) 339 return if $line_before > 0 && $count == $count_before; 340 341 # do not report indentation at normal indentation level while hanging expression indent would be required 342 return if $expr_indent != 0 && $count == $stmt_indent; 343 344 # do not report if contents have been shifted left of nested expr indent (but not as far as stmt indent) 345 # apparently aligned to the right in order to fit within line length limit 346 return if $stmt_indent < $count && $count < $expr_indent && 347 length($contents) == MAX_LINE_LENGTH + length("\n"); 348 } 349 350 report("indent = $count != $ref_indent for $ref_desc". 351 ($alt_desc eq "" 352 || $alt_indent == $ref_indent # prevent showing alternative that happens to have equal value 353 ? "" : " or $alt_indent for $alt_desc")) 354 if $count != $ref_indent && $count != $alt_indent; 355} 356 357# submodules handling indentation within expressions @@@@@@@@@@@@@@@@@@@@@@@@@@@ 358 359sub update_nested_indents { # may reset $in_paren_expr and in this case also resets $in_expr 360 my $str = shift; 361 my $start = shift; # defaults to 0 362 my $terminator_position = -1; 363 for (my $i = $start; $i < length($str); $i++) { 364 my $c; 365 my $curr = substr($str, $i); 366 if ($curr =~ m/^(.*?)([{}()?:;\[\]])(.*)$/) { # match from position $i the first {}()?:;[] 367 $c = $2; 368 } else { 369 last; 370 } 371 my ($head, $tail) = (substr($str, 0, $i).$1, $3); 372 $i += length($1) + length($2) - 1; 373 374 # stop at terminator outside 'for(..;..;..)', assuming that 'for' is followed by '(' 375 return $i if $c eq ";" && (!$in_paren_expr || @nested_indents == 0); 376 377 my $in_stmt = $in_expr || @nested_symbols != 0; # not: || $in_typedecl != 0 378 if ($c =~ m/[{([?]/) { # $c is '{', '(', '[', or '?' 379 if ($c eq "{") { # '{' in any context 380 # cancel newly hanging_offset if opening brace '{' is after non-whitespace non-comment: 381 $hanging_offset -= INDENT_LEVEL if $hanging_offset > 0 && $head =~ m/[^\s\@]/; 382 push @nested_block_indents, $block_indent; 383 push @nested_hanging_offsets, $in_expr ? $hanging_offset : 0; 384 push @nested_in_typedecl, $in_typedecl if $in_typedecl != 0; 385 $block_indent += INDENT_LEVEL + $hanging_offset; 386 $hanging_offset = 0; 387 } 388 if ($c ne "{" || $in_stmt) { # for '{' inside stmt/expr (not: decl), for '(', '[', or '?' anywhere 389 $tail =~ m/^([\s@]*)([^\s\@])/; 390 push @nested_indents, defined $2 391 ? $i + 1 + length($1) # actual indentation of following non-space non-comment 392 : $c ne "{" ? +($i + 1) # just after '(' or '[' if only whitespace thereafter 393 : -($i + 1); # allow also $stmt_indent if '{' with only whitespace thereafter 394 push @nested_symbols, $c; # done also for '?' to be able to check correct nesting 395 push @nested_conds_indents, $i if $c eq "?"; # remember special alternative indent for ':' 396 } 397 } elsif ($c =~ m/[})\]:]/) { # $c is '}', ')', ']', or ':' 398 my $opening_c = ($c =~ tr/})]:/{([/r); 399 if (($c ne ":" || $in_stmt # ignore ':' outside stmt/expr/decl 400 # in the presence of ':', one could add this sanity check: 401 # && !(# ':' after initial label/case/default 402 # $head =~ m/^([\s@]*)(case\W.*$|\w+$)/ || # this matching would not work for 403 # # multi-line expr after 'case' 404 # # bitfield length within unsigned type decl 405 # $tail =~ m/^[\s@]*\d+/ # this matching would need improvement 406 # ) 407 )) { 408 if ($c ne "}" || $in_stmt) { # for '}' inside stmt/expr/decl, ')', ']', or ':' 409 if (@nested_symbols != 0 && 410 @nested_symbols[-1] == $opening_c) { # for $c there was a corresponding $opening_c 411 pop @nested_indents; 412 pop @nested_symbols; 413 pop @nested_conds_indents if $opening_c eq "?"; 414 } else { 415 report("unexpected '$c' @ ".($in_paren_expr ? "(expr)" : "expr")); 416 next; 417 } 418 } 419 if ($c eq "}") { # '}' at block level but also inside stmt/expr/decl 420 if (@nested_block_indents == 0) { 421 report("unexpected '}'"); 422 } else { 423 $block_indent = pop @nested_block_indents; 424 $hanging_offset = pop @nested_hanging_offsets; 425 $in_typedecl = pop @nested_in_typedecl if @nested_in_typedecl != 0; 426 } 427 } 428 if ($in_paren_expr && !grep(/\(/, @nested_symbols)) { # end of (expr) 429 check_nested_nonblock_indents("(expr)"); 430 $in_paren_expr = $in_expr = 0; 431 report("code after (expr)") 432 if $tail =~ m/^([^{]*)/ && $1 =~ m/[^\s\@;]/; # non-space non-';' before any '{' 433 } 434 } 435 } 436 } 437 return -1; 438} 439 440sub check_nested_nonblock_indents { 441 my $position = shift; 442 while (@nested_symbols != 0) { 443 my $symbol = pop @nested_symbols; 444 report("unclosed '$symbol' in $position"); 445 if ($symbol eq "{") { # repair stack of blocks 446 $block_indent = pop @nested_block_indents; 447 $hanging_offset = pop @nested_hanging_offsets; 448 $in_typedecl = pop @nested_in_typedecl if @nested_in_typedecl != 0; 449 } 450 } 451 @nested_indents = (); 452 @nested_conds_indents = (); 453} 454 455# start of main program @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 456 457reset_file_state(); 458 459while (<>) { # loop over all lines of all input files 460 $self_test = $ARGV =~ m/check-format-test/; 461 $line++; 462 s/\r$//; # strip any trailing CR '\r' (which are typical on Windows systems) 463 $contents = $_; 464 465 # check for illegal characters 466 if (m/(.