Sweden-Number/tools/wrc/ppl.l

1457 lines
41 KiB
Plaintext

/*
* Wrc preprocessor lexical analysis
*
* Copyright 1999-2000 Bertho A. Stultiens (BS)
*
* 24-Apr-2000 BS - Started from scratch to restructure everything
* and reintegrate the source into the wine-tree.
* 04-Jan-2000 BS - Added comments about the lexicographical
* grammar to give some insight in the complexity.
* 28-Dec-1999 BS - Eliminated backing-up of the flexer by running
* `flex -b' on the source. This results in some
* weirdo extra rules, but a much faster scanner.
* 23-Dec-1999 BS - Started this file
*
*-------------------------------------------------------------------------
* The preprocessor's lexographical grammar (approximately):
*
* pp := {ws} # {ws} if {ws} {expr} {ws} \n
* | {ws} # {ws} ifdef {ws} {id} {ws} \n
* | {ws} # {ws} ifndef {ws} {id} {ws} \n
* | {ws} # {ws} elif {ws} {expr} {ws} \n
* | {ws} # {ws} else {ws} \n
* | {ws} # {ws} endif {ws} \n
* | {ws} # {ws} include {ws} < {anytext} > \n
* | {ws} # {ws} include {ws} " {anytext} " \n
* | {ws} # {ws} define {ws} {anytext} \n
* | {ws} # {ws} define( {arglist} ) {ws} {expansion} \n
* | {ws} # {ws} pragma {ws} {anytext} \n
* | {ws} # {ws} ident {ws} {anytext} \n
* | {ws} # {ws} error {ws} {anytext} \n
* | {ws} # {ws} warning {ws} {anytext} \n
* | {ws} # {ws} line {ws} " {anytext} " {number} \n
* | {ws} # {ws} {number} " {anytext} " {number} [{number} [{number}]] \n
* | {ws} # {ws} \n
*
* ws := [ \t\r\f\v]*
*
* expr := {expr} [+-*%^/|&] {expr}
* | {expr} {logor|logand} {expr}
* | [!~+-] {expr}
* | {expr} ? {expr} : {expr}
*
* logor := ||
*
* logand := &&
*
* id := [a-zA-Z_][a-zA-Z0-9_]*
*
* anytext := [^\n]* (see note)
*
* arglist :=
* | {id}
* | {arglist} , {id}
* | {arglist} , {id} ...
*
* expansion := {id}
* | # {id}
* | {anytext}
* | {anytext} ## {anytext}
*
* number := [0-9]+
*
* Note: "anytext" is not always "[^\n]*". This is because the
* trailing context must be considered as well.
*
* The only certain assumption for the preprocessor to make is that
* directives start at the beginning of the line, followed by a '#'
* and end with a newline.
* Any directive may be suffixed with a line-continuation. Also
* classical comment / *...* / (note: no comments within comments,
* therefore spaces) is considered to be a line-continuation
* (according to gcc and egcs AFAIK, ANSI is a bit vague).
* Comments have not been added to the above grammer for simplicity
* reasons. However, it is allowed to enter comment anywhere within
* the directives as long as they do not interfere with the context.
* All comments are considered to be deletable whitespace (both
* classical form "/ *...* /" and C++ form "//...\n").
*
* All recursive scans, except for macro-expansion, are done by the
* parser, whereas the simple state transitions of non-recursive
* directives are done in the scanner. This results in the many
* exclusive start-conditions of the scanner.
*
* Macro expansions are slightly more difficult because they have to
* prescan the arguments. Parameter substitution is literal if the
* substitution is # or ## (either side). This enables new identifiers
* to be created (see 'info cpp' node Macro|Pitfalls|Prescan for more
* information).
*
* FIXME: Variable macro parameters is recognized, but not yet
* expanded. I have to reread the ANSI standard on the subject (yes,
* ANSI defines it).
*
* The following special defines are supported:
* __FILE__ -> "thissource.c"
* __LINE__ -> 123
* __DATE__ -> "May 1 2000"
* __TIME__ -> "23:59:59"
* These macros expand, as expected, into their ANSI defined values.
*
* The same include prevention is implemented as gcc and egcs does.
* This results in faster processing because we do not read the text
* at all. Some wine-sources attempt to include the same file 4 or 5
* times. This strategy also saves a lot blank output-lines, which in
* its turn improves the real resource scanner/parser.
*
*/
/*
* Special flex options and exclusive scanner start-conditions
*/
%option stack
%option never-interactive
%x pp_pp
%x pp_eol
%x pp_inc
%x pp_dqs
%x pp_sqs
%x pp_iqs
%x pp_comment
%x pp_def
%x pp_define
%x pp_macro
%x pp_mbody
%x pp_macign
%x pp_macscan
%x pp_macexp
%x pp_if
%x pp_ifd
%x pp_line
%x pp_defined
%x pp_ignore
ws [ \v\f\t\r]
cident [a-zA-Z_][0-9a-zA-Z_]*
ul [uUlL]|[uUlL][lL]|[lL][uU]|[lL][lL][uU]|[uU][lL][lL]|[lL][uU][lL]
%{
#include "config.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#include <assert.h>
#include "utils.h"
#include "wrc.h"
#include "preproc.h"
#include "ppy.tab.h"
/*
* Make sure that we are running an appropriate version of flex.
*/
#if !defined(YY_FLEX_MAJOR_VERSION) || (1000 * YY_FLEX_MAJOR_VERSION + YY_FLEX_MINOR_VERSION < 2005)
#error Must use flex version 2.5.1 or higher (yy_scan_* routines are required).
