#include #include #include #include "ftnchek.h" #include "symtab.h" #include "tokdefs.h" #include "forlex.h" #include "advance.h" PROTO(PRIVATE void find_comments,(srcLine *Buf)); PROTO(PRIVATE void find_fixed_contins,(srcLine *Buf)); PROTO(PRIVATE void find_free_contins,(srcLine *Buf)); PROTO(PRIVATE int line_is_blank,(char *s)); /* Function to read in a source file and store it in a linked list of srcLine elements. Returns pointer to head of the list. If file is empty, returns NULL. */ srcLine* gulp_srcfile(FILE *fd) { char linebuf[MAXLINE+1]; LINENO_t line_num=0; int line_len; srcLine *firstLine=NULL, *prevLine=NULL, *thisLine; while( fgets(linebuf, sizeof(linebuf), fd ) != NULL ) { LINENO_t new_line_num = 0; line_len = strlen(linebuf); /* this counts the \n */ /* If input line was too long to fit into buffer, gobble what was not read and discard it. */ if(line_len == MAXLINE && linebuf[MAXLINE-1] != ENDL) { int c; while( (c = getc(fd)) != ENDL && c != EOF ) continue; } else { /* Replace \n by null. line_len>0 is guaranteed */ if(linebuf[line_len-1] == ENDL) linebuf[--line_len] = '\0'; /* handle */ if( line_len > 1 && linebuf[line_len-1] == '\r' ) { linebuf[--line_len] = '\0'; } } /* Allocate a new struct for this line */ if( (thisLine = (srcLine *)malloc(sizeof(srcLine))) == (srcLine *)NULL || (thisLine->line = (char *)malloc(line_len+1)) == (char *)NULL ) { fflush(list_fd); fprintf(stderr,"Oops: Out of memory in gulp_srcfile\n"); exit(1); } /* Remember the first line as header */ if( firstLine == NULL ) firstLine = thisLine; /* Fill in the struct */ strcpy(thisLine->line,linebuf); ++line_num; /* Advance the line count */ thisLine->line_num = line_num; thisLine->printed = FALSE; thisLine->contin = FALSE; /* this will be set by lexer */ thisLine->cpp_line_directive = FALSE; /* Link it into the list */ if( prevLine != NULL ) prevLine->next = thisLine; thisLine->prev = prevLine; thisLine->next = NULL; prevLine = thisLine; /* Let # line directives override the above increment of line_num. The given value will take effect on the line following the # line directive. */ if( linebuf[0] == '#' ) { char *s = linebuf; do { ++s; } while( isspace(*s) ); /* Skip space after the '#' */ if(strncmp(s,"line",4) == 0) { /* Look for the keyword "line" */ s += 4; /* Skip the word "line" */ while( isspace(*s) ) ++s; /* Skip space after the word "line" */ } if( isdigit(*s) ) { /* See that we are now looking at a number */ /* Get the line number */ new_line_num=0; while( isdigit(*s) ) new_line_num = new_line_num*10 + BCD(*s++); thisLine->cpp_line_directive = TRUE; /* Note we did it */ } } if( new_line_num != 0 ) { /* This is a # line directive */ line_num = new_line_num-1; /* number it 1 less than next */ } } (void) fclose(fd); /* Now go thru and mark comment and continuation lines */ if( firstLine != (srcLine *)NULL ) { find_comments(firstLine); if( free_form ) { /* free source form */ find_free_contins(firstLine); } else { /* fixed source form */ find_fixed_contins(firstLine); } } #ifdef DEBUG_GULP_SRCFILE if(debug_latest) { /* print the source file marking comments & contins */ srcLine *p; for(p = firstLine; p != NULL; p = p->next ) { fprintf(list_fd,"\n%4d%c", p->line_num, p->comment?(p->f90_comment?'!':'C'): (p->contin?'&':' ') ); fprintf(list_fd,"%s",p->line); if(! p->comment ) { int i, last_i; last_i = strlen(p->line)-1; if(last_i < p->start_index) last_i = p->start_index; if(last_i < p->end_index) last_i = p->end_index; if(last_i >= MAXLINE) /* just in case something's bogus */ last_i = MAXLINE-1; for(i=0; i<=last_i; i++) linebuf[i] = ' '; if( p->start_index >= MAXLINE || p->end_index >= MAXLINE ) linebuf[last_i] = '!'; /* signals a bug */ else { if(p->start_index == p->end_index) /* coincident */ linebuf[p->start_index] = '%'; else { linebuf[p->start_index] = '^'; /* points to start */ linebuf[p->end_index] = '$'; /* points to end */ } } linebuf[last_i+1] = '\0'; fprintf(list_fd,"\n%5s",""); fprintf(list_fd,"%s",linebuf); } #include "PRINT_srcBuf.h" PRINT_srcBuf(p); } } #endif return firstLine; } /* Function to find lines that are entirely comment (incl blank lines) */ PRIVATE void find_comments(srcLine *Buf) { while( Buf != (srcLine *)NULL ) { char *s = Buf->line; int i,c= makeupper(s[0]); int allspace; COLNO_t col; Buf->comment = FALSE; Buf->blank = FALSE; Buf->d_comment = FALSE; Buf->f90_comment = FALSE; Buf->empty_contin = FALSE; if( !free_form ) { /* Handle F77 standard comments here. */ if( c == 'C' || c == '*' ) Buf->comment = TRUE; /* Tolerate D comment lines. There is no provision for optionally treating them as source code lines. */ if( c == 'D' ) { Buf->comment = TRUE; Buf->d_comment = TRUE; } } /* Now see if line is blank or only contains an inline comment. */ allspace = TRUE; for(i=0,col=1; s[i] != '\0'; i++) { if( !isspace(s[i])) { allspace = FALSE; /* Initial "!" starts a comment, except in col. 6 of fixed form it must be taken as continuation mark */ if(s[i]==INLINE_COMMENT_CHAR && (free_form || col != 6) ) { Buf->comment = TRUE; Buf->f90_comment = TRUE; } /* Standard 3.3.1.3 prohibits an & on a line with nothing but blanks or commentary. We allow it but flag it here. It is treated as a comment rather than a continuation. */ else if( s[i] == '&' && free_form ) { i++; while( s[i] != '\0' && (isspace(s[i])) ) { i++; } if( s[i] == '\0' || s[i] == INLINE_COMMENT_CHAR ) { Buf->f90_comment = (s[i] == INLINE_COMMENT_CHAR); Buf->comment = TRUE; /* treat as comment line */ Buf->empty_contin = TRUE; } else break; } else break; } else { col = NXTCOL(s[i],col); } } /* end for */ if( allspace ) { Buf->blank = TRUE; /* blank line */ Buf->comment = TRUE; } #if 0 fprintf(list_fd,"\n%c",Buf->comment? 'C':(Buf->contin?'&':' ')); fprintf(list_fd,"%s",Buf->line); #endif Buf = Buf->next; /* advance to next line */ } } /* Functions to find and mark continuation lines. Note that first line of continued statement is NOT marked. These functions also set the cursors start_index and end_index of each line to the first and last significant character respectively of line. (It is possible for start_index to point to '\0' in some cases. Also, end_index is set to -1 for fixed-form continuation line that is blank.) For comment lines, the cursors are set to harmless values, but they should not be needed. The column number start_col corresponding to start_index after accounting for tabs (including -source=dec-tab convention) is also set, but there is no need for an end_col. The main difference between the two functions, besides the fundamental one of the free-form final '&' vs. fixed-form column-6 mark, is that in fixed-form we need to ignore spaces and pay attention to column numbers to avoid running past column 72 (or in general max_stmt_col). These functions would be simple except that a '!' or '&' must not be treated as ending a line if it is inside a string or a hollerith. (In this comment, an H edit descriptor is considered a hollerith.) Strings are pretty easy to deal with, but holleriths cause a lot of trouble. In order not to see a hollerith in a variable like X1H1 or in DO 4 H=1,10 we require it to be preceded by a punctuation character. It is worth noting that the code allows a hollerith to immediately follow a string, which is improper but accepted by some compilers (and our lexer/parser) in a FORMAT. I believe that all cases where a hollerith could be matched by this code in a place where no hollerith should be are illegal, so all we need to do is make the parser squawk, which will happen. Likewise it will always spot a hollerith if it sees an integer preceded by punctuation and followed by 'H', which can never miss a valid one. In any event a false identification or failure to