*?)([\x00-\x09\x0B-\x1F\x7F-\xFF])/) { 467 my $col = length($1); 468 report(($2 eq "\x09" ? "TAB" : $2 eq "\x0D" ? "CR " : $2 =~ m/[\x00-\x1F]/ ? "non-printable" 469 : "non-7bit char") . " at column $col") ; 470 } 471 472 # check for whitespace at EOL 473 report("trailing whitespace at EOL") if m/\s\n$/; 474 475 # assign to $count the actual indentation level of the current line 476 chomp; # remove trailing NL '\n' 477 m/^(\s*)/; 478 $count = length($1); # actual indentation 479 $has_label = 0; 480 $local_offset = 0; 481 482 # character/string literals @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 483 484 s/\\["']/@@/g; # blind all '"' and "'" escaped by '\' (typically within character literals or string literals) 485 486 # handle multi-line string literals to avoid confusion on starting/ending '"' and trailing '\' 487 if ($in_multiline_string) { 488 if (s#^([^"]*)"#($1 =~ tr/"/@/cr).'@'#e) { # string literal terminated by '"' 489 # string contents and its terminating '"' have been blinded as '@' 490 $count = -1; # do not check indentation 491 } else { 492 report("multi-line string literal not terminated by '\"' and trailing '\' is missing") 493 unless s#^([^\\]*)\s*\\\s*$#$1#; # strip trailing '\' plus any whitespace around 494 goto LINE_FINISHED; 495 } 496 } 497 498 # blind contents of character and string literals as @, preserving length (but not spaces) 499 # this prevents confusing any of the matching below, e.g., of whitespace and comment delimiters 500 s#('[^']*')#$1 =~ tr/'/@/cr#eg; # handle all intra-line character literals 501 s#("[^"]*")#$1 =~ tr/"/@/cr#eg; # handle all intra-line string literals 502 $in_multiline_string = # handle trailing string literal terminated by '\' 503 s#^(([^"]*"[^"]*")*[^"]*)("[^"]*)\\(\s*)$#$1.($3 =~ tr/"/@/cr).'"'.$4#e; 504 # its contents have been blinded and the trailing '\' replaced by '"' 505 506 # strip any other trailing '\' along with any whitespace around it such that it does not interfere with various 507 # matching below; the later handling of multi-line macro definitions uses $contents where it is not stripped 508 s#^(.*?)\s*\\\s*$#$1#; # trailing '\' possibly preceded and/or followed by whitespace 509 510 # comments @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 511 512 # do/prepare checks within multi-line comments 513 my $self_test_exception = $self_test ? "@" : ""; 514 if ($in_comment > 0) { # this still includes the last line of multi-line commment 515 my ($head, $any_symbol, $cmt_text) = m/^(\s*)(.?)(.*)$/; 516 if ($any_symbol eq "*") { 517 report("no space after leading '*' in multi-line comment") if $cmt_text =~ m|^[^/\s$self_test_exception]|; 518 } else { 519 report("no leading '*' in multi-line comment"); 520 } 521 $in_comment++; 522 } 523 524 # detect end of comment, must be within multi-line comment, check if it is preceded by non-whitespace text 525 if ((my ($head, $tail) = m|^(.*?)\*/(.*)$|) && $1 ne '/') { # ending comment: '*/' 526 report("neither space nor '*' before '*/'") if $head =~ m/[^*\s]$/; 527 report("no space after '*/'") if $tail =~ m/^[^\s,;)}\]]/; # no space or ,;)}] after '*/' 528 if (!($head =~ m|/\*|)) { # not begin of comment '/*', which is is handled below 529 if ($in_comment == 0) { 530 report("unexpected '*/' outside comment"); 531 $_ = "$head@@".$tail; # blind the "*/" 532 } else { 533 report("text before '*/' in multi-line comment") if ($head =~ m/\S/); # non-SPC before '*/' 534 $in_comment = -1; # indicate that multi-line comment ends on current line 535 if ($count > 0) { 536 # make indentation of end of multi-line comment appear like of leading intra-line comment 537 $head =~ s/^(\s*)\s/$1@/; # replace the last leading space by '@' 538 $count--; 539 $in_comment = -2; # indicate that multi-line comment ends on current line, with tweak 540 } 541 my $cmt_text = $head; 542 $_ = blind_nonspace($cmt_text)."@@".$tail; 543 } 544 } 545 } 546 547 # detect begin of comment, check if it is followed by non-space text 548 MATCH_COMMENT: 549 if (my ($head, $opt_minus, $tail) = m|^(.*?)/\*(-?)(.*)$|) { # begin of comment: '/*' 550 report("no space before '/*'") 551 if $head =~ m/[^\s(\*]$/; # not space, '(', or or '*' (needed to allow '*/') before comment delimiter 552 report("neither space nor '*' after '/*' or '/*-'") if $tail =~ m/^[^\s*$self_test_exception]/; 553 my $cmt_text = $opt_minus.$tail; # preliminary 554 if ($in_comment > 0) { 555 report("unexpected '/*' inside multi-line comment"); 556 } elsif ($tail =~ m|^(.*?)\*/(.*)$|) { # comment end: */ on same line 557 report("unexpected '/*' inside intra-line comment") if $1 =~ /\/\*/; 558 # blind comment text, preserving length and spaces 559 ($cmt_text, my $rest) = ($opt_minus.$1, $2); 560 $_ = "$head@@".blind_nonspace($cmt_text)."@@".$rest; 561 goto MATCH_COMMENT; 562 } else { # begin of multi-line comment 563 my $self_test_exception = $self_test ? "(@\d?)?" : ""; 564 report("text after '/*' in multi-line comment") 565 unless $tail =~ m/^$self_test_exception.?