#endif
#define YY_USE_PROTOS
#define YY_NO_UNPUT
#define YY_READ_BUF_SIZE 65536 /* So we read most of a file at once */
#define yy_current_state() YY_START
#define yy_pp_state(x) yy_pop_state(); yy_push_state(x)
/*
* Always update the current character position within a line
*/
#define YY_USER_ACTION char_number+=ppleng;
/*
* Buffer management for includes and expansions
*/
#define MAXBUFFERSTACK 128 /* Nesting more than 128 includes or macro expansion textss is insane */
typedef struct bufferstackentry {
YY_BUFFER_STATE bufferstate; /* Buffer to switch back to */
pp_entry_t *define; /* Points to expanding define or NULL if handling includes */
int line_number; /* Line that we were handling */
int char_number; /* The current position on that line */
char *filename; /* Filename that we were handling */
int if_depth; /* How many #if:s deep to check matching #endif:s */
int ncontinuations; /* Remember the continuation state */
int should_pop; /* Set if we must pop the start-state on EOF */
/* Include management */
int include_state;
char *include_ppp;
char *include_filename;
int include_ifdepth;
int seen_junk;
} bufferstackentry_t;
#define ALLOCBLOCKSIZE (1 << 10) /* Allocate these chunks at a time for string-buffers */
/*
* Macro expansion nesting
* We need the stack to handle expansions while scanning
* a macro's arguments. The TOS must always be the macro
* that receives the current expansion from the scanner.
*/
#define MAXMACEXPSTACK 128 /* Nesting more than 128 macro expansions is insane */
typedef struct macexpstackentry {
pp_entry_t *ppp; /* This macro we are scanning */
char **args; /* With these arguments */
char **ppargs; /* Resulting in these preprocessed arguments */
int *nnls; /* Number of newlines per argument */
int nargs; /* And this many arguments scanned */
int parentheses; /* Nesting level of () */
int curargsize; /* Current scanning argument's size */
int curargalloc; /* Current scanning argument's block allocated */
char *curarg; /* Current scanning argument's content */
} macexpstackentry_t;
#define MACROPARENTHESES() (top_macro()->parentheses)
/*
* Prototypes
*/
static void newline(int);
static int make_number(int radix, YYSTYPE *val, char *str, int len);
static void put_buffer(char *s, int len);
/* Buffer management */
static void push_buffer(pp_entry_t *ppp, char *filename, char *incname, int pop);
static bufferstackentry_t *pop_buffer(void);
/* String functions */
static void new_string(void);
static void add_string(char *str, int len);
static char *get_string(void);
static void put_string(void);
static int string_start(void);
/* Macro functions */
static void push_macro(pp_entry_t *ppp);
static macexpstackentry_t *top_macro(void);
static macexpstackentry_t *pop_macro(void);
static void free_macro(macexpstackentry_t *mep);
static void add_text_to_macro(char *text, int len);
static void macro_add_arg(int last);
static void macro_add_expansion(void);
/* Expansion */
static void expand_special(pp_entry_t *ppp);
static void expand_define(pp_entry_t *ppp);
static void expand_macro(macexpstackentry_t *mep);
/*
* Local variables
*/
static int ncontinuations;
static int strbuf_idx = 0;
static int strbuf_alloc = 0;
static char *strbuffer = NULL;
static int str_startline;
static macexpstackentry_t *macexpstack[MAXMACEXPSTACK];
static int macexpstackidx = 0;
static bufferstackentry_t bufferstack[MAXBUFFERSTACK];
static int bufferstackidx = 0;
/*
* Global variables
*/
/*
* Trace the include files to prevent double reading.
* This save 20..30% of processing time for most stuff
* that uses complex includes.
* States:
* -1 Don't track or seen junk
* 0 New include, waiting for "#ifndef __xxx_h"
* 1 Seen #ifndef, waiting for "#define __xxx_h ..."
* 2 Seen #endif, waiting for EOF
*/
int include_state = -1;
char *include_ppp = NULL; /* The define to be set from the #ifndef */
int include_ifdepth = 0; /* The level of ifs at the #ifdef */
int seen_junk = 0; /* Set when junk is seen */
includelogicentry_t *includelogiclist = NULL;
%}
/*
**************************************************************************
* The scanner starts here
**************************************************************************
*/
%%
/*
* Catch line-continuations.
* Note: Gcc keeps the line-continuations in, for example, strings
* intact. However, I prefer to remove them all so that the next
* scanner will not need to reduce the continuation state.
*
* <*>\\\n newline(0);
*/
/*
* Detect the leading # of a preprocessor directive.
*/
<INITIAL,pp_ignore>^{ws}*# seen_junk++; yy_push_state(pp_pp);
/*
* Scan for the preprocessor directives
*/
<pp_pp>{ws}*include{ws}* if(yy_top_state() != pp_ignore) {yy_pp_state(pp_inc); return tINCLUDE;} else {yy_pp_state(pp_eol);}
<pp_pp>{ws}*define{ws}* yy_pp_state(yy_current_state() != pp_ignore ? pp_def : pp_eol);
<pp_pp>{ws}*error{ws}* yy_pp_state(pp_eol); if(yy_top_state() != pp_ignore) return tERROR;
<pp_pp>{ws}*warning{ws}* yy_pp_state(pp_eol); if(yy_top_state() != pp_ignore) return tWARNING;
<pp_pp>{ws}*pragma{ws}* yy_pp_state(pp_eol); if(yy_top_state() != pp_ignore) return tPRAGMA;
<pp_pp>{ws}*ident{ws}* yy_pp_state(pp_eol); if(yy_top_state() != pp_ignore) return tPPIDENT;
<pp_pp>{ws}*undef{ws}* if(yy_top_state() != pp_ignore) {yy_pp_state(pp_ifd); return tUNDEF;} else {yy_pp_state(pp_eol);}
<pp_pp>{ws}*ifdef{ws}* yy_pp_state(pp_ifd); return tIFDEF;
<pp_pp>{ws}*ifndef{ws}* seen_junk--; yy_pp_state(pp_ifd); return tIFNDEF;
<pp_pp>{ws}*if{ws}* yy_pp_state(pp_if); return tIF;
<pp_pp>{ws}*elif{ws}* yy_pp_state(pp_if); return tELIF;
<pp_pp>{ws}*else{ws}* return tELSE;
<pp_pp>{ws}*endif{ws}* return tENDIF;
<pp_pp>{ws}*line{ws}* if(yy_top_state() != pp_ignore) {yy_pp_state(pp_line); return tLINE;} else {yy_pp_state(pp_eol);}
<pp_pp>{ws}+ if(yy_top_state() != pp_ignore) {yy_pp_state(pp_line); return tGCCLINE;} else {yy_pp_state(pp_eol);}
<pp_pp>{ws}*[a-z]+ pperror("Invalid preprocessor token '%s'", pptext);
<pp_pp>\r?\n newline(1); yy_pop_state(); return tNL; /* This could be the null-token */
<pp_pp>\\\r?\n newline(0);
<pp_pp>\\\r? pperror("Preprocessor junk '%s'", pptext);
<pp_pp>. return *pptext;
/*
* Handle #include and #line
*/
<pp_line>[0-9]+ return make_number(10, &pplval, pptext, ppleng);
<pp_inc>\< new_string(); add_string(pptext, ppleng); yy_push_state(pp_iqs);
<pp_inc,pp_line>\" new_string(); add_string(pptext, ppleng); yy_push_state(pp_dqs);
<pp_inc,pp_line>{ws}+ ;
<pp_inc,pp_line>\n newline(1); yy_pop_state(); return tNL;
<pp_inc,pp_line>\\\r?\n newline(0);
<pp_inc,pp_line>(\\\r?)|(.) pperror(yy_current_state() == pp_inc ? "Trailing junk in #include" : "Trailing junk in #line");
/*
* Ignore all input when a false clause is parsed
*/
<pp_ignore>[^#/\\\n]+ ;
<pp_ignore>\n newline(1);
<pp_ignore>\\\r?\n newline(0);
<pp_ignore>(\\\r?)|(.) ;
/*
* Handle #if and #elif.
