X-Git-Url: https://diplodocus.org/git/nmh/blobdiff_plain/030c6e4df060fcbc33db4495148267e90c960d9d..9322ba2854211794c27fae9468768b80b767c211:/sbr/m_getfld.c diff --git a/sbr/m_getfld.c b/sbr/m_getfld.c index c6716284..4976b1c5 100644 --- a/sbr/m_getfld.c +++ b/sbr/m_getfld.c @@ -41,12 +41,9 @@ state to FLD. void m_unknown(FILE *iob): Determines the message delimiter string - for the maildrop. Called by inc, scan, and msh when reading from a + for the maildrop. Called by inc and scan when reading from a maildrop file. - void m_eomsbr (int (*action)(int)): Sets the hook to check for end - of message in a maildrop. Called only by msh. - State variables =============== m_getfld() retains state internally between calls in the @@ -61,7 +58,6 @@ int edelimlen char *msg_delim int msg_style - int (*eom_action)(int) Usage ===== @@ -94,7 +90,7 @@ be parsed as well. Unfortunately the speed issue finally caught up with us since this routine is at the very heart of MH. - To speed things up considerably, the routine Eom() was made an auxilary + To speed things up considerably, the routine Eom() was made an auxiliary function called by the macro eom(). Unless we are bursting a maildrop, the eom() macro returns FALSE saying we aren't at the end of the message. @@ -163,7 +159,7 @@ names are typically short (~8 char) and the loop that extracts them might terminate on a colon, newline or max width. I considered using a Vax "scanc" to locate the end of the field followed by a - "bcopy" but the routine call overhead on a Vax is too large for this + "memmove" but the routine call overhead on a Vax is too large for this to work on short names. If Berkeley ever makes "inline" part of the C optimiser (so things like "scanc" turn into inline instructions) a change here would be worthwhile. @@ -179,7 +175,7 @@ so message bodies average at least a few hundred characters. Assuming your system uses reasonably sized stdio buffers (1K or more), this routine should be able to remove the body in large - (>500 byte) chunks. The makes the cost of a call to "bcopy" + (>500 byte) chunks. The makes the cost of a call to "memmove" small but there is a premium on checking for the eom in packed maildrops. The eom pattern is always a simple string so we can construct an efficient pattern matcher for it (e.g., a Vax "matchc" @@ -207,12 +203,10 @@ * static prototypes */ struct m_getfld_state; -static int m_Eom (m_getfld_state_t, int); -static char *matchc(int, char *, int, char *); +static int m_Eom (m_getfld_state_t); #define eom(c,s) (s->msg_style != MS_DEFAULT && \ - (((c) == *s->msg_delim && m_Eom(s,c)) || \ - (s->eom_action && (*s->eom_action)(c)))) + ((c) == *s->msg_delim && m_Eom(s))) /* This replaces the old approach, with its direct access to stdio * internals. It uses one fread() to load a buffer that we manage. @@ -222,9 +216,9 @@ static char *matchc(int, char *, int, char *); * separate messages in a maildrop, such as mbox "From ". * * Some of the tests in the test suite assume a MSG_INPUT_SIZE - * of 4096. + * of 8192. */ -#define MSG_INPUT_SIZE 4096 +#define MSG_INPUT_SIZE NMH_BUFSIZ #define MAX_DELIMITER_SIZE 5 struct m_getfld_state { @@ -233,7 +227,7 @@ struct m_getfld_state { char *end; /* One past the last character read in. */ /* The following support tracking of the read position in the input file stream so that callers can interleave m_getfld() - calls with ftell() and fseek(). ytes_read replaces the old + calls with ftell() and fseek(). bytes_read replaces the old m_getfld() msg_count global. last_caller_pos is stored when leaving m_getfld()/m_unknown(), then checked on the next entry. last_internal_pos is used to remember the position used @@ -250,10 +244,10 @@ struct m_getfld_state { * The "full" delimiter string for a packed maildrop consists * of a newline followed by the actual delimiter. E.g., the * full string for a Unix maildrop would be: "\n\nFrom ". - * "Fdelim" points to the start of the full string and is used + * "fdelim" points to the start of the full string and is used * in the BODY case of the main routine to search the buffer for * a possible eom. Msg_delim points to the first character of - * the actual delim. string (i.e., fdelim+1). Edelim + * the actual delim. string (i.e., fdelim+1). edelim * points to the 2nd character of actual delimiter string. It * is used in m_Eom because the first character of the string * has been read and matched before m_Eom is called. @@ -264,7 +258,6 @@ struct m_getfld_state { int fdelimlen; char *edelim; int edelimlen; - int (*eom_action)(int); int state; int track_filepos; }; @@ -274,7 +267,8 @@ void m_getfld_state_init (m_getfld_state_t *gstate, FILE *iob) { m_getfld_state_t s; - s = *gstate = (m_getfld_state_t) mh_xmalloc(sizeof (struct m_getfld_state)); + NEW(s); + *gstate = s; s->readpos = s->end = s->msg_buf; s->bytes_read = s->total_bytes_read = 0; s->last_caller_pos = s->last_internal_pos = 0; @@ -284,7 +278,6 @@ m_getfld_state_init (m_getfld_state_t *gstate, FILE *iob) { s->msg_delim = ""; s->fdelim = s->delimend = s->edelim = NULL; s->fdelimlen = s->edelimlen = 0; - s->eom_action = NULL; s->state = FLD; s->track_filepos = 0; } @@ -313,7 +306,10 @@ void m_getfld_state_destroy (m_getfld_state_t *gstate) { m_getfld_state_t s = *gstate; if (s) { - if (s->fdelim) free (s->fdelim-1); + if (s->fdelim) { + free (s->fdelim-1); + free (s->pat_map); + } free (s); *gstate = 0; } @@ -351,7 +347,8 @@ void m_getfld_state_destroy (m_getfld_state_t *gstate) { static void enter_getfld (m_getfld_state_t *gstate, FILE *iob) { m_getfld_state_t s; - off_t pos = ftello (iob); + off_t pos; + off_t pos_movement; if (! *gstate) { m_getfld_state_init (gstate, iob); @@ -367,46 +364,51 @@ enter_getfld (m_getfld_state_t *gstate, FILE *iob) { readpos shift code being currently unused. */ s->iob = iob; - if (s->track_filepos && (pos != 0 || s->last_internal_pos != 0)) { - if (s->last_internal_pos == 0) { - s->total_bytes_read = pos; - } else { - off_t pos_movement = pos - s->last_caller_pos; /* Can be < 0. */ - - if (pos_movement == 0) { - pos = s->last_internal_pos; - } else { - /* The current file stream position differs from the - last one, so caller must have called ftell/o(). - Or, this is the first call and the file position - was not at 0. */ - - if (s->readpos + pos_movement >= s->msg_buf && - s->readpos + pos_movement < s->end) { - /* This is currently unused. It could be used by - parse_mime() if it was changed to use a global - m_getfld_state. */ - /* We can shift readpos and remain within the - bounds of msg_buf. */ - s->readpos += pos_movement; - s->total_bytes_read += pos_movement; - pos = s->last_internal_pos; - } else { - size_t num_read; - - /* This seek skips past an integral number of - chunks of size MSG_INPUT_SIZE. */ - fseeko (iob, pos/MSG_INPUT_SIZE * MSG_INPUT_SIZE, SEEK_SET); - num_read = fread (s->msg_buf, 1, MSG_INPUT_SIZE, iob); - s->readpos = s->msg_buf + pos % MSG_INPUT_SIZE; - s->end = s->msg_buf + num_read; - s->total_bytes_read = pos; - } - } + if (!s->track_filepos) + return; - fseeko (iob, pos, SEEK_SET); - } + pos = ftello(iob); + if (pos == 0 && s->last_internal_pos == 0) + return; + + if (s->last_internal_pos == 0) { + s->total_bytes_read = pos; + return; + } + + pos_movement = pos - s->last_caller_pos; /* Can be < 0. */ + if (pos_movement == 0) { + pos = s->last_internal_pos; + } else { + /* The current file stream position differs from the + last one, so caller must have called ftell/o(). + Or, this is the first call and the file position + was not at 0. */ + + if (s->readpos + pos_movement >= s->msg_buf && + s->readpos + pos_movement < s->end) { + /* This is currently unused. It could be used by + parse_mime() if it was changed to use a global + m_getfld_state. */ + /* We can shift readpos and remain within the + bounds of msg_buf. */ + s->readpos += pos_movement; + s->total_bytes_read += pos_movement; + pos = s->last_internal_pos; + } else { + size_t num_read; + + /* This seek skips past an integral number of + chunks of size MSG_INPUT_SIZE. */ + fseeko (iob, pos/MSG_INPUT_SIZE * MSG_INPUT_SIZE, SEEK_SET); + num_read = fread (s->msg_buf, 1, MSG_INPUT_SIZE, iob); + s->readpos = s->msg_buf + pos % MSG_INPUT_SIZE; + s->end = s->msg_buf + num_read; + s->total_bytes_read = pos; + } } + + fseeko (iob, pos, SEEK_SET); } static void @@ -430,8 +432,9 @@ read_more (m_getfld_state_t s) { ssize_t retain = s->edelimlen; size_t num_read; - if (retain < s->end - s->readpos) retain = s->end - s->readpos; - /* assert (retain <= s->readpos - s->msg_buf <= sizeof msg_buf); */ + if (retain < s->end - s->readpos) + retain = s->end - s->readpos; + assert (retain <= s->readpos - s->msg_buf); /* Move what we want to retain at end of the buffer to the beginning. */ memmove (s->msg_buf, s->readpos - retain, retain); @@ -448,28 +451,19 @@ read_more (m_getfld_state_t s) { but EOF is typically 0xffffffff. */ static int Getc (m_getfld_state_t s) { - if (s->end - s->readpos < 1) { - if (read_more (s) == 0) { - /* Pretend that we read a character. That's what stdio does. */ - ++s->readpos; - return EOF; - } - } + if ((s->end - s->readpos < 1 && read_more (s) == 0) || + s->readpos >= s->end) + return EOF; - ++s->bytes_read; - return s->readpos < s->end ? (unsigned char) *s->readpos++ : EOF; + s->bytes_read++; + return (unsigned char)*s->readpos++; } static int Peek (m_getfld_state_t s) { - if (s->end - s->readpos < 1) { - if (read_more (s) == 0) { - /* Pretend that we read a character. That's what stdio does. */ - ++s->readpos; - return EOF; - } + if (s->end - s->readpos < 1 && read_more (s) == 0) { + return EOF; } - return s->readpos < s->end ? (unsigned char) *s->readpos : EOF; } @@ -477,10 +471,9 @@ static int Ungetc (int c, m_getfld_state_t s) { if (s->readpos == s->msg_buf) { return EOF; - } else { - --s->bytes_read; - return *--s->readpos = (unsigned char) c; } + --s->bytes_read; + return *--s->readpos = (unsigned char) c; } @@ -489,26 +482,24 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, FILE *iob) { m_getfld_state_t s; - register char *cp; - register int max, n, c; + char *cp; + int max, n, c; enter_getfld (gstate, iob); s = *gstate; - if ((c = Getc(s)) < 0) { + if ((c = Getc(s)) == EOF) { *bufsz = *buf = 0; leave_getfld (s); return s->state = FILEEOF; } if (eom (c, s)) { - if (! s->eom_action) { - /* flush