\s*$/; 566 # tail not essentially empty, first char already checked 567 # adapt to actual indentation of first line 568 $comment_indent = length($head) + 1; 569 $_ = "$head@@".blind_nonspace($cmt_text); 570 $in_comment = 1; 571 $leading_comment = $head =~ m/^\s*$/; # there is code before beginning delimiter 572 $formatted_comment = $opt_minus eq "-"; 573 } 574 } 575 576 if ($in_comment > 1) { # still inside multi-line comment (not at its begin or end) 577 m/^(\s*)\*?(\s*)(.*)$/; 578 $_ = $1."@".$2.blind_nonspace($3); 579 } 580 581 # handle special case of line after '#ifdef __cplusplus' (which typically appears in header files) 582 if ($ifdef__cplusplus) { 583 $ifdef__cplusplus = 0; 584 $_ = "$1 $2" if $contents =~ m/^(\s*extern\s*"C"\s*)\{(\s*)$/; # ignore opening brace in 'extern "C" {' 585 goto LINE_FINISHED if m/^\s*\}\s*$/; # ignore closing brace '}' 586 } 587 588 # check for over-long lines, 589 # while allowing trailing (also multi-line) string literals to go past $max_length 590 my $len = length; # total line length (without trailing '\n') 591 if ($len > $max_length && 592 !(m/^(.*)"[^"]*"\s*[\)\}\]]*[,;]?\s*$/ # string literal terminated by '"' (or '\'), then maybe )}],; 593 && length($1) < $max_length) 594 # this allows over-long trailing string literals with beginning col before $max_length 595 ) { 596 report("line length = $len > ".MAX_LINE_LENGTH); 597 } 598 599 # handle C++ / C99 - style end-of-line comments 600 if (my ($head, $cmt_text) = m|^(.*?)//(.*$)|) { 601 report("'//' end-of-line comment"); # the '//' comment style is not allowed for C90 602 # blind comment text, preserving length and spaces 603 $_ = "$head@@".blind_nonspace($cmt_text); 604 } 605 606 # at this point all non-space portions of any types of comments have been blinded as @ 607 608 goto LINE_FINISHED if m/^\s*$/; # essentially empty line: just whitespace (and maybe a trailing '\') 609 610 # intra-line whitespace nits @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 611 612 my $in_multiline_comment = ($in_comment > 1 || $in_comment < 0); # $in_multiline_comment refers to line before 613 if (!$sloppy_SPC && !($in_multiline_comment && $formatted_comment)) { 614 sub extra_SPC { 615 my $intra_line = shift; 616 return "extra space".($intra_line =~ m/@\s\s/ ? 617 $in_comment != 0 ? " in multi-line comment" 618 : " in intra-line comment" : ""); 619 } 620 sub split_line_head { # split line contents into header containing leading spaces and the first non-space char, and the rest of the line 621 my $comment_symbol = 622 $in_comment != 0 ? "@" : ""; # '@' will match the blinded leading '*' in multi-line comment 623 # $in_comment may pertain to the following line due to delayed check 624 # do not check for extra SPC in leading spaces including any '#' (or '*' within multi-line comment) 625 shift =~ m/^(\s*([#$comment_symbol]\s*)?)(.*?)\s*$/; 626 return ($1, $3); 627 } 628 my ($head , $intra_line ) = split_line_head($_); 629 my ($head1, $intra_line1) = split_line_head($contents_before_ ) if $line_before > 0; 630 my ($head2, $intra_line2) = split_line_head($contents_before_2) if $line_before2 > 0; 631 if ($line_before > 0) { # check with one line delay, such that at least $contents_before is available 632 sub column_alignments_only { # return 1 if the given line has multiple consecutive spaces only at columns that match the reference line 633 # all parameter strings are assumed to contain contents after blinding comments etc. 634 my $head = shift; # leading spaces and the first non-space char 635 my $intra = shift; # the rest of the line contents 636 my $contents = shift; # reference line 637 # check if all extra SPC in $intra is used only for multi-line column alignment with $contents 638 my $offset = length($head); 639 for (my $col = 0; $col < length($intra) - 2; $col++) { 640 my $substr = substr($intra, $col); 641 next unless $substr =~ m/^\s\s\S/; # extra SPC (but not in leading spaces of the line) 642 next if !$eol_cmt && $substr =~ m/^[@\s]+$/; # end-of-line comment 643 return 0 unless substr($contents, $col + $offset + 1, 2) =~ m/\s\S/; # reference line contents do not match 644 } 645 return 1; 646 } 647 report_flexibly($line_before, extra_SPC($intra_line1), $contents_before) if $intra_line1 =~ m/\s\s\S/ && 648 !( column_alignments_only($head1, $intra_line1, $_ ) # compare with $line 649 || ($line_before2 > 0 && 650 column_alignments_only($head1, $intra_line1, $contents_before_2))); # compare w/ $line_before2 651 report(extra_SPC($intra_line)) if $intra_line =~ m/\s\s\S/ && eof 652 && ! column_alignments_only($head , $intra_line , $contents_before_ ) ; # compare w/ $line_before 653 } elsif (eof) { # special case: just one line exists 654 report(extra_SPC($intra_line)) if $intra_line =~ m/\s\s\S/; 655 } 656 # ignore paths in #include 657 $intra_line =~ s/^(include\s*)(".*?"|<.*?>)/$1/e if $head =~ m/#/; 658 # treat op= and comparison operators as simple '=', simplifying matching below 659 $intra_line =~ s/([\+\-\*\/\/%\&\|\^\!<>=]|<<|>>)=/=/g; 660 # treat (type) variables within macro, indicated by trailing '\', as 'int' simplifying matching below 661 $intra_line =~ s/[A-Z_]+/int/g if $contents =~ m/^(.