bc: parse file arguments piecemeal (do not read entire file)

function                                             old     new   delta
bc_read_line                                         336     406     +70
zbc_vm_execute_FILE                                    -      67     +67
zbc_lex_next                                        2309    2318      +9
zbc_program_exec                                    4002    4008      +6
bc_program_index                                      66      64      -2
bc_vm_run                                            139     124     -15
zbc_vm_file                                          208      32    -176
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 3/3 up/down: 152/-193)          Total: -41 bytes
   text	   data	    bss	    dec	    hex	filename
 981736	    485	   7296	 989517	  f194d	busybox_old
 981667	    485	   7296	 989448	  f1908	busybox_unstripped

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This commit is contained in:
Denys Vlasenko 2018-12-17 09:51:43 +01:00
parent c5774a3458
commit e4ba4c4371

View File

@ -694,7 +694,6 @@ struct globals {
IF_FEATURE_BC_SIGNALS(smallint ttyin;) IF_FEATURE_BC_SIGNALS(smallint ttyin;)
IF_FEATURE_CLEAN_UP(smallint exiting;) IF_FEATURE_CLEAN_UP(smallint exiting;)
smallint in_read; smallint in_read;
smallint use_stdin;
BcParse prs; BcParse prs;
BcProgram prog; BcProgram prog;
@ -704,7 +703,8 @@ struct globals {
unsigned err_line; unsigned err_line;
BcVec files; BcVec files;
BcVec stdin_buffer; BcVec input_buffer;
FILE *input_fp;
char *env_args; char *env_args;
@ -1317,8 +1317,8 @@ static int bad_input_byte(char c)
return 0; return 0;
} }
// Note: it _appends_ data from the stdin to vec. // Note: it _appends_ data from fp to vec.
static void bc_read_line(BcVec *vec) static void bc_read_line(BcVec *vec, FILE *fp)
{ {
again: again:
fflush_and_check(); fflush_and_check();
@ -1326,6 +1326,17 @@ static void bc_read_line(BcVec *vec)
#if ENABLE_FEATURE_BC_SIGNALS #if ENABLE_FEATURE_BC_SIGNALS
if (G_interrupt) { // ^C was pressed if (G_interrupt) { // ^C was pressed
intr: intr:
if (fp != stdin) {
// ^C while running a script (bc SCRIPT): die.
// We do not return to interactive prompt:
// user might be running us from a shell,
// and SCRIPT might be intended to terminate
// (e.g. contain a "halt" stmt).
// ^C dropping user into a bc prompt instead of
// the shell would be unexpected.
xfunc_die();
}
// ^C while interactive input
G_interrupt = 0; G_interrupt = 0;
// GNU bc says "interrupted execution." // GNU bc says "interrupted execution."
// GNU dc says "Interrupt!" // GNU dc says "Interrupt!"
@ -1333,14 +1344,14 @@ static void bc_read_line(BcVec *vec)
} }
# if ENABLE_FEATURE_EDITING # if ENABLE_FEATURE_EDITING
if (G_ttyin) { if (G_ttyin && fp == stdin) {
int n, i; int n, i;
# define line_buf bb_common_bufsiz1 # define line_buf bb_common_bufsiz1
n = read_line_input(G.line_input_state, "", line_buf, COMMON_BUFSIZE); n = read_line_input(G.line_input_state, "", line_buf, COMMON_BUFSIZE);
if (n <= 0) { // read errors or EOF, or ^D, or ^C if (n <= 0) { // read errors or EOF, or ^D, or ^C
if (n == 0) // ^C if (n == 0) // ^C
goto intr; goto intr;
bc_vec_pushZeroByte(vec); bc_vec_pushZeroByte(vec); // ^D or EOF (or error)
return; return;
} }
i = 0; i = 0;
@ -1359,60 +1370,38 @@ static void bc_read_line(BcVec *vec)
bool bad_chars = 0; bool bad_chars = 0;
size_t len = vec->len; size_t len = vec->len;
IF_FEATURE_BC_SIGNALS(errno = 0;)
do { do {
c = fgetc(stdin); #if ENABLE_FEATURE_BC_SIGNALS