* These require conditionals to be evaluated, but we do not
* want to jam the scanner normally when we see these tokens.
* Note: tIDENT is handled below.
*/
<pp_if>0[0-7]*{ul}? return make_number(8, &pplval, pptext, ppleng);
<pp_if>0[0-7]*[8-9]+{ul}? pperror("Invalid octal digit");
<pp_if>[1-9][0-9]*{ul}? return make_number(10, &pplval, pptext, ppleng);
<pp_if>0[xX][0-9a-fA-F]+{ul}? return make_number(16, &pplval, pptext, ppleng);
<pp_if>0[xX] pperror("Invalid hex number");
<pp_if>defined yy_push_state(pp_defined); return tDEFINED;
<pp_if>"<<" return tLSHIFT;
<pp_if>">>" return tRSHIFT;
<pp_if>"&&" return tLOGAND;
<pp_if>"||" return tLOGOR;
<pp_if>"==" return tEQ;
<pp_if>"!=" return tNE;
<pp_if>"<=" return tLTE;
<pp_if>">=" return tGTE;
<pp_if>\n newline(1); yy_pop_state(); return tNL;
<pp_if>\\\r?\n newline(0);
<pp_if>\\\r? pperror("Junk in conditional expression");
<pp_if>{ws}+ ;
<pp_if>\' new_string(); add_string(pptext, ppleng); yy_push_state(pp_sqs);
<pp_if>\" pperror("String constants not allowed in conditionals");
<pp_if>. return *pptext;
/*
* Handle #ifdef, #ifndef and #undef
* to get only an untranslated/unexpanded identifier
*/
<pp_ifd>{cident} pplval.cptr = xstrdup(pptext); return tIDENT;
<pp_ifd>{ws}+ ;
<pp_ifd>\n newline(1); yy_pop_state(); return tNL;
<pp_ifd>\\\r?\n newline(0);
<pp_ifd>(\\\r?)|(.) pperror("Identifier expected");
/*
* Handle the special 'defined' keyword.
* This is necessary to get the identifier prior to any
* substitutions.
*/
<pp_defined>{cident} yy_pop_state(); pplval.cptr = xstrdup(pptext); return tIDENT;
<pp_defined>{ws}+ ;
<pp_defined>(\()|(\)) return *pptext;
<pp_defined>\\\r?\n newline(0);
<pp_defined>(\\.)|(\n)|(.) pperror("Identifier expected");
/*
* Handle #error, #warning, #pragma and #ident.
* Pass everything literally to the parser, which
* will act appropriately.
* Comments are stripped from the literal text.
*/
<pp_eol>[^/\\\n]+ if(yy_top_state() != pp_ignore) { pplval.cptr = xstrdup(pptext); return tLITERAL; }
<pp_eol>\/[^/\\\n*]* if(yy_top_state() != pp_ignore) { pplval.cptr = xstrdup(pptext); return tLITERAL; }
<pp_eol>(\\\r?)|(\/[^/*]) if(yy_top_state() != pp_ignore) { pplval.cptr = xstrdup(pptext); return tLITERAL; }
<pp_eol>\n newline(1); yy_pop_state(); if(yy_current_state() != pp_ignore) { return tNL; }
<pp_eol>\\\r?\n newline(0);
/*
* Handle left side of #define
*/
<pp_def>{cident}\( pplval.cptr = xstrdup(pptext); pplval.cptr[ppleng-1] = '\0'; yy_pp_state(pp_macro); return tMACRO;
<pp_def>{cident} pplval.cptr = xstrdup(pptext); yy_pp_state(pp_define); return tDEFINE;
<pp_def>{ws}+ ;
<pp_def>\\\r?\n newline(0);
<pp_def>(\\\r?)|(\n)|(.) perror("Identifier expected");
/*
* Scan the substitution of a define
*/
<pp_define>[^'"/\\\n]+ pplval.cptr = xstrdup(pptext); return tLITERAL;
<pp_define>(\\\r?)|(\/[^/*]) pplval.cptr = xstrdup(pptext); return tLITERAL;
<pp_define>\\\r?\n{ws}+ newline(0); pplval.cptr = xstrdup(" "); return tLITERAL;
<pp_define>\\\r?\n newline(0);
<pp_define>\n newline(1); yy_pop_state(); return tNL;
<pp_define>\' new_string(); add_string(pptext, ppleng); yy_push_state(pp_sqs);
<pp_define>\" new_string(); add_string(pptext, ppleng); yy_push_state(pp_dqs);
/*
* Scan the definition macro arguments
*/
<pp_macro>\){ws}* yy_pp_state(pp_mbody); return tMACROEND;
<pp_macro>{ws}+ ;
<pp_macro>{cident} pplval.cptr = xstrdup(pptext); return tIDENT;
<pp_macro>, return ',';
<pp_macro>"..." return tELIPSIS;
<pp_macro>(\\\r?)|(\n)|(.)|(\.\.?) pperror("Argument identifier expected");
<pp_macro>\\\r?\n newline(0);
/*
* Scan the substitution of a macro
*/
<pp_mbody>[^a-zA-Z0-9'"#/\\\n]+ pplval.cptr = xstrdup(pptext); return tLITERAL;
<pp_mbody>{cident} pplval.cptr = xstrdup(pptext); return tIDENT;
<pp_mbody>\#\# return tCONCAT;
<pp_mbody>\# return tSTRINGIZE;
<pp_mbody>[0-9][^'"#/\\\n]* pplval.cptr = xstrdup(pptext); return tLITERAL;
<pp_mbody>(\\\r?)|(\/[^/*'"#\\\n]*) pplval.cptr = xstrdup(pptext); return tLITERAL;
<pp_mbody>\\\r?\n{ws}+ newline(0); pplval.cptr = xstrdup(" "); return tLITERAL;
<pp_mbody>\\\r?\n newline(0);
<pp_mbody>\n newline(1); yy_pop_state(); return tNL;
<pp_mbody>\' new_string(); add_string(pptext, ppleng); yy_push_state(pp_sqs);
<pp_mbody>\" new_string(); add_string(pptext, ppleng); yy_push_state(pp_dqs);
/*
* Macro expansion text scanning.