*?)\s*\\\s*$/; 662 # treat double &&, ||, <<, and >> as single ones, simplifying matching below 663 $intra_line =~ s/(&&|\|\||<<|>>)/substr($1, 0, 1)/eg; 664 # remove blinded comments etc. directly after [{( 665 while ($intra_line =~ s/([\[\{\(])@+\s?/$1/e) {} # /g does not work here 666 # remove blinded comments etc. directly before ,;)}] 667 while ($intra_line =~ s/\s?@+([,;\)\}\]])/$1/e) {} # /g does not work here 668 # treat remaining blinded comments and string literal contents as (single) space during matching below 669 $intra_line =~ s/@+/ /g; # note that extra SPC has already been handled above 670 $intra_line =~ s/\s+$//; # strip any (resulting) space at EOL 671 $intra_line =~ s/(for\s*\([^;]*);;(\))/"$1$2"/eg; # strip trailing ';;' in for (;;) 672 $intra_line =~ s/(for\s*\([^;]+;[^;]+);(\))/"$1$2"/eg; # strip trailing ';' in for (;;) 673 $intra_line =~ s/(=\s*)\{ /"$1@ "/eg; # do not report {SPC in initializers such as ' = { 0, };' 674 $intra_line =~ s/, \};/, @;/g; # do not report SPC} in initializers such as ' = { 0, };' 675 report("space before '$1'") if $intra_line =~ m/[\w)\]]\s+(\+\+|--)/; # postfix ++/-- with preceding space 676 report("space after '$1'") if $intra_line =~ m/(\+\+|--)\s+[a-zA-Z_(]/; # prefix ++/-- with following space 677 $intra_line =~ s/\.\.\./@/g; # blind '...' 678 report("space before '$1'") if $intra_line =~ m/\s(\.|->)/; # '.' or '->' with preceding space 679 report("space after '$1'") if $intra_line =~ m/(\.|->)\s/; # '.' or '->' with following space 680 $intra_line =~ s/\-\>|\+\+|\-\-/@/g; # blind '->,', '++', and '--' 681 report("space before '$2'") if $intra_line =~ m/[^:]\s+(;)/; # space before ';' but not after ':' 682 report("space before '$1'") if $intra_line =~ m/\s([,)\]])/; # space before ,)] 683 report("space after '$1'") if $intra_line =~ m/([(\[~!])\s/; # space after ([~! 684 report("space after '$1'") if $intra_line =~ m/(defined)\s/; # space after 'defined' 685 report("no space before '=' or '<op>='") if $intra_line =~ m/\S(=)/; # '=' etc. without preceding space 686 report("no space before '$1'") if $intra_line =~ m/\S([|\/%<>^\?])/; # |/%<>^? without preceding space 687 # TODO ternary ':' without preceding SPC, while allowing no SPC before ':' after 'case' 688 report("no space before binary '$1'") if $intra_line =~ m/[^\s{()\[]([+\-])/;# +/- without preceding space or {()[ 689 # or ')' (which is used f type casts) 690 report("no space before binary '$1'") if $intra_line =~ m/[^\s{()\[*!]([*])/; # '*' without preceding space or {()[*! 691 report("no space before binary '$1'") if $intra_line =~ m/[^\s{()\[]([&])/; # '&' without preceding space or {()[ 692 report("no space after ternary '$1'") if $intra_line =~ m/(:)[^\s\d]/; # ':' without following space or digit 693 report("no space after '$1'") if $intra_line =~ m/([,;=|\/%<>^\?])\S/; # ,;=|/%<>^? without following space 694 report("no space after binary '$1'") if $intra_line=~m/[^{(\[]([*])[^\sa-zA-Z_(),*]/;# '*' w/o space or \w(),* after 695 # TODO unary '*' must not be followed by SPC 696 report("no space after binary '$1'") if $intra_line=~m/([&])[^\sa-zA-Z_(]/; # '&' w/o following space or \w( 697 # TODO unary '&' must not be followed by SPC 698 report("no space after binary '$1'") if $intra_line=~m/[^{(\[]([+\-])[^\s\d(]/; # +/- w/o following space or \d( 699 # TODO unary '+' and '-' must not be followed by SPC 700 report("no space after '$2'") if $intra_line =~ m/(^|\W)(if|while|for|switch|case)[^\w\s]/; # kw w/o SPC 701 report("no space after '$2'") if $intra_line =~ m/(^|\W)(return)[^\w\s;]/; # return w/o SPC or ';' 702 report("space after function/macro name") 703 if $intra_line =~ m/(\w+)\s+\(/ # fn/macro name with space before '(' 704 && !($1 =~ m/^(if|while|for|switch|return|typedef|void|char|unsigned|int|long|float|double)$/) # not keyword 705 && !(m/^\s*#\s*define\s/); # we skip macro definitions here because macros 706 # without parameters but with body beginning with '(', e.g., '#define X (1)', 707 # would lead to false positives - TODO also check for macros with parameters 708 report("no space before '{'") if $intra_line =~ m/[^\s{(\[]\{/; # '{' without preceding space or {([ 709 report("no space after '}'") if $intra_line =~ m/\}[^\s,;\])}]/; # '}' without following space or ,;])} 710 } 711 712 # preprocessor directives @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 713 714 # handle preprocessor directives 715 if (m/^\s*#(\s*)(\w+)/) { # line beginning with '#' 716 my $space_count = length($1); # maybe could also use indentation before '#' 717 my $directive = $2; 718 report("indent = $count != 0 for '#'") if $count != 0; 719 $directive_nesting-- if $directive =~ m/^(else|elif|endif)$/; 720 if ($directive_nesting < 0) { 721 $directive_nesting = 0; 722 report("unexpected '#$directive'"); 723 } 724 report("'#' directive nesting = $space_count != $directive_nesting") if $space_count != $directive_nesting; 725 $directive_nesting++ if $directive =~ m/^if|ifdef|ifndef|else|elif$/; 726 $ifdef__cplusplus = m/^\s*#\s*ifdef\s+__cplusplus\s*$/; 727 goto POSTPROCESS_DIRECTIVE unless $directive =~ m/^define$/; # skip normal code handling except for #define 728 # TODO improve handling of indents of preprocessor directives ('\', $in_directive != 0) vs. normal C code 729 $count = -1; # do not check indentation of #define 730 } 731 732 # adapt required indentation @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 733 734 s/(\w*ASN1_[A-Z_]+END\w*([^(]|\(.