#if ENABLE_FEATURE_BC_SIGNALS && !ENABLE_FEATURE_EDITING if (G_interrupt) {
// Both conditions appear simultaneously, check both just in case // ^C was pressed: ignore entire line, get another one
if (errno == EINTR || G_interrupt) { vec->len = len;
// ^C was pressed
clearerr(stdin);
goto intr; goto intr;
} }
#endif #endif
c = fgetc(fp);
if (c == EOF) { if (c == EOF) {
if (ferror(stdin)) if (ferror(fp))
quit(); // this emits error message bb_perror_msg_and_die("input error");
// Note: EOF does not append '\n', therefore: // Note: EOF does not append '\n'
// printf 'print 123\n' | bc - works
// printf 'print 123' | bc - fails (syntax error)
break; break;
} }
bad_chars |= bad_input_byte(c); bad_chars |= bad_input_byte(c);
bc_vec_pushByte(vec, (char)c); bc_vec_pushByte(vec, (char)c);
} while (c != '\n'); } while (c != '\n');
if (bad_chars) { if (bad_chars) {
// Bad chars on this line, ignore entire line // Bad chars on this line
if (!G.prog.file) { // stdin
// ignore entire line, get another one
vec->len = len; vec->len = len;
goto again; goto again;
} }
bb_perror_msg_and_die("file '%s' is not text", G.prog.file);
}
bc_vec_pushZeroByte(vec); bc_vec_pushZeroByte(vec);
} }
} }
static char* bc_read_file(const char *path)
{
char *buf;
size_t size = ((size_t) -1);
size_t i;
// Never returns NULL (dies on errors)
buf = xmalloc_xopen_read_close(path, &size);
for (i = 0; i < size; ++i) {
char c = buf[i];
if ((c < ' ' && c != '\t' && c != '\r' && c != '\n') // also allow '\v' '\f'?
|| c > 0x7e
) {
free(buf);
buf = NULL;
break;
}
}
return buf;
}
static void bc_num_setToZero(BcNum *n, size_t scale) static void bc_num_setToZero(BcNum *n, size_t scale)
{ {
n->len = 0; n->len = 0;
@ -2912,7 +2901,7 @@ static bool bc_lex_more_input(BcLex *l)
size_t str; size_t str;
bool comment; bool comment;
bc_vec_pop_all(&G.stdin_buffer); bc_vec_pop_all(&G.input_buffer);
// This loop is complex because the vm tries not to send any lines that end // This loop is complex because the vm tries not to send any lines that end
// with a backslash to the parser. The reason for that is because the parser // with a backslash to the parser. The reason for that is because the parser
@ -2921,18 +2910,18 @@ static bool bc_lex_more_input(BcLex *l)
comment = false; comment = false;
str = 0; str = 0;
for (;;) { for (;;) {
size_t prevlen = G.stdin_buffer.len; size_t prevlen = G.input_buffer.len;
char *string; char *string;
bc_read_line(&G.stdin_buffer); bc_read_line(&G.input_buffer, G.input_fp);
// No more input means EOF // No more input means EOF
if (G.stdin_buffer.len <= prevlen + 1) // (we expect +1 for NUL byte) if (G.input_buffer.len <= prevlen + 1) // (we expect +1 for NUL byte)
break; break;
string = G.stdin_buffer.v + prevlen; string = G.input_buffer.v + prevlen;
while (*string) { while (*string) {
char c = *string; char c = *string;
if (string == G.stdin_buffer.v || string[-1] != '\\') { if (string == G.input_buffer.v || string[-1] != '\\') {
if (IS_BC) if (IS_BC)
str ^= (c == '"'); str ^= (c == '"');
else { else {
@ -2954,7 +2943,7 @@ static bool bc_lex_more_input(BcLex *l)
} }
} }
if (str != 0 || comment) { if (str != 0 || comment) {
G.stdin_buffer.len--; // backstep over the trailing NUL byte G.input_buffer.len--; // backstep over the trailing NUL byte
continue; continue;
} }
@ -2963,21 +2952,20 @@ static bool bc_lex_more_input(BcLex *l)
// if it is not, then it's EOF, and looping back // if it is not, then it's EOF, and looping back