identify a hollerith can only cause trouble if a '!' or (for free-form) '&' is in the span of the (real or falsely supposed) hollerith. In both functions (free and fixed versions) the variable inside_number is set TRUE while scanning an integer that could be the length part of a hollerith; num_val accumulates the value of this integer; prev_sig_char remembers the previous significant character, which must be punctuation (ispunct) in order for inside_number to be turned on by seeing a digit. Similarly, inside_quote is TRUE while scanning a string, and quoteChar remembers the opening quote type (' or ") to determine when the closing quote is found. */ PRIVATE void find_free_contins(srcLine *Buf) { int to_be_continued = FALSE; /* remembers '&' of previous line */ int inside_quote = FALSE; int quoteChar; /* opening quote character */ int inside_number = FALSE; /* used for continued hollerith numbers */ int prev_sig_char = ' '; /* previous significant character */ int num_val; /* value of number recently seen */ while( Buf != (srcLine *)NULL ) { Buf->overlength = FALSE; if( Buf->comment) { /* skip comment lines */ Buf->contin = FALSE; Buf->start_index = 0; Buf->end_index = -1; Buf->start_col = 0; } else { /* not comment */ COLNO_t col; int i, c; char *s = Buf->line; Buf->contin = to_be_continued; /* mark this line */ Buf->end_index = -1; /* look for optional initial '&' on continued line */ i = 0; col = 1; if( inside_quote ) { /* Record the start of significant text. Inside a quote: line must be continued, and these may be overruled by finding a leading '&' later (which std requires but we don't) . */ Buf->start_index = i; Buf->start_col = col; Buf->contin_wo_amp = TRUE; } while( isspace(s[i]) ) {/* skip leading space */ col = NXTCOL(s[i],col); i++; } if( to_be_continued ) { /* deal with continued statements */ if( s[i] == '&' ) { Buf->contin_wo_amp = FALSE; i++; /* resume after the '&' */ col++; if( inside_quote ) { /* blanks are significant inside strings */ Buf->start_index = i; Buf->start_col = col; } } /* Handle the unlikely case that a hollerith is split within or just after the initial number. The usual case is to skip over blanks following the initial '&' if any. */ if( ! inside_number || !( (i == 0 || s[i-1] == '&') && (makeupper(s[i]) == 'H' || isdigit(s[i])) ) ) { inside_number = FALSE; while( isspace(s[i]) ) {/* skip space after & */ col = NXTCOL(s[i],col); i++; } } } else { inside_quote = FALSE; /* clean up unfinished business */ inside_number = FALSE; prev_sig_char = ' '; } /* Record the start of significant text. Continued quoted strings were handled above; all other situations are done here. We have at all events skipped over all leading space and, if a continuation line, over any initial '&' and following nonquoted space. */ if( ! inside_quote ) { Buf->start_index = i; Buf->start_col = col; } to_be_continued = FALSE; for( ; (c=s[i]) != '\0'; i++ ) { if( inside_quote ) { if( c == quoteChar ) { if( s[i+1] == quoteChar ) { /* escaped quote */ i++; col++; } else inside_quote = FALSE; } /* inside quote '&' is continuation if rest of line blank */ else if( c == '&' && line_is_blank(s+i+1) ) { to_be_continued = TRUE; break; } /* escaped quote via Unix backslash */ else if( source_unix_backslash && c == '\\' && s[i+1] == quoteChar ) { i++; col++; } } else if( c == '\'' || c == '"' ) { inside_quote = TRUE; quoteChar = c; inside_number = FALSE; } /* Not in a quoted string or starting one */ else { if( isdigit(c) ) { /* number found: see if hollerith */ if( inside_number ) { /* this may be a continued number */ num_val = num_val*10 + BCD(c); } else { /* hollerith can only follow punctuation */ if( ispunct(prev_sig_char) ) { num_val=BCD(c); /* start it off */ inside_number = TRUE; } } } if( inside_number ) { if( ! (isdigit(c) || c == '&') ) { inside_number = FALSE; /* cannot be continued */ } if( makeupper(c) == 'H' ) { /* skip past hollerith */ int hollerith_count=0; /* We can't just say i += num since line may end before then. (AFAIK holleriths cannot be continued in free form, so premature end of line can be taken as end of stmt.) */ i++; col++; /* move past 'H' */ for(hollerith_count = num_val; (c=s[i]) != '\0' && hollerith_count>0 ; /* Hmmm: how should tabs in holleriths be counted? As one column or many? We say many. */ hollerith_count--, i++, col = NXTCOL(c,col)) { Buf->end_index = i; } hollerith_count = 0; /* if line ends before H does */ } } /* end of dealing with number and maybe hollerith */ /* Start of a comment: quit i loop. */ if( c == INLINE_COMMENT_CHAR ) { Buf->f90_comment = TRUE; break; } else if( c == '&' ) { /* Found freeform contin mark. Check * that naught but comments follow. */ to_be_continued = TRUE; for(i++, col++; (c=s[i]) != '\0' && isspace(c); i++, col = NXTCOL(c,col)) continue; /* If something does follow the contin mark, set end_index to point to it so the parser will squawk. This error is too rare to deserve specific handling. */ if(c == INLINE_COMMENT_CHAR) Buf->f90_comment = TRUE; else if(c != '\0') Buf->end_index = i; break; } } /* end of stuff outside quotes */ /* Re-test for end of stmt needed here due to advancing i above. This should never happen when inside_quote but source may be illegal. */ if( c == '\0' ) break; /* As long as interesting stuff is seen, * update end index. This should never * leave end_index at -1 since we know line * is not a comment so not all blank. */ if( inside_quote || ! isspace(c) ) { Buf->end_index = i; prev_sig_char = c; /* set up std violation warning if significant chars past std max line length */ if( col > std_max_stmt_col ) Buf->overlength = TRUE; } } /* end for i */ } /* end else not comment */ Buf = Buf->next; } /* end while Buf */ }/*find_free_contins*/ PRIVATE void find_fixed_contins(srcLine *Buf) { int inside_quote = FALSE; int quoteChar; /* opening quote character */ int hollerith_count=0; /* used to deal with continued holleriths */ int inside_number = FALSE; /* used for continued hollerith numbers */ int prev_sig_char = ' '; /* previous significant character */ int num_val; /* value of number recently seen */ int tab_count = 0; /* tabs seen, for -port=tab option */ while( Buf != (srcLine *)NULL ) { Buf->overlength = FALSE; Buf->contin = FALSE; if( Buf->comment) { /* skip comment lines */ Buf->start_index = 0; Buf->end_index = -1; Buf->start_col = 0; } else { COLNO_t col; int i,c; char *s = Buf->line; /* Handle DEC tabs: followed by nonzero digit is a continuation line */ if( source_dec_tab && s[0] == '\t' ) { ++tab_count; if( isadigit((int)s[1]) && s[1] != '0' ) { Buf->contin = TRUE; Buf->start_index = 2; /* stmt char after contin mark */ Buf->start_col = 8; /* treat start of stmt as col 8 */ } } else { /* skip to col 6 */ for(i=0,col=1; col < 6 && s[i] != '\0'; i++) { col = PORT_NXTCOL(s[i],col); } c = s[i]; if( col == 6 && c != '\0' && !isspace(c) && c != '0' #ifdef ALLOW_UNIX_CPP /* Veto if it is a preprocessor line */ && s[0] != '#' #endif ) { Buf->contin = TRUE; Buf->start_index = i+1; /* record where stmt resumes */ Buf->start_col = 7; } } /* Locate the start and end of significant * source text. */ if( Buf->contin ) { /* resume after any continuation mark */ i=Buf->start_index; col=Buf->start_col; } else { /* For DEC tabs, an initial tab puts us in column 7 */ #ifdef DEC_TABS if( source_dec_tab && s[0] == '\t' ) { i = 1; col = 7; } else #endif { i=0; /* otherwise start at column 1 */ col=1; } hollerith_count = 0; /* clear unfinished business */ inside_quote = FALSE; inside_number = FALSE; prev_sig_char = ' '; } if( hollerith_count>0 ) { for(; /* Eat any continued hollerith */ (c=s[i]) != '\0' && hollerith_count>0 && col<=max_stmt_col; hollerith_count--, i++, col=PORT_NXTCOL(c,col)) { Buf->end_index = i; } if( hollerith_count>0 ) { /* fill out implied blanks */ hollerith_count -= (max_stmt_col - col + 1); } } /* Skip over leading blank space */ else if( ! inside_quote ) { for( ; (c=s[i]) != '\0' && isspace(c); i++, col=PORT_NXTCOL(c,col) ) { continue; } Buf->start_index = i; Buf->start_col = col; } /* In fixed form, end_index can fail to get set if line consists of no significant characters other than a continuation mark. This is OK if it is equal to start_index so an EOL will be sent by advance. */ Buf->end_index = Buf->start_index; /* Now scan rest of line for end of significant text */ for( ; (c=s[i]) != '\0'; i++, col=PORT_NXTCOL(c,col) ) { if( inside_quote ) { /* Handle escaped quote. This code does not spot escaped quote split across continuation, but it gets the begin and end markers right anyway. Unix backslash style can't be split since it would escape newline instead. */ if( c == quoteChar ) { if( s[i+1] == quoteChar ) {/* escaped quote */ i++; col++; } else inside_quote = FALSE; } /* escaped quote via Unix backslash */ else if( source_unix_backslash && c == '\\' && s[i+1] == quoteChar ) { i++; col++; } } else if( c == '\'' || c == '"' ) { inside_quote = TRUE; quoteChar = c; inside_number = FALSE; } /* Not in a quoted string or starting one */ else { if( isdigit(c) ) { /* number found: see if hollerith */ if( inside_number ) { /* this may be a continued number */ num_val = num_val*10 + BCD(c); } else { /* hollerith can only follow punctuation */ if( ispunct(prev_sig_char) ) { num_val=BCD(c); /* start it off */ inside_number = TRUE; } } } if( inside_number ) { if( !(isdigit(c) || isspace(c) || c == INLINE_COMMENT_CHAR) ) { inside_number = FALSE; /* cannot be continued */ } if( makeupper(c) == 'H' ) { /* skip past hollerith */ Buf->end_index = i; /* the H might be last thing on line */ /* We can't just set i = i+num since line may end before then. */ i++; col++; /* move past 'H' */ for(hollerith_count = num_val; (c=s[i]) != '\0' && hollerith_count>0 && col<=max_stmt_col; /* Hmmm: how should tabs in holleriths be counted? As one column or many? We say many. */ hollerith_count--, i++, col = PORT_NXTCOL(c,col)) { Buf->end_index = i; } if( hollerith_count>0 ) { /* fill out implied blanks */ hollerith_count -= (max_stmt_col - col + 1); } } } /* end of dealing with number and maybe hollerith */ /* Start of a comment: quit i loop. */ if( c == INLINE_COMMENT_CHAR ) { Buf->f90_comment = TRUE; break; } } /* end of stuff outside quotes */ /* Re-test for end of stmt needed here due to advancing i above */ if( c == '\0' ) break; /* As long as interesting stuff is seen, * update end index. This should never * leave end_index undefined since we know line * is not a comment so not all blank. */ if( (inside_quote || ! isspace(c)) ) { if( col <= max_stmt_col ) { Buf->end_index = i; prev_sig_char = c; } /* Set up std violation warning if significant chars past std max line length. The second test is to avoid warning if nonsignificant blanks follow end of statement inside quote. */ if( col > std_max_stmt_col && (col <= max_stmt_col || ! isspace(c)) ) Buf->overlength = TRUE; } } /* end for i */ /* A continuation line that is blank except for the continuation mark will leave end_index unchanged from initial -1. This is not illegal in fixed form: f77 std 3.2.3. So we do not treat them as comments, unlike counterpart in freeform. */ if( Buf->contin && Buf->end_index == -1 ) { Buf->empty_contin = TRUE; } } /* end else not comment */ Buf = Buf->next; } /* end while Buf */ if(tab_count > 0 && tab_filename == NULL) tab_filename = current_filename; /* for portability warning */ }/*find_fixed_contins*/ /* Function to see if line is blank (under freeform rules) */ PRIVATE int line_is_blank(char *s) { int i, c; for(i=0; (c=s[i]) != '\0' && isspace(c); i++) continue; return (c == '\0'); }