*?\)|$))/$1;/g; # treat *ASN1_*END*(..) macro calls as if followed by ';' 735 736 my $nested_indents_position = 0; 737 738 # update indents according to leading closing brace(s) '}' or label or switch case 739 my $in_stmt = $in_expr || @nested_symbols != 0 || $in_typedecl != 0; 740 if ($in_stmt) { # expr/stmt/type decl/var def/fn hdr, i.e., not at block level 741 if (m/^([\s@]*\})/) { # leading '}', any preceding blinded comment must not be matched 742 my $head = $1; 743 update_nested_indents($head); 744 $nested_indents_position = length($head); 745 if (@nested_symbols >= 1) { 746 $hanging_symbol = @nested_symbols[-1]; 747 $expr_indent = @nested_indents[-1]; 748 } else { # typically end of initialiizer expr or enum 749 $expr_indent = 0; 750 } 751 } elsif (m/^([\s@]*)(static_)?ASN1_ITEM_TEMPLATE_END(\W|$)/) { # workaround for ASN1 macro indented as '}' 752 $local_offset = -INDENT_LEVEL; 753 $expr_indent = 0; 754 } elsif (m/;.*?\}/) { # expr ends with ';' before '}' 755 report("code before '}'"); 756 } 757 } 758 if (@in_do_hanging_offsets != 0 && # note there is nothing like "unexpected 'while'" 759 m/^[\s@]*while(\W|$)/) { # leading 'while' 760 $hanging_offset = pop @in_do_hanging_offsets; 761 } 762 if ($if_maybe_terminated) { 763 if (m/(^|\W)else(\W|$)/) { # (not necessarily leading) 'else' 764 if (@in_if_hanging_offsets == 0) { 765 report("unexpected 'else'"); 766 } else { 767 $hanging_offset = pop @in_if_hanging_offsets; 768 } 769 } else { 770 @in_if_hanging_offsets = (); # note there is nothing like "unclosed 'if'" 771 $hanging_offset = 0; 772 } 773 } 774 if (!$in_stmt) { # at block level, i.e., outside expr/stmt/type decl/var def/fn hdr 775 $if_maybe_terminated = 0; 776 if (my ($head, $before, $tail) = m/^([\s@]*([^{}]*)\})[\s@]*(.*)$/) { # leading closing '}', but possibly 777 # with non-whitespace non-'{' before 778 report("code after '}'") unless $tail eq "" || $tail =~ m/(else|while|OSSL_TRACE_END)(\W|$)/; 779 my $outermost_level = @nested_block_indents == 1 && @nested_block_indents[0] == 0; 780 if (!$sloppy_bodylen && $outermost_level && $line_body_start != 0) { 781 my $body_len = $line - $line_body_start - 1; 782 report_flexibly($line_function_start, "function body length = $body_len > ".MAX_BODY_LENGTH." lines", 783 $last_function_header) if $body_len > MAX_BODY_LENGTH; 784 $line_body_start = 0; 785 } 786 if ($before ne "") { # non-whitespace non-'{' before '}' 787 report("code before '}'"); 788 } else { # leading '}', any preceding blinded comment must not be matched 789 $local_offset = $block_indent + $hanging_offset - INDENT_LEVEL; 790 update_nested_indents($head); 791 $nested_indents_position = length($head); 792 $local_offset -= ($block_indent + $hanging_offset); 793 # in effect $local_offset = -INDENT_LEVEL relative to $block_indent + $hanging_offset values before 794 } 795 } 796 797 # handle opening brace '{' after if/else/while/for/switch/do on line before 798 if ($hanging_offset > 0 && m/^[\s@]*{/ && # leading opening '{' 799 $line_before > 0 && 800 $contents_before_ =~ m/(^|^.*\W)(if|else|while|for|switch|do)(\W.*$|$)/) { 801 $keyword_opening_brace = $1; 802 $hanging_offset -= INDENT_LEVEL; # cancel newly hanging_offset 803 } 804 805 if (m/^[\s@]*(case|default)(\W.*$|$)/) { # leading 'case' or 'default' 806 my $keyword = $1; 807 report("code after $keyword: ") if $2 =~ /:.*[^\s@].*$/; 808 $local_offset = -INDENT_LEVEL; 809 } else { 810 if (m/^([\s@]*)(\w+):/) { # (leading) label, cannot be "default" 811 $local_offset = -INDENT_LEVEL; 812 $has_label = 1; 813 } 814 } 815 } 816 817 # potential adaptations of indent in first line of macro body in multi-line macro definition 818 if ($in_directive > 0 && $in_macro_header > 0) { 819 if ($in_macro_header > 1) { # still in macro definition header 820 $in_macro_header += parens_balance($_); 821 } else { # begin of macro body 822 $in_macro_header = 0; 823 if ($count == $block_indent - $directive_offset # body began with same indentation as preceding code 824 && $sloppy_macro) { # workaround for this situation is enabled 825 $block_indent -= $directive_offset; 826 $directive_offset = 0; 827 } 828 } 829 } 830 831 # check required indentation @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 832 833 check_indent() if $count >= 0; # not for #define and not if multi-line string literal is continued 834 835 $in_comment = 0 if $in_comment < 0; # multi-line comment has ended 836 837 # do some further checks @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 838 839 my $outermost_level = $block_indent == 0 + ($in_directive > 0 ? $directive_offset : 0); 840 841 report("more than one stmt") if !m/(^|\W)for(\W.*|$)/ && # no 'for' - TODO improve matching 842 m/;.*;/; # two or more terminators ';', so more than one statement 843 844 # check for code block containing a single line/statement 845 if ($line_before2 > 0 && !