// to bc_read_line() will detect it: // to bc_read_line() will detect it:
string -= 2; string -= 2;
if (string >= G.stdin_buffer.v && *string == '\\') { if (string >= G.input_buffer.v && *string == '\\') {
G.stdin_buffer.len--; G.input_buffer.len--;
continue; continue;
} }
break; break;
} }
l->buf = G.stdin_buffer.v; l->buf = G.input_buffer.v;
l->i = 0; l->i = 0;
//bb_error_msg("G.stdin_buffer.len:%d '%s'", G.stdin_buffer.len, G.stdin_buffer.v); // bb_error_msg("G.input_buffer.len:%d '%s'", G.input_buffer.len, G.input_buffer.v);
l->len = G.stdin_buffer.len - 1; // do not include NUL l->len = G.input_buffer.len - 1; // do not include NUL
G.use_stdin = (l->len != 0); return l->len != 0;
return G.use_stdin;
} }
static BC_STATUS zbc_lex_next(BcLex *l) static BC_STATUS zbc_lex_next(BcLex *l)
@ -2989,22 +2977,23 @@ static BC_STATUS zbc_lex_next(BcLex *l)
l->line += l->newline; l->line += l->newline;
G.err_line = l->line; G.err_line = l->line;
l->t.t = BC_LEX_EOF;
//this NL handling is bogus
l->newline = (l->i == l->len);
if (l->newline) {
if (!G.use_stdin || !bc_lex_more_input(l))
RETURN_STATUS(BC_STATUS_SUCCESS);
// here it's guaranteed that l->i is below l->len
l->newline = false; l->newline = false;
}
// Loop until failure or we don't have whitespace. This // Loop until failure or we don't have whitespace. This
// is so the parser doesn't get inundated with whitespace. // is so the parser doesn't get inundated with whitespace.
// Comments are also BC_LEX_WHITESPACE tokens and eaten here. // Comments are also BC_LEX_WHITESPACE tokens and eaten here.
s = BC_STATUS_SUCCESS; s = BC_STATUS_SUCCESS;
do { do {
l->t.t = BC_LEX_EOF;
if (l->i == l->len) {
if (!G.input_fp)
RETURN_STATUS(BC_STATUS_SUCCESS);
if (!bc_lex_more_input(l)) {
G.input_fp = NULL;
RETURN_STATUS(BC_STATUS_SUCCESS);
}
// here it's guaranteed that l->i is below l->len
}
dbg_lex("next string to parse:'%.*s'", dbg_lex("next string to parse:'%.*s'",
(int)(strchrnul(l->buf + l->i, '\n') - (l->buf + l->i)), (int)(strchrnul(l->buf + l->i, '\n') - (l->buf + l->i)),
l->buf + l->i); l->buf + l->i);
@ -5351,7 +5340,7 @@ static BC_STATUS zbc_program_read(void)
G.in_read = 1; G.in_read = 1;
bc_char_vec_init(&buf); bc_char_vec_init(&buf);
bc_read_line(&buf); bc_read_line(&buf, stdin);
bc_parse_create(&parse, BC_PROG_READ); bc_parse_create(&parse, BC_PROG_READ);
bc_lex_file(&parse.l); bc_lex_file(&parse.l);
@ -6931,62 +6920,46 @@ static BC_STATUS zbc_vm_process(const char *text)
# define zbc_vm_process(...) (zbc_vm_process(__VA_ARGS__), BC_STATUS_SUCCESS) # define zbc_vm_process(...) (zbc_vm_process(__VA_ARGS__), BC_STATUS_SUCCESS)
#endif #endif
static BC_STATUS zbc_vm_file(const char *file) static BC_STATUS zbc_vm_execute_FILE(FILE *fp, const char *filename)
{ {
// So far bc/dc have no way to include a file from another file, // So far bc/dc have no way to include a file from another file,
// therefore we know G.prog.file == NULL on entry // therefore we know G.prog.file == NULL on entry
//const char *sv_file; //const char *sv_file;
char *data;
BcStatus s; BcStatus s;
BcFunc *main_func;
BcInstPtr *ip;
data = bc_read_file(file); G.prog.file = filename;
if (!data) RETURN_STATUS(bc_error_fmt("file '%s' is not text", file)); G.input_fp = fp;
//sv_file = G.prog.file;
G.prog.file = file;
bc_lex_file(&G.prs.l); bc_lex_file(&G.prs.l);
s = zbc_vm_process(data);
if (s) goto err;
main_func = bc_program_func(BC_PROG_MAIN); do {
ip = bc_vec_item(&G.prog.stack, 0); s = zbc_vm_process("");
// We do not stop looping on errors here if reading stdin.