* This state is active just after the identifier is scanned
* that triggers an expansion. We *must* delete the leading
* whitespace before we can start scanning for arguments.
*
* If we do not see a '(' as next trailing token, then we have
* a false alarm. We just continue with a nose-bleed...
*/
<pp_macign>{ws}*/\( yy_pp_state(pp_macscan);
<pp_macign>{ws}*\n {
if(yy_top_state() != pp_macscan)
newline(0);
}
<pp_macign>{ws}*\\\r?\n newline(0);
<pp_macign>{ws}+|{ws}*\\\r?|. {
macexpstackentry_t *mac = pop_macro();
yy_pop_state();
put_buffer(mac->ppp->ident, strlen(mac->ppp->ident));
put_buffer(pptext, ppleng);
free_macro(mac);
}
/*
* Macro expansion argument text scanning.
* This state is active when a macro's arguments are being read for expansion.
*/
<pp_macscan>\( {
if(++MACROPARENTHESES() > 1)
add_text_to_macro(pptext, ppleng);
}
<pp_macscan>\) {
if(--MACROPARENTHESES() == 0)
{
yy_pop_state();
macro_add_arg(1);
}
else
add_text_to_macro(pptext, ppleng);
}
<pp_macscan>, {
if(MACROPARENTHESES() > 1)
add_text_to_macro(pptext, ppleng);
else
macro_add_arg(0);
}
<pp_macscan>\" new_string(); add_string(pptext, ppleng); yy_push_state(pp_dqs);
<pp_macscan>\' new_string(); add_string(pptext, ppleng); yy_push_state(pp_sqs);
<pp_macscan>"/*" yy_push_state(pp_comment); add_text_to_macro(" ", 1);
<pp_macscan>\n line_number++; char_number = 1; add_text_to_macro(pptext, ppleng);
<pp_macscan>([^/(),\\\n"']+)|(\/[^/*(),\\\n'"]*)|(\\\r?)|(.) add_text_to_macro(pptext, ppleng);
<pp_macscan>\\\r?\n newline(0);
/*
* Comment handling (almost all start-conditions)
*/
<INITIAL,pp_pp,pp_ignore,pp_eol,pp_inc,pp_if,pp_ifd,pp_defined,pp_def,pp_define,pp_macro,pp_mbody>"/*" yy_push_state(pp_comment);
<pp_comment>[^*\n]*|"*"+[^*/\n]* ;
<pp_comment>\n newline(0);
<pp_comment>"*"+"/" yy_pop_state();
/*
* Remove C++ style comment (almost all start-conditions)
*/
<INITIAL,pp_pp,pp_ignore,pp_eol,pp_inc,pp_if,pp_ifd,pp_defined,pp_def,pp_define,pp_macro,pp_mbody,pp_macscan>"//"[^\n]* {
if(pptext[ppleng-1] == '\\')
ppwarning("C++ style comment ends with an escaped newline (escape ignored)");
}
/*
* Single, double and <> quoted constants
*/
<INITIAL,pp_macexp>\" seen_junk++; new_string(); add_string(pptext, ppleng); yy_push_state(pp_dqs);
<INITIAL,pp_macexp>\' seen_junk++; new_string(); add_string(pptext, ppleng); yy_push_state(pp_sqs);
<pp_dqs>[^"\\\n]+ add_string(pptext, ppleng);
<pp_dqs>\" {
add_string(pptext, ppleng);
yy_pop_state();
switch(yy_current_state())
{
case pp_pp:
case pp_define:
case pp_mbody:
case pp_inc:
case pp_line:
pplval.cptr = get_string();
return tDQSTRING;
default:
put_string();
}
}
<pp_sqs>[^'\\\n]+ add_string(pptext, ppleng);
<pp_sqs>\' {
add_string(pptext, ppleng);
yy_pop_state();
switch(yy_current_state())
{
case pp_if:
case pp_define:
case pp_mbody:
pplval.cptr = get_string();
return tSQSTRING;
default:
put_string();
}
}
<pp_iqs>[^\>\\\n]+ add_string(pptext, ppleng);
<pp_iqs>\> {
add_string(pptext, ppleng);
yy_pop_state();
pplval.cptr = get_string();
return tIQSTRING;
}
<pp_dqs>\\\r?\n {
/*
* This is tricky; we need to remove the line-continuation
* from preprocessor strings, but OTOH retain them in all
* other strings. This is because the resource grammar is
* even more braindead than initially analysed and line-
* continuations in strings introduce, sigh, newlines in
* the output. There goes the concept of non-breaking, non-
* spacing whitespace.