$outermost_level && # within function body 846 $in_typedecl == 0 && @nested_indents == 0 && # neither within type declaration nor inside stmt/expr 847 m/^[\s@]*\}/) { # leading closing brace '}', any preceding blinded comment must not be matched 848 # TODO extend detection from single-line to potentially multi-line statement 849 if ($line_opening_brace > 0 && 850 ($line_opening_brace == $line_before2 || 851 $line_opening_brace == $line_before) 852 && $contents_before =~ m/;/) { # there is at least one terminator ';', so there is some stmt 853 # TODO do not report cases where a further else branch 854 # follows with a block containg more than one line/statement 855 report_flexibly($line_before, "'$keyword_opening_brace' { 1 stmt }", $contents_before); 856 } 857 } 858 859 report("single-letter name '$2'") if (m/(^|.*\W)([IO])(\W.*|$)/); # single-letter name 'I' or 'O' # maybe re-add 'l'? 860 # constant on LHS of comparison or assignment, e.g., NULL != x or 'a' < c, but not a + 1 == b 861 report("constant on LHS of '$2'") 862 if (m/(['"]|([\+\-\*\/\/%\&\|\^<>]\s*)?\W[0-9]+L?|NULL)\s*([\!<>=]=|[<=>][^<>])/ && $2 eq ""); 863 864 # TODO report #if 0 and #if 1 865 866 # TODO report empty line within local variable definitions 867 868 # TODO report missing empty line after local variable definitions 869 870 # TODO report needless use of parentheses, while 871 # macro parameters should always be in parens (except when passed on), e.g., '#define ID(x) (x)' 872 873 # adapt required indentation for following lines @@@@@@@@@@@@@@@@@@@@@@@@@@@ 874 875 # set $in_expr, $in_paren_expr, and $hanging_offset for if/while/for/switch, return/enum, and assignment RHS 876 my $paren_expr_start = 0; 877 my $return_enum_start = 0; 878 my $assignment_start = 0; 879 my $tmp = $_; 880 $tmp =~ s/[\!<>=]=/@@/g; # blind (in-)equality symbols like '<=' as '@@' to prevent matching them as '=' below 881 if (m/^((^|.*\W)(if|while|for|switch))(\W.*|$)$/) { # (last) if/for/while/switch 882 $paren_expr_start = 1; 883 } elsif (m/^((^|.*\W)(return|enum))(\W.*|$)/ # (last) return/enum 884 && !$in_expr && @nested_indents == 0 && parens_balance($1) == 0) { # not nested enum 885 $return_enum_start = 1; 886 } elsif ($tmp =~ m/^(([^=]*)(=))(.*)$/ # (last) '=', i.e., assignment 887 && !$in_expr && @nested_indents == 0 && parens_balance($1) == 0) { # not nested assignment 888 $assignment_start = 1; 889 } 890 if ($paren_expr_start || $return_enum_start || $assignment_start) 891 { 892 my ($head, $mid, $tail) = ($1, $3, $4); 893 $keyword_opening_brace = $mid if $mid ne "="; 894 # to cope with multi-line expressions, do this also if !($tail =~ m/\{/) 895 push @in_if_hanging_offsets, $hanging_offset if $mid eq "if"; 896 897 # already handle $head, i.e., anything before expression 898 update_nested_indents($head, $nested_indents_position); 899 $nested_indents_position = length($head); 900 # now can set $in_expr and $in_paren_expr 901 $in_expr = 1; 902 $in_paren_expr = 1 if $paren_expr_start; 903 if ($mid eq "while" && @in_do_hanging_offsets != 0) { 904 $hanging_offset = pop @in_do_hanging_offsets; 905 } else { 906 $hanging_offset += INDENT_LEVEL; # tentatively set hanging_offset, may be canceled by following '{' 907 } 908 } 909 910 # set $hanging_offset and $keyword_opening_brace for do/else 911 if (my ($head, $mid, $tail) = m/(^|^.*\W)(else|do)(\W.*|$)$/) { # last else/do, where 'do' is preferred 912 my $code_before = $head =~ m/[^\s\@}]/; # leading non-whitespace non-comment non-'}' 913 report("code before '$mid'") if $code_before; 914 report("code after '$mid'" ) if $tail =~ m/[^\s\@{]/# trailing non-whitespace non-comment non-'{' (non-'\') 915 && !($mid eq "else" && $tail =~ m/[\s@]*if(\W|$)/); 916 if ($mid eq "do") { # workarounds for code before 'do' 917 if ($head =~ m/(^|^.*\W)(else)(\W.*$|$)/) { # 'else' ... 'do' 918 $hanging_offset += INDENT_LEVEL; # tentatively set hanging_offset, may be canceled by following '{' 919 } 920 if ($head =~ m/;/) { # terminator ';' ... 'do' 921 @in_if_hanging_offsets = (); # note there is nothing like "unclosed 'if'" 922 $hanging_offset = 0; 923 } 924 } 925 push @in_do_hanging_offsets, $hanging_offset if $mid eq "do"; 926 if ($code_before && $mid eq "do") { 927 $hanging_offset = length($head) - $block_indent; 928 } 929 if (!$in_paren_expr) { 930 $keyword_opening_brace = $mid if $tail =~ m/\{/; 931 $hanging_offset += INDENT_LEVEL; 932 } 933 } 934 935 # set $in_typedecl and potentially $hanging_offset for type declaration 936 if (!$in_expr && @nested_indents == 0 # not in expression 937 && m/(^|^.*\W)(typedef|struct|union|enum)(\W.*|$)$/ 938 && parens_balance($1) == 0 # not in newly started expression or function arg list 939 && ($2 eq "typedef" || !($3 =~ m/\s*\w++\s*(.)/ && $1 ne "{")) # 'struct'/'union'/'enum' <name> not followed by '{' 940 # not needed: && $keyword_opening_brace = $2 if $3 =~ m/\{/; 941 ) { 942 $in_typedecl++; 943 $hanging_offset += INDENT_LEVEL if m/\*.