module main import ( os term ) ////////////////////////////////////////////////////////////////////////////////////////////////// // NB: The code in this file is organized in layers (between the ///// lines). // This allows for easier keeping in sync of error/warn functions. // The functions in each of the layers, call the functions from the layers *below*. // The functions in each of the layers, also have more details about the warn/error situation, // so they can display more informative message, so please call the lowest level variant you can. ////////////////////////////////////////////////////////////////////////////////////////////////// // TLDR: If you have a token index, call: // p.error_with_token_index(msg, token_index) // ... not just : // p.error(msg) ////////////////////////////////////////////////////////////////////////////////////////////////// fn (p mut Parser) error(s string) { // no positioning info, so just assume that the last token was the culprit: p.error_with_token_index(s, p.token_idx-1 ) } fn (p mut Parser) warn(s string) { p.warn_with_token_index(s, p.token_idx-1 ) } ////////////////////////////////////////////////////////////////////////////////////////////////// fn (p mut Parser) production_error_with_token_index(e string, tokenindex int) { if p.pref.is_prod { p.error_with_token_index( e, tokenindex ) }else { p.warn_with_token_index( e, tokenindex ) } } ////////////////////////////////////////////////////////////////////////////////////////////////// fn (p mut Parser) error_with_token_index(s string, tokenindex int) { p.error_with_position(s, p.scanner.get_scanner_pos_of_token( p.tokens[ tokenindex ] ) ) } fn (p mut Parser) warn_with_token_index(s string, tokenindex int) { p.warn_with_position(s, p.scanner.get_scanner_pos_of_token( p.tokens[ tokenindex ] ) ) } ////////////////////////////////////////////////////////////////////////////////////////////////// fn (p mut Parser) error_with_position(s string, sp ScannerPos) { p.print_error_context() e := normalized_error( s ) p.scanner.goto_scanner_position( sp ) p.scanner.error_with_col(e, sp.pos - sp.last_nl_pos) } fn (p mut Parser) warn_with_position(s string, sp ScannerPos) { // on a warning, restore the scanner state after printing the warning: cpos := p.scanner.get_scanner_pos() e := normalized_error( s ) p.scanner.goto_scanner_position( sp ) p.scanner.warn_with_col(e, sp.pos - sp.last_nl_pos) p.scanner.goto_scanner_position( cpos ) } ////////////////////////////////////////////////////////////////////////////////////////////////// fn (s &Scanner) error(msg string) { s.error_with_col(msg, 0) } fn (s &Scanner) warn(msg string) { s.warn_with_col(msg, 0) } ////////////////////////////////////////////////////////////////////////////////////////////////// fn (s &Scanner) warn_with_col(msg string, col int) { fullpath := s.get_error_filepath() color_on := s.is_color_output_on() final_message := if color_on { term.bold(term.bright_blue( msg )) } else { msg } eprintln('warning: ${fullpath}:${s.line_nr+1}:${col}: $final_message') } fn (s &Scanner) error_with_col(msg string, col int) { fullpath := s.get_error_filepath() color_on := s.is_color_output_on() final_message := if color_on { term.red( term.bold( msg ) ) } else { msg } // The filepath:line:col: format is the default C compiler // error output format. It allows editors and IDE's like // emacs to quickly find the errors in the output // and jump to their source with a keyboard shortcut. // NB: using only the filename may lead to inability of IDE/editors // to find the source file, when the IDE has a different working folder than v itself. eprintln('${fullpath}:${s.line_nr + 1}:${col}: $final_message') if s.should_print_line_on_error && s.file_lines.len > 0 { context_start_line := imax(0, (s.line_nr - error_context_before + 1 )) context_end_line := imin(s.file_lines.len, (s.line_nr + error_context_after + 1 )) for cline := context_start_line; cline < context_end_line; cline++ { line := '${(cline+1):5d}| ' + s.file_lines[ cline ] coloredline := if cline == s.line_nr && color_on { term.red(line) } else { line } eprintln( coloredline ) if cline != s.line_nr { continue } // The pointerline should have the same spaces/tabs as the offending // line, so that it prints the ^ character exactly on the *same spot* // where it is needed. That is the reason we can not just // use strings.repeat(` `, col) to form it. mut pointerline := []string for i , c in line { if i < col { x := if c.is_space() { c } else { ` ` } pointerline << x.str() continue } pointerline << if color_on { term.bold( term.blue('^') ) } else { '^' } break } eprintln( ' ' + pointerline.join('') ) } } exit(1) } ////////////////////////////////////////////////////////////////////////////////////////////////// /// Misc error helper functions, can be called by any of the functions above [inline] fn (p &Parser) cur_tok_index() int { return p.token_idx - 1 } [inline] fn