*/
switch(yy_top_state())
{
case pp_pp:
case pp_define:
case pp_mbody:
case pp_inc:
case pp_line:
newline(0);
break;
default:
add_string(pptext, ppleng);
newline(-1);
}
}
<pp_iqs,pp_dqs,pp_sqs>\\. add_string(pptext, ppleng);
<pp_iqs,pp_dqs,pp_sqs>\n {
newline(1);
add_string(pptext, ppleng);
ppwarning("Newline in string constant encounterd (started line %d)", string_start());
}
/*
* Identifier scanning
*/
<INITIAL,pp_if,pp_inc,pp_macexp>{cident} {
pp_entry_t *ppp;
seen_junk++;
if(!(ppp = pplookup(pptext)))
{
if(yy_current_state() == pp_inc)
pperror("Expected include filename");
if(yy_current_state() == pp_if)
{
pplval.cptr = xstrdup(pptext);
return tIDENT;
}
else
put_buffer(pptext, ppleng);
}
else if(!ppp->expanding)
{
switch(ppp->type)
{
case def_special:
expand_special(ppp);
break;
case def_define:
expand_define(ppp);
break;
case def_macro:
yy_push_state(pp_macign);
push_macro(ppp);
break;
default:
internal_error(__FILE__, __LINE__, "Invalid define type %d\n", ppp->type);
}
}
}
/*
* Everything else that needs to be passed and
* newline and continuation handling
*/
<INITIAL,pp_macexp>[^a-zA-Z_#'"/\\\n \r\t\f\v]+|(\/|\\)[^a-zA-Z_/*'"\\\n \r\t\v\f]* seen_junk++; put_buffer(pptext, ppleng);
<INITIAL,pp_macexp>{ws}+ put_buffer(pptext, ppleng);
<INITIAL>\n newline(1);
<INITIAL>\\\r?\n newline(0);
<INITIAL>\\\r? seen_junk++; put_buffer(pptext, ppleng);
/*
* Special catcher for macro argmument expansion to prevent
* newlines to propagate to the output or admin.
*/
<pp_macexp>(\n)|(.)|(\\\r?(\n|.)) put_buffer(pptext, ppleng);
/*
* This is a 'catch-all' rule to discover errors in the scanner
* in an orderly manner.
*/
<*>. seen_junk++; ppwarning("Unmatched text '%c' (0x%02x); please report\n", isprint(*pptext & 0xff) ? *pptext : ' ', *pptext);
<<EOF>> {
YY_BUFFER_STATE b = YY_CURRENT_BUFFER;
bufferstackentry_t *bep = pop_buffer();
if((!bep && get_if_depth()) || (bep && get_if_depth() != bep->if_depth))
ppwarning("Unmatched #if/#endif at end of file");
if(!bep)
{
if(YY_START != INITIAL)
pperror("Unexpected end of file during preprocessing");
yyterminate();
}
else if(bep->should_pop == 2)
{
macexpstackentry_t *mac;
mac = pop_macro();
expand_macro(mac);
}
pp_delete_buffer(b);
}
%%
/*
**************************************************************************
* Support functions
**************************************************************************
*/
#ifndef ppwrap
int ppwrap(void)
{
return 1;
}
#endif
/*
*-------------------------------------------------------------------------
* Output newlines or set them as continuations
*
* Input: -1 - Don't count this one, but update local position (see pp_dqs)
* 0 - Line-continuation seen and cache output
* 1 - Newline seen and flush output
*-------------------------------------------------------------------------
*/
static void newline(int dowrite)
{
line_number++;
char_number = 1;
if(dowrite == -1)
return;
ncontinuations++;
if(dowrite)
{
for(;ncontinuations; ncontinuations--)
put_buffer("\n", 1);
}
}
/*
*-------------------------------------------------------------------------
* Make a number out of an any-base and suffixed string
*
* Possible number extensions:
* - "" int
* - "L" long int
* - "LL" long long int
* - "U" unsigned int
* - "UL" unsigned long int
* - "ULL" unsigned long long int
* - "LU" unsigned long int
* - "LLU" unsigned long long int
* - "LUL" invalid
*
* FIXME:
* The sizes of resulting 'int' and 'long' are compiler specific.
* I depend on sizeof(int) > 2 here (although a relatively safe
* assumption).
* Long longs are not yet implemented because this is very compiler
* specific and I don't want to think too much about the problems.
*
*-------------------------------------------------------------------------
*/
static int make_number(int radix, YYSTYPE *val, char *str, int len)
{
int is_l = 0;
int is_ll = 0;
int is_u = 0;
char ext[4];
ext[3] = '\0';
ext[2] = toupper(str[len-1]);
ext[1] = len > 1 ? toupper(str[len-2]) : ' ';
ext[0] = len > 2 ? toupper(str[len-3]) : ' ';
if(!strcmp(ext, "LUL"))
pperror("Invalid constant suffix");
else if(!strcmp(ext, "LLU") || !strcmp(ext, "ULL"))
{
is_ll++;
is_u++;
}
else if(!strcmp(ext+1, "LU") || !strcmp(ext+1, "UL"))
{
is_l++;
is_u++;
}
else if(!strcmp(ext+1, "LL"))
{
is_ll++;
}
else if(!strcmp(ext+2, "L"))
{
is_l++;
}
else if(!strcmp(ext+2, "U"))
{
is_u++;
}
if(is_ll)
internal_error(__FILE__, __LINE__, "long long constants not implemented yet");
if(is_u && is_l)
{
val->ulong = strtoul(str, NULL, radix);
return tULONG;
}
else if(!is_u && is_l)
{
val->slong = strtol(str, NULL, radix);
return tSLONG;
}
else if(is_u && !is_l)
{
val->uint = (unsigned int)strtoul(str, NULL, radix);
if(!win32 && val->uint > 65535)
{
pperror("Constant overflow");
}
return tUINT;
}
/* Else it must be an int... */
val->sint = (int)strtol(str, NULL, radix);
if(!win32 && (val->sint < -32768 || val->sint > 32768))
{
/*
* Note: test must be > 32768 because unary minus
* is handled as an expression! This can result in
* failure and must be checked in the parser.
*/
pperror("Constant overflow");
}
return tSINT;
}
/*
*-------------------------------------------------------------------------
* Macro and define expansion support
*
* FIXME: Variable macro arguments.