*\(/; # '*' followed by '(' - seems consistent with Emacs C mode 944 } 945 946 my $bak_in_expr = $in_expr; 947 my $terminator_position = update_nested_indents($_, $nested_indents_position); 948 949 if ($bak_in_expr) { 950 # on end of non-if/while/for/switch (multi-line) expression (i.e., return/enum/assignment) and 951 # on end of statement/type declaration/variable definition/function header 952 if ($terminator_position >= 0 && ($in_typedecl == 0 || @nested_indents == 0)) { 953 check_nested_nonblock_indents("expr"); 954 $in_expr = 0; 955 } 956 } else { 957 check_nested_nonblock_indents($in_typedecl == 0 ? "stmt" : "decl") if $terminator_position >= 0; 958 } 959 960 # on ';', which terminates the current statement/type declaration/variable definition/function declaration 961 if ($terminator_position >= 0) { 962 my $tail = substr($_, $terminator_position + 1); 963 if (@in_if_hanging_offsets != 0) { 964 if ($tail =~ m/\s*else(\W|$)/) { 965 pop @in_if_hanging_offsets; 966 $hanging_offset -= INDENT_LEVEL; 967 } elsif ($tail =~ m/[^\s@]/) { # code (not just comment) follows 968 @in_if_hanging_offsets = (); # note there is nothing like "unclosed 'if'" 969 $hanging_offset = 0; 970 } else { 971 $if_maybe_terminated = 1; 972 } 973 } elsif ($tail =~ m/^[\s@]*$/) { # ';' has been trailing, i.e. there is nothing but whitespace and comments 974 $hanging_offset = 0; # reset in case of terminated assignment ('=') etc. 975 } 976 $in_typedecl-- if $in_typedecl != 0 && @nested_in_typedecl == 0; # TODO handle multiple type decls per line 977 m/(;[^;]*)$/; # match last ';' 978 $terminator_position = length($_) - length($1) if $1; 979 # new $terminator_position value may be after the earlier one in case multiple terminators on current line 980 # TODO check treatment in case of multiple terminators on current line 981 update_nested_indents($_, $terminator_position + 1); 982 } 983 984 # set hanging expression indent according to nested indents - TODO maybe do better in update_nested_indents() 985 # also if $in_expr is 0: in statement/type declaration/variable definition/function header 986 $expr_indent = 0; 987 for (my $i = -1; $i >= -@nested_symbols; $i--) { 988 if (@nested_symbols[$i] ne "?") { # conditionals '?' ... ':' are treated specially in check_indent() 989 $hanging_symbol = @nested_symbols[$i]; 990 $expr_indent = $nested_indents[$i]; 991 # $expr_indent is guaranteed to be != 0 unless @nested_indents contains just outer conditionals 992 last; 993 } 994 } 995 996 # remember line number and header containing name of last function defined for reports w.r.t. MAX_BODY_LENGTH 997 if ($outermost_level && m/(\w+)\s*\(/ && $1 ne "STACK_OF") { 998 $line_function_start = $line; 999 $last_function_header = $contents; 1000 } 1001 1002 # special checks for last, typically trailing opening brace '{' in line 1003 if (my ($head, $tail) = m/^(.*)\{(.*)$/) { # match last ... '{' 1004 if ($in_directive == 0 && !$in_expr && $in_typedecl == 0) { 1005 if ($outermost_level) { 1006 if (!$assignment_start && !$bak_in_expr) { 1007 # at end of function definition header (or stmt or var definition) 1008 report("'{' not at beginning") if $head ne ""; 1009 $line_body_start = $contents =~ m/LONG BODY/ ? 0 : $line; 1010 } 1011 } else { 1012 $line_opening_brace = $line if $keyword_opening_brace =~ m/do|while|for/; 1013 # using, not assigning, $keyword_opening_brace here because it could be on an earlier line 1014 $line_opening_brace = $line if $keyword_opening_brace =~ m/if|else/ && $extended_1_stmt && 1015 # TODO prevent false positives for if/else where braces around single-statement branches 1016 # should be avoided but only if all branches have just single statements 1017 # The following helps detecting the exception when handling multiple 'if ... else' branches: 1018 !($keyword_opening_brace eq "else" && $line_opening_brace < $line_before2); 1019 } 1020 report("code after '{'") if $tail=~ m/[^\s\@]/ && # trailing non-whitespace non-comment (non-'\') 1021 !($tail=~ m/\}/); # no '}' after last '{' 1022 } 1023 } 1024 1025 # check for opening brace after if/while/for/switch/do not on same line 1026 # note that "no '{' on same line after '} else'" is handled further below 1027 if (/^[\s@]*{/ && # leading '{' 1028 $line_before > 0 && !($contents_before_ =~ m/^\s*#/) && # not preprocessor directive '#if 1029 (my ($head, $mid, $tail) = ($contents_before_ =~ m/(^|^.*\W)(if|while|for|switch|do)(\W.*$|$)/))) { 1030 my $brace_after = $tail =~ /^[\s@]*{/; # any whitespace or comments then '{' 1031 report("'{' not on same line as preceding '$mid'") if !$brace_after; 1032 } 1033 # check for closing brace on line before 'else' not followed by leading '{' 1034 elsif (my ($head, $tail) = m/(^|^.*\W)else(\W.*$|$)/) { 1035 if (parens_balance($tail) == 0 && # avoid false positive due to unfinished expr on current line 1036 !($tail =~ m/{/) && # after 'else' no '{' on same line 1037 !($head =~ m/}[\s@]*$/) && # not: '}' then any whitespace or comments before 'else' 1038 $line_before > 0 && $contents_before_ =~ /}[\s@]*$/) { # trailing '}' on line before 1039 report("no '{' after '} else'"); 1040 } 1041 } 1042 1043 # check for closing brace before 'while' not on same line 1044 if (my ($head, $tail) = m/(^|^.*\W)while(\W.