imax(a,b int) int { return if a > b { a } else { b } } [inline] fn imin(a,b int) int { return if a < b { a } else { b } } fn (s &Scanner) get_error_filepath() string { if s.should_print_relative_paths_on_error { return s.file_path } return os.realpath( s.file_path ) } fn (s &Scanner) is_color_output_on() bool { return s.should_print_errors_in_color && term.can_show_color_on_stderr() } fn (p mut Parser) print_error_context(){ // Dump all vars and types for debugging if p.pref.is_debug { // os.write_to_file('/var/tmp/lang.types', '')//pes(p.table.types)) os.write_file('fns.txt', p.table.debug_fns()) } if p.pref.is_verbose || p.pref.is_debug { println('pass=$p.pass fn=`$p.cur_fn.name`\n') } p.cgen.save() // V up hint cur_path := os.getwd() if !p.pref.is_repl && !p.pref.is_test && ( p.file_path.contains('v/compiler') || cur_path.contains('v/compiler') ){ println('\n=========================') println('It looks like you are building V. It is being frequently updated every day.') println('If you didn\'t modify V\'s code, most likely there was a change that ') println('lead to this error.') println('\nRun `v up`, that will most likely fix it.') //println('\nIf this doesn\'t help, re-install V from source or download a precompiled' + ' binary from\nhttps://vlang.io.') println('\nIf this doesn\'t help, please create a GitHub issue.') println('=========================\n') } if p.pref.is_debug { print_backtrace() } // p.scanner.debug_tokens() } fn normalized_error( s string ) string { // Print `[]int` instead of `array_int` in errors return s.replace('array_', '[]') .replace('__', '.') .replace('Option_', '?') .replace('main.', '') } ////////////////////////////////////////////////////////////////////////////////////////////////// // The goal of ScannerPos is to track the current scanning position, // so that if there is an error found later, v could show a more accurate // position about where the error initially was. // NB: The fields of ScannerPos *should be kept synchronized* with the // corresponding fields in Scanner. struct ScannerPos { mut: pos int line_nr int last_nl_pos int } fn (s ScannerPos) str() string { return 'ScannerPos{ ${s.pos:5d} , ${s.line_nr:5d} , ${s.last_nl_pos:5d} }' } fn (s &Scanner) get_scanner_pos() ScannerPos { return ScannerPos{ pos: s.pos line_nr: s.line_nr last_nl_pos: s.last_nl_pos } } fn (s mut Scanner) goto_scanner_position(scp ScannerPos) { s.pos = scp.pos s.line_nr = scp.line_nr s.last_nl_pos = scp.last_nl_pos } // get_scanner_pos_of_token rescans *the whole source* till it reaches {t.line_nr, t.col} . fn (s mut Scanner) get_scanner_pos_of_token(t &Token) ScannerPos { // This rescanning is done just once on error, so it is fine for now. // Be careful for the performance implications, if you want to // do it more frequently. The alternative would be to store // the scanpos (12 bytes) for each token, and there are potentially many tokens. tline := t.line_nr tcol := if t.line_nr == 0 { t.col + 1 } else { t.col - 1 } // save the current scanner position, it will be restored later cpos := s.get_scanner_pos() mut sptoken := ScannerPos{} // Starting from the start, scan the source lines // till the desired tline is reached, then // s.pos + tcol would be the proper position // of the token. Continue scanning for some more lines of context too. s.goto_scanner_position(ScannerPos{}) s.file_lines = []string mut prevlinepos := 0 // NB: TCC BUG workaround: removing the `mut ate:=0 ate++` line // below causes a bug in v, when v is compiled with tcc, and v // wants to report the error: 'the following imports were never used:' // // This can be reproduced, if you follow the steps: // a) ./v -cc tcc -o v compiler ; // b) ./v vlib/builtin/hashmap_test.v' // // In this case, prevlinepos gets a random value on each run. // Any kind of operation may be used seemingly, as long as // there is a new stack allocation that will 'protect' prevlinepos. ////////////////////////////////////////////////////////////////// mut ate:=0 ate++ // This var will be smashed by TCC, instead of /////////////////// prevlinepos. The cause is the call to /////////////////// s.get_scanner_pos() /////////////////// which just returns a struct, and that works /////////////////// in gcc and clang, but causes the TCC problem. for { prevlinepos = s.pos if s.pos >= s.text.len { break } if s.line_nr > tline + 10 { break } //////////////////////////////////////// if tline == s.line_nr { sptoken = s.get_scanner_pos() sptoken.pos += tcol } s.ignore_line() s.eat_single_newline() sline := s.text.substr( prevlinepos, s.pos )//.trim_right('\r\n') s.file_lines << sline } ////////////////////////////////////////////////// s.goto_scanner_position(cpos) return sptoken } fn (s mut Scanner) eat_single_newline(){ if s.pos >= s.text.len { return } if s.expect('\r\n', s.pos) { s.pos += 2 return } if s.text[ s.pos ] == `\n` { s.pos ++ return } if s.text[ s.pos ] == `\r` { s.pos ++ return } }