*-------------------------------------------------------------------------
*/
static void expand_special(pp_entry_t *ppp)
{
char *dbgtext = "?";
static char *buf = NULL;
assert(ppp->type == def_special);
if(!strcmp(ppp->ident, "__LINE__"))
{
dbgtext = "def_special(__LINE__)";
buf = xrealloc(buf, 32);
sprintf(buf, "%d", line_number);
}
else if(!strcmp(ppp->ident, "__FILE__"))
{
dbgtext = "def_special(__FILE__)";
buf = xrealloc(buf, strlen(input_name) + 3);
sprintf(buf, "\"%s\"", input_name);
}
else if(!strcmp(ppp->ident, "__DATE__"))
{
dbgtext = "def_special(__DATE__)";
buf = xrealloc(buf, 32);
strftime(buf, 32, "\"%b %d %Y\"", localtime(&now));
}
else if(!strcmp(ppp->ident, "__TIME__"))
{
dbgtext = "def_special(__TIME__)";
buf = xrealloc(buf, 32);
strftime(buf, 32, "\"%H:%M:%S\"", localtime(&now));
}
else
internal_error(__FILE__, __LINE__, "Special macro '%s' not found...\n", ppp->ident);
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "expand_special(%d): %s:%d: '%s' -> '%s'\n",
macexpstackidx,
input_name,
line_number,
ppp->ident,
buf ? buf : "");
if(buf && buf[0])
{
push_buffer(ppp, NULL, NULL, 0);
yy_scan_string(buf);
}
}
static void expand_define(pp_entry_t *ppp)
{
assert(ppp->type == def_define);
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "expand_define(%d): %s:%d: '%s' -> '%s'\n",
macexpstackidx,
input_name,
line_number,
ppp->ident,
ppp->subst.text);
if(ppp->subst.text && ppp->subst.text[0])
{
push_buffer(ppp, NULL, NULL, 0);
yy_scan_string(ppp->subst.text);
}
}
static int curdef_idx = 0;
static int curdef_alloc = 0;
static char *curdef_text = NULL;
static void add_text(char *str, int len)
{
if(len == 0)
return;
if(curdef_idx >= curdef_alloc || curdef_alloc - curdef_idx < len)
{
curdef_alloc += (len + ALLOCBLOCKSIZE-1) & ~(ALLOCBLOCKSIZE-1);
curdef_text = xrealloc(curdef_text, curdef_alloc * sizeof(curdef_text[0]));
if(curdef_alloc > 65536)
ppwarning("Reallocating macro-expansion buffer larger than 64kB");
}
memcpy(&curdef_text[curdef_idx], str, len);
curdef_idx += len;
}
static mtext_t *add_expand_text(mtext_t *mtp, macexpstackentry_t *mep, int *nnl)
{
char *cptr;
char *exp;
int tag;
int n;
if(mtp == NULL)
return NULL;
switch(mtp->type)
{
case exp_text:
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "add_expand_text: exp_text: '%s'\n", mtp->subst.text);
add_text(mtp->subst.text, strlen(mtp->subst.text));
break;
case exp_stringize:
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "add_expand_text: exp_stringize(%d): '%s'\n",
mtp->subst.argidx,
mep->args[mtp->subst.argidx]);
cptr = mep->args[mtp->subst.argidx];
add_text("\"", 1);
while(*cptr)
{
if(*cptr == '"' || *cptr == '\\')
add_text("\\", 1);
add_text(cptr, 1);
cptr++;
}
add_text("\"", 1);
break;
case exp_concat:
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "add_expand_text: exp_concat\n");
/* Remove trailing whitespace from current expansion text */
while(curdef_idx)
{
if(isspace(curdef_text[curdef_idx-1] & 0xff))
curdef_idx--;
else
break;
}
/* tag current position and recursively expand the next part */
tag = curdef_idx;
mtp = add_expand_text(mtp->next, mep, nnl);
/* Now get rid of the leading space of the expansion */
cptr = &curdef_text[tag];
n = curdef_idx - tag;
while(n)
{
if(isspace(*cptr & 0xff))
{
cptr++;
n--;
}
else
break;
}
if(cptr != &curdef_text[tag])
{
memmove(&curdef_text[tag], cptr, n);
curdef_idx -= (curdef_idx - tag) - n;
}
break;
case exp_subst:
if((mtp->next && mtp->next->type == exp_concat) || (mtp->prev && mtp->prev->type == exp_concat))
exp = mep->args[mtp->subst.argidx];
else
exp = mep->ppargs[mtp->subst.argidx];
if(exp)
{
add_text(exp, strlen(exp));
*nnl -= mep->nnls[mtp->subst.argidx];
cptr = strchr(exp, '\n');
while(cptr)
{
*cptr = ' ';
cptr = strchr(cptr+1, '\n');
}
mep->nnls[mtp->subst.argidx] = 0;
}
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "add_expand_text: exp_subst(%d): '%s'\n", mtp->subst.argidx, exp);
break;
default:
internal_error(__FILE__, __LINE__, "Invalid expansion type (%d) in macro expansion\n", mtp->type);
}
return mtp;
}
static void expand_macro(macexpstackentry_t *mep)
{
mtext_t *mtp;
int n, k;
char *cptr;
int nnl = 0;
pp_entry_t *ppp = mep->ppp;
int nargs = mep->nargs;
assert(ppp->type == def_macro);
assert(ppp->expanding == 0);
if((ppp->nargs >= 0 && nargs != ppp->nargs) || (ppp->nargs < 0 && nargs < -ppp->nargs))
pperror("Too %s macro arguments (%d)", nargs < abs(ppp->nargs) ? "few" : "many", nargs);
for(n = 0; n < nargs; n++)
nnl += mep->nnls[n];
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "expand_macro(%d): %s:%d: '%s'(%d,%d) -> ...\n",
macexpstackidx,
input_name,
line_number,
ppp->ident,
mep->nargs,
nnl);
curdef_idx = 0;
for(mtp = ppp->subst.mtext; mtp; mtp = mtp->next)
{
if(!(mtp = add_expand_text(mtp, mep, &nnl)))
break;
}
for(n = 0; n < nnl; n++)
add_text("\n", 1);
/* To make sure there is room and termination (see below) */
add_text(" \0", 2);
/* Strip trailing whitespace from expansion */
for(k = curdef_idx, cptr = &curdef_text[curdef_idx-1]; k > 0; k--, cptr--)
{
if(!isspace(*cptr & 0xff))
break;
}
/*
* We must add *one* whitespace to make sure that there
* is a token-seperation after the expansion.