*$|$)/) { 1045 my $brace_before = $head =~ m/}[\s@]*$/; # '}' then any whitespace or comments 1046 # possibly 'if (...)' (with potentially inner '(' and ')') then any whitespace or comments then '{' 1047 if (!$brace_before && 1048 # does not work here: @in_do_hanging_offsets != 0 && #'while' terminates loop 1049 parens_balance($tail) == 0 && # avoid false positive due to unfinished expr on current line 1050 $tail =~ /;/ && # 'while' terminates loop (by ';') 1051 $line_before > 0 && 1052 $contents_before_ =~ /}[\s@]*$/) { # on line before: '}' then any whitespace or comments 1053 report("'while' not on same line as preceding '}'"); 1054 } 1055 } 1056 1057 # check for missing brace on same line before or after 'else' 1058 if (my ($head, $tail) = m/(^|^.*\W)else(\W.*$|$)/) { 1059 my $brace_before = $head =~ /}[\s@]*$/; # '}' then any whitespace or comments 1060 my $brace_after = $tail =~ /^[\s@]*if[\s@]*\(.*\)[\s@]*{|[\s@]*{/; 1061 # possibly 'if (...)' (with potentially inner '(' and ')') then any whitespace or comments then '{' 1062 if (!$brace_before) { 1063 if ($line_before > 0 && $contents_before_ =~ /}[\s@]*$/) { 1064 report("'else' not on same line as preceding '}'"); 1065 } elsif (parens_balance($tail) == 0) { # avoid false positive due to unfinished expr on current line 1066 report("no '}' on same line before 'else ... {'") if $brace_after; 1067 } 1068 } elsif (parens_balance($tail) == 0) { # avoid false positive due to unfinished expr on current line 1069 report("no '{' on same line after '} else'") if $brace_before && !$brace_after; 1070 } 1071 } 1072 1073 POSTPROCESS_DIRECTIVE: 1074 # on begin of multi-line preprocessor directive, adapt indent 1075 # need to use original line contents because trailing '\' may have been stripped above 1076 if ($contents =~ m/^(.*?)[\s@]*\\[\s@]*$/) { # trailing '\' (which is not stripped from $contents), 1077 # typically used in macro definitions (or other preprocessor directives) 1078 if ($in_directive == 0) { 1079 $in_macro_header = m/^\s*#\s*define(\W|$)?(.*)/ ? 1 + parens_balance($2) : 0; # '#define' is beginning 1080 $directive_offset = INDENT_LEVEL; 1081 $block_indent += $directive_offset; 1082 } 1083 $in_directive += 1; 1084 } 1085 1086 # post-processing at end of line @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 1087 1088 LINE_FINISHED: 1089 # on end of multi-line preprocessor directive, adapt indent 1090 if ($in_directive > 0 && 1091 # need to use original line contents because trailing \ may have been stripped 1092 !($contents =~ m/^(.*?)[\s@]*\\[\s@]*$/)) { # no trailing '\' 1093 $block_indent -= $directive_offset; 1094 $in_directive = 0; 1095 # macro body typically does not include terminating ';' 1096 $hanging_offset = 0; # compensate for this in case macro ends, e.g., as 'while (0)' 1097 } 1098 1099 if (m/^\s*$/) { # at begin of file essentially empty line: just whitespace (and maybe a '\') 1100 report("leading ".($1 eq "" ? "empty" :"whitespace")." line") if $line == 1 && !$sloppy_SPC; 1101 } else { 1102 if ($line_before > 0) { 1103 my $linediff = $line - $line_before - 1; 1104 report("$linediff empty lines before") if $linediff > 1 && !$sloppy_SPC; 1105 } 1106 $line_before2 = $line_before; 1107 $contents_before2 = $contents_before; 1108 $contents_before_2 = $contents_before_; 1109 $line_before = $line; 1110 $contents_before = $contents; 1111 $contents_before_ = $_; 1112 $count_before = $count; 1113 } 1114 1115 if ($self_test) { # debugging 1116 my $should_report = $contents =~ m/\*@(\d)?/ ? 1 : 0; 1117 $should_report = +$1 if $should_report != 0 && defined $1; 1118 print("$ARGV:$line:$num_reports_line reports on:$contents") 1119 if $num_reports_line != $should_report; 1120 } 1121 $num_reports_line = 0; 1122 1123 # post-processing at end of file @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 1124 1125 if (eof) { 1126 # check for essentially empty line (which may include a '\') just before EOF 1127 report(($1 eq "\n" ? "empty line" : $2 ne "" ? "'\\'" : "whitespace")." at EOF") 1128 if $contents =~ m/^(\s*(\\?)\s*)$/ && !$sloppy_SPC; 1129 1130 # report unclosed expression-level nesting 1131 check_nested_nonblock_indents("expr at EOF"); # also adapts @nested_block_indents 1132 1133 # sanity-check balance of block-level { ... } via final $block_indent at end of file 1134 report_flexibly($line, +@nested_block_indents." unclosed '{'", "(EOF)\n") if @nested_block_indents != 0; 1135 1136 # sanity-check balance of #if ... #endif via final preprocessor directive indent at end of file 1137 report_flexibly($line, "$directive_nesting unclosed '#if'", "(EOF)\n") if $directive_nesting != 0; 1138 1139 reset_file_state(); 1140 } 1141} 1142 1143# final summary report @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 1144 1145my $num_other_reports = $num_reports - $num_indent_reports - $num_nesting_issues 1146 - $num_syntax_issues - $num_SPC_reports - $num_length_reports; 1147print "$num_reports ($num_indent_reports indentation, $num_nesting_issues directive nesting, ". 1148 "$num_syntax_issues syntax, $num_SPC_reports whitespace, $num_length_reports length, $num_other_reports other)". 1149 " issues have been found by $0\n" if $num_reports != 0 && !$self_test; 1150