if (main_func->code.len < ip->idx) // Example: start interactive bc and enter "return".
s = bc_error_fmt("file '%s' is not executable", file); // It should say "'return' not in a function"
// but should not exit.
err: } while (G.input_fp == stdin);
//G.prog.file = sv_file;
G.prog.file = NULL; G.prog.file = NULL;
free(data); RETURN_STATUS(s);
}
#if ERRORS_ARE_FATAL
# define zbc_vm_execute_FILE(...) (zbc_vm_execute_FILE(__VA_ARGS__), BC_STATUS_SUCCESS)
#endif
static BC_STATUS zbc_vm_file(const char *file)
{
BcStatus s;
FILE *fp;
fp = xfopen_for_read(file);
s = zbc_vm_execute_FILE(fp, file);
fclose(fp);
RETURN_STATUS(s); RETURN_STATUS(s);
} }
#if ERRORS_ARE_FATAL #if ERRORS_ARE_FATAL
# define zbc_vm_file(...) (zbc_vm_file(__VA_ARGS__), BC_STATUS_SUCCESS) # define zbc_vm_file(...) (zbc_vm_file(__VA_ARGS__), BC_STATUS_SUCCESS)
#endif #endif
static BC_STATUS zbc_vm_stdin(void)
{
BcStatus s;
//G.prog.file = NULL; - already is
bc_lex_file(&G.prs.l);
G.use_stdin = 1;
do {
s = zbc_vm_process("");
// We do not stop looping on errors here.
// Example: start interactive bc and enter "return".
// It should say "'return' not in a function"
// but should not exit.
} while (G.use_stdin);
RETURN_STATUS(s);
}
#if ERRORS_ARE_FATAL
# define zbc_vm_stdin(...) (zbc_vm_stdin(__VA_ARGS__), BC_STATUS_SUCCESS)
#endif
#if ENABLE_BC #if ENABLE_BC
static void bc_vm_info(void) static void bc_vm_info(void)
{ {
@ -7257,7 +7230,7 @@ static BC_STATUS zbc_vm_exec(void)
} }
if (IS_BC || (option_mask32 & BC_FLAG_I)) if (IS_BC || (option_mask32 & BC_FLAG_I))
s = zbc_vm_stdin(); s = zbc_vm_execute_FILE(stdin, /*filename:*/ NULL);
RETURN_STATUS(s); RETURN_STATUS(s);
} }
@ -7287,7 +7260,7 @@ static void bc_program_free(void)
bc_num_free(&G.prog.last); bc_num_free(&G.prog.last);
bc_num_free(&G.prog.zero); bc_num_free(&G.prog.zero);
bc_num_free(&G.prog.one); bc_num_free(&G.prog.one);
bc_vec_free(&G.stdin_buffer); bc_vec_free(&G.input_buffer);
} }
static void bc_vm_free(void) static void bc_vm_free(void)
@ -7352,7 +7325,7 @@ static void bc_program_init(void)
bc_vec_init(&G.prog.stack, sizeof(BcInstPtr), NULL); bc_vec_init(&G.prog.stack, sizeof(BcInstPtr), NULL);
bc_vec_push(&G.prog.stack, &ip); bc_vec_push(&G.prog.stack, &ip);
bc_char_vec_init(&G.stdin_buffer); bc_char_vec_init(&G.input_buffer);
} }
static int bc_vm_init(const char *env_len) static int bc_vm_init(const char *env_len)