*/
*(++cptr) = ' ';
*(++cptr) = '\0';
k++;
/* Strip leading whitespace from expansion */
for(n = 0, cptr = curdef_text; n < k; n++, cptr++)
{
if(!isspace(*cptr & 0xff))
break;
}
if(k - n > 0)
{
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "expand_text: '%s'\n", curdef_text + n);
push_buffer(ppp, NULL, NULL, 0);
/*yy_scan_bytes(curdef_text + n, k - n);*/
yy_scan_string(curdef_text + n);
}
}
/*
*-------------------------------------------------------------------------
* String collection routines
*-------------------------------------------------------------------------
*/
static void new_string(void)
{
#ifdef DEBUG
if(strbuf_idx)
ppwarning("new_string: strbuf_idx != 0");
#endif
strbuf_idx = 0;
str_startline = line_number;
}
static void add_string(char *str, int len)
{
if(len == 0)
return;
if(strbuf_idx >= strbuf_alloc || strbuf_alloc - strbuf_idx < len)
{
strbuf_alloc += (len + ALLOCBLOCKSIZE-1) & ~(ALLOCBLOCKSIZE-1);
strbuffer = xrealloc(strbuffer, strbuf_alloc * sizeof(strbuffer[0]));
if(strbuf_alloc > 65536)
ppwarning("Reallocating string buffer larger than 64kB");
}
memcpy(&strbuffer[strbuf_idx], str, len);
strbuf_idx += len;
}
static char *get_string(void)
{
char *str = (char *)xmalloc(strbuf_idx + 1);
memcpy(str, strbuffer, strbuf_idx);
str[strbuf_idx] = '\0';
#ifdef DEBUG
strbuf_idx = 0;
#endif
return str;
}
static void put_string(void)
{
put_buffer(strbuffer, strbuf_idx);
#ifdef DEBUG
strbuf_idx = 0;
#endif
}
static int string_start(void)
{
return str_startline;
}
/*
*-------------------------------------------------------------------------
* Buffer management
*-------------------------------------------------------------------------
*/
static void push_buffer(pp_entry_t *ppp, char *filename, char *incname, int pop)
{
if(ppdebug)
printf("push_buffer(%d): %p %p %p %d\n", bufferstackidx, ppp, filename, incname, pop);
if(bufferstackidx >= MAXBUFFERSTACK)
internal_error(__FILE__, __LINE__, "Buffer stack overflow");
memset(&bufferstack[bufferstackidx], 0, sizeof(bufferstack[0]));
bufferstack[bufferstackidx].bufferstate = YY_CURRENT_BUFFER;
bufferstack[bufferstackidx].define = ppp;
bufferstack[bufferstackidx].line_number = line_number;
bufferstack[bufferstackidx].char_number = char_number;
bufferstack[bufferstackidx].if_depth = get_if_depth();
bufferstack[bufferstackidx].should_pop = pop;
bufferstack[bufferstackidx].filename = input_name;
bufferstack[bufferstackidx].ncontinuations = ncontinuations;
bufferstack[bufferstackidx].include_state = include_state;
bufferstack[bufferstackidx].include_ppp = include_ppp;
bufferstack[bufferstackidx].include_filename = incname;
bufferstack[bufferstackidx].include_ifdepth = include_ifdepth;
bufferstack[bufferstackidx].seen_junk = seen_junk;
if(ppp)
ppp->expanding = 1;
else if(filename)
{
/* These will track the pperror to the correct file and line */
line_number = 1;
char_number = 1;
input_name = filename;
ncontinuations = 0;
}
else if(!pop)
internal_error(__FILE__, __LINE__, "Pushing buffer without knowing where to go to");
bufferstackidx++;
}
static bufferstackentry_t *pop_buffer(void)
{
if(bufferstackidx < 0)
internal_error(__FILE__, __LINE__, "Bufferstack underflow?");
if(bufferstackidx == 0)
return NULL;
bufferstackidx--;
if(bufferstack[bufferstackidx].define)
bufferstack[bufferstackidx].define->expanding = 0;
else
{
line_number = bufferstack[bufferstackidx].line_number;
char_number = bufferstack[bufferstackidx].char_number;
input_name = bufferstack[bufferstackidx].filename;
ncontinuations = bufferstack[bufferstackidx].ncontinuations;
if(!bufferstack[bufferstackidx].should_pop)
{
fclose(ppin);
fprintf(ppout, "# %d \"%s\" 2\n", line_number, input_name);
/* We have EOF, check the include logic */
if(include_state == 2 && !seen_junk && include_ppp)
{
pp_entry_t *ppp = pplookup(include_ppp);
if(ppp)
{
includelogicentry_t *iep = xmalloc(sizeof(includelogicentry_t));
iep->ppp = ppp;
ppp->iep = iep;
iep->filename = bufferstack[bufferstackidx].include_filename;
iep->next = includelogiclist;
if(iep->next)
iep->next->prev = iep;
includelogiclist = iep;
if(debuglevel & DEBUGLEVEL_PPMSG)
fprintf(stderr, "pop_buffer: %s:%d: includelogic added, include_ppp='%s', file='%s'\n", input_name, line_number, include_ppp, iep->filename);
}
else if(bufferstack[bufferstackidx].include_filename)
free(bufferstack[bufferstackidx].include_filename);
}
if(include_ppp)
free(include_ppp);
include_state = bufferstack[bufferstackidx].include_state;
include_ppp = bufferstack[bufferstackidx].include_ppp;
include_ifdepth = bufferstack[bufferstackidx].include_ifdepth;
seen_junk = bufferstack[bufferstackidx].seen_junk;
}
}
if(ppdebug)
printf("pop_buffer(%d): %p %p (%d, %d, %d) %p %d\n",
bufferstackidx,
bufferstack[bufferstackidx].bufferstate,
bufferstack[bufferstackidx].define,
bufferstack[bufferstackidx].line_number,
bufferstack[bufferstackidx].char_number,
bufferstack[bufferstackidx].if_depth,
bufferstack[bufferstackidx].filename,
bufferstack[bufferstackidx].should_pop);
pp_switch_to_buffer(bufferstack[bufferstackidx].bufferstate);
if(bufferstack[bufferstackidx].should_pop)
{
if(yy_current_state() == pp_macexp)
macro_add_expansion();
else
internal_error(__FILE__, __LINE__, "Pop buffer and state without macro expansion state");
yy_pop_state();
}
return &bufferstack[bufferstackidx];
}
/*
*-------------------------------------------------------------------------
* Macro nestng support
*-------------------------------------------------------------------------
*/
static void push_macro(pp_entry_t *ppp)
{
if(macexpstackidx >= MAXMACEXPSTACK)
pperror("Too many nested macros");
macexpstack[macexpstackidx] = xmalloc(sizeof(macexpstack[0][0]));
macexpstack[macexpstackidx]->ppp = ppp;
macexpstackidx++;
}
static macexpstackentry_t *top_macro(void)
{
return macexpstackidx > 0 ? macexpstack[macexpstackidx-1] : NULL;
}
static macexpstackentry_t *pop_macro(void)
{
if(macexpstackidx <= 0)
internal_error(__FILE__, __LINE__, "Macro expansion stack underflow\n");
return macexpstack[--macexpstackidx];
}
static void free_macro(macexpstackentry_t *mep)
{
int i;
for(i = 0; i < mep->nargs; i++)
free(mep->args[i]);
if(mep->args)
free(mep->args);
if(mep->nnls)
free(mep->nnls);
if(mep->curarg)
free(mep->curarg);
free(mep);
}
static void add_text_to_macro(char *text, int len)
{
macexpstackentry_t *mep = top_macro();
assert(mep->ppp->expanding == 0);
if(mep->curargalloc - mep->curargsize <= len)
{
mep->curargalloc += ALLOCBLOCKSIZE;
mep->curarg = xrealloc(mep->curarg, mep->curargalloc * sizeof(mep->curarg[0]));
}
memcpy(mep->curarg + mep->curargsize, text, len+1); /* +1 for '\0' */
mep->curargsize += len;
}
static void macro_add_arg(int last)
{
int nnl = 0;
char *cptr;
macexpstackentry_t *mep = top_macro();
assert(mep->ppp->expanding == 0);
mep->args = xrealloc(mep->args, (mep->nargs+1) * sizeof(mep->args[0]));
mep->ppargs = xrealloc(mep->ppargs, (mep->nargs+1) * sizeof(mep->ppargs[0]));
mep->nnls = xrealloc(mep->nnls, (mep->nargs+1) * sizeof(mep->nnls[0]));
mep->args[mep->nargs] = xstrdup(mep->curarg ? mep->curarg : "");
cptr = mep->args[mep->nargs]-1;
while((cptr = strchr(cptr+1, '\n')))
{
nnl++;
}
mep->nnls[mep->nargs] = nnl;
mep->nargs++;
free(mep->curarg);
mep->curargalloc = mep->curargsize = 0;
mep->curarg = NULL;
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "macro_add_arg: %s:%d: %d -> '%s'\n",
input_name,
line_number,
mep->nargs-1,
mep->args[mep->nargs-1]);
/* Each macro argument must be expanded to cope with stingize */
if(last || mep->args[mep->nargs-1][0])
{
yy_push_state(pp_macexp);
push_buffer(NULL, NULL, NULL, last ? 2 : 1);
yy_scan_string(mep->args[mep->nargs-1]);
/*mep->bufferstackidx = bufferstackidx; But not nested! */
}
}
static void macro_add_expansion(void)
{
macexpstackentry_t *mep = top_macro();
assert(mep->ppp->expanding == 0);
mep->ppargs[mep->nargs-1] = xstrdup(mep->curarg ? mep->curarg : "");
free(mep->curarg);
mep->curargalloc = mep->curargsize = 0;
mep->curarg = NULL;
if(debuglevel & DEBUGLEVEL_PPLEX)
fprintf(stderr, "macro_add_expansion: %s:%d: %d -> '%s'\n",
input_name,
line_number,
mep->nargs-1,
mep->ppargs[mep->nargs-1]);
}
/*
*-------------------------------------------------------------------------
* Output management
*-------------------------------------------------------------------------
*/
static void put_buffer(char *s, int len)
{
if(top_macro())
add_text_to_macro(s, len);
else
fwrite(s, 1, len, ppout);
}
/*
*-------------------------------------------------------------------------
* Include management
*-------------------------------------------------------------------------
*/
void do_include(char *fname, int type)
{
char *newpath;
int n;
includelogicentry_t *iep;
for(iep = includelogiclist; iep; iep = iep->next)
{
if(!strcmp(iep->filename, fname))
{
/*
* We are done. The file was included before.
* If the define was deleted, then this entry would have
* been deleted too.
*/
return;
}
}
n = strlen(fname);
if(n <= 2)
pperror("Empty include filename");
/* Undo the effect of the quotation */
fname[n-1] = '\0';
if((ppin = open_include(fname+1, type, &newpath)) == NULL)
pperror("Unable to open include file %s", fname+1);
fname[n-1] = *fname; /* Redo the quotes */
push_buffer(NULL, newpath, fname, 0);
seen_junk = 0;
include_state = 0;
include_ppp = NULL;
if(debuglevel & DEBUGLEVEL_PPMSG)
fprintf(stderr, "do_include: %s:%d: include_state=%d, include_ppp='%s', include_ifdepth=%d\n", input_name, line_number, include_state, include_ppp, include_ifdepth);
pp_switch_to_buffer(pp_create_buffer(ppin, YY_BUF_SIZE));
fprintf(ppout, "# 1 \"%s\" 1%s\n", newpath, type ? "" : " 3");
}
/*
*-------------------------------------------------------------------------
* Push/pop preprocessor ignore state when processing conditionals
* which are false.
*-------------------------------------------------------------------------
*/
void push_ignore_state(void)
{
yy_push_state(pp_ignore);
}
void pop_ignore_state(void)
{
yy_pop_state();
}