null messages */ - while ((c = Getc(s)) >= 0 && eom (c, s)) - ; + /* flush null messages */ + while ((c = Getc(s)) != EOF && eom (c, s)) + ; - if (c >= 0) - Ungetc(c, s); - } + if (c != EOF) + Ungetc(c, s); *bufsz = *buf = 0; leave_getfld (s); return s->state = FILEEOF; @@ -518,16 +509,15 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, case FLD: if (c == '\n' || c == '-') { /* we hit the header/body separator */ - while (c != '\n' && (c = Getc(s)) >= 0) continue; - - if (c < 0 || (c = Getc(s)) < 0 || eom (c, s)) { - if (! s->eom_action) { - /* flush null messages */ - while ((c = Getc(s)) >= 0 && eom (c, s)) - ; - if (c >= 0) - Ungetc(c, s); - } + while (c != '\n' && (c = Getc(s)) != EOF) + ; + + if (c == EOF || (c = Getc(s)) == EOF || eom (c, s)) { + /* flush null messages */ + while ((c = Getc(s)) != EOF && eom (c, s)) + ; + if (c != EOF) + Ungetc(c, s); *bufsz = *buf = 0; leave_getfld (s); return s->state = FILEEOF; @@ -591,13 +581,17 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, memcpy (buf, name, n - 1); buf[n - 1] = '\n'; buf[n] = '\0'; + /* Indicate this wasn't a header field using a character + that can't appear in a header field. */ + name[0] = ':'; /* The last character read was '\n'. s->bytes_read (and n) include that, but it was not put into the name array in the for loop above. So subtract 1. */ *bufsz = --s->bytes_read; /* == n - 1 */ leave_getfld (s); return s->state = BODY; - } else if (max <= n) { + } + if (max <= n) { /* By design, the loop above discards the last character it had read. It's in c, use it. */ *cp++ = c; @@ -612,7 +606,7 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, while (isspace ((unsigned char) *--cp) && cp >= name) continue; *++cp = 0; /* readpos points to the first character of the field body. */ - /* fall through */ + /* FALLTHRU */ case FLDPLUS: { /* @@ -627,12 +621,23 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, n = 0; for (finished = 0; ! finished; ) { while (c != '\n' && c != EOF && n++ < max) { - *cp++ = c = Getc (s); + if ((c = Getc (s)) != EOF) + *cp++ = c; } - if (c != EOF) c = Peek (s); + if (c != EOF) + c = Peek (s); if (max < n) { - /* the dest buffer is full */ + /* The dest buffer is full. Need to back the read + pointer up by one because when m_getfld() is + reentered, it will read a character. Then + we'll jump right to the FLDPLUS handling code, + which will not store that character, but + instead move on to the next one. */ + if (s->readpos > s->msg_buf) { + --s->readpos; + --s->bytes_read; + } s->state = FLDPLUS; finished = 1; } else if (c != ' ' && c != '\t') { @@ -658,6 +663,7 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, */ char *bp; + name[0] = '\0'; max = *bufsz-1; /* Back up and store the current position. */ bp = --s->readpos; @@ -677,7 +683,7 @@ m_getfld (m_getfld_state_t *gstate, char name[NAMESZ], char *buf, int *bufsz, */ char *ep; - if ((ep = matchc( s->fdelimlen, s->fdelim, c, bp ))) + if ((ep = memmem(bp, c, s->fdelim, s->fdelimlen))) c = ep - bp + 1; else { /* @@ -761,11 +767,11 @@ void m_unknown(m_getfld_state_t *gstate, FILE *iob) { m_getfld_state_t s; - register int c; + int c; char text[MAX_DELIMITER_SIZE]; char from[] = "From "; - register char *cp; - register char *delimstr; + char *cp; + char *delimstr; unsigned int i; enter_getfld (gstate, iob); @@ -787,49 +793,68 @@ m_unknown(m_getfld_state_t *gstate, FILE *iob) s->msg_style = MS_UNKNOWN; for (i = 0, cp = text; i < sizeof text; ++i, ++cp) { - if ((*cp = Getc (s)) == EOF) { + if ((c = Getc (s)) == EOF) { + *cp = '\0'; break; + } else { + *cp = c; } } if (i == sizeof from-1 && strncmp (text, "From ", sizeof from-1) == 0) { s->msg_style = MS_MBOX; delimstr = "\nFrom "; - while ((c = Getc (s)) != '\n' && c >= 0) continue; + while ((c = Getc(s)) != EOF && c != '\n') + ; } else { /* not a Unix style maildrop */ s->readpos -= s->bytes_read; + s->bytes_read = 0; delimstr = mmdlm2; s->msg_style = MS_MMDF; } + c = strlen (delimstr); - s->fdelim = mh_xmalloc (c + 3); + s->fdelim = mh_xmalloc (c + 3); /* \0, \n, delimstr, \0 */ *s->fdelim++ = '\0'; *s->fdelim = '\n'; - s->msg_delim = s->fdelim+1; - s->edelim = s->msg_delim+1; s->fdelimlen = c + 1; - s->edelimlen = c - 1; /* == strlen (delimstr) */ + s->msg_delim = s->fdelim+1; strcpy (s->msg_delim, delimstr); + s->edelim = s->msg_delim+1; + s->edelimlen = c - 1; s->delimend = s->msg_delim + s->edelimlen; if (s->edelimlen <= 1) adios (NULL, "maildrop delimiter must be at least 2 bytes"); + + /* Now malloc'd memory at s->fdelim-1 referenced several times: + * + * delimstr "\nFrom " "\001\001\001\001\n" + * c 6 5 + * s->fdelim \0"\n\nFrom " \0"\n\001\001\001\001\n" + * s->fdelimlen 6 5 + * s->msg_delim "\nFrom " "\001\001\001\001\n" + * s->edelim "From " "\001\001\001\n" + * s->edelimlen 5 4 + * s->delimend " " "\n" + */ + /* * build a Boyer-Moore end-position map for the matcher in m_getfld. * N.B. - we don't match just the first char (since it's the newline * separator) or the last char (since the matchc would have found it * if it was a real delim). */ - s->pat_map = (char **) calloc (256, sizeof(char *)); + s->pat_map = (char **) mh_xcalloc (256, sizeof(char *)); for (cp = s->fdelim + 1; cp < s->delimend; cp++ ) s->pat_map[(unsigned char)*cp] = cp; if (s->msg_style == MS_MMDF) { /* flush extra msg hdrs */ - while ((c = Getc(s)) >= 0 && eom (c, s)) + while ((c = Getc(s)) != EOF && eom (c, s)) ; - if (c >= 0) + if (c != EOF) Ungetc(c, s); } @@ -837,37 +862,25 @@ m_unknown(m_getfld_state_t *gstate, FILE *iob) } -void -m_eomsbr (m_getfld_state_t s, int (*action)(int)) -{ - if ((s->eom_action = action)) { - s->msg_style = MS_MSH; - *s->msg_delim = 0; - s->fdelimlen = 1; - s->delimend = s->fdelim; - } else { - s->msg_style = MS_MMDF; - s->msg_delim = s->fdelim + 1; - s->fdelimlen = strlen (s->fdelim); - s->delimend = s->msg_delim + s->edelimlen; - } -} - - /* * test for msg delimiter string */ static int -m_Eom (m_getfld_state_t s, int c) +m_Eom (m_getfld_state_t s) { - register int i; + int i; char text[MAX_DELIMITER_SIZE]; char *cp; for (i = 0, cp = text; i < s->edelimlen; ++i, ++cp) { - if ((*cp = Getc (s)) == EOF) { + int c2; + + if ((c2 = Getc (s)) == EOF) { + *cp = '\0'; break; + } else { + *cp = c2; } } @@ -883,38 +896,15 @@ m_Eom (m_getfld_state_t s, int c) Note that on input, a character had already been read with Getc(). It will be unget by m_getfld () on return. */ s->readpos -= s->bytes_read - 1; + s->bytes_read = 1; return 0; } if (s->msg_style == MS_MBOX) { - while ((c = Getc (s)) != '\n') - if (c < 0) - break; + int c; + while ((c = Getc(s)) != EOF && c != '\n') + ; } return 1; } - - -static char * -matchc(int patln, char *pat, int strln, char *str) -{ - register char *es = str + strln - patln; - register char *sp; - register char *pp; - register char *ep = pat + patln; - register char pc = *pat++; - - for(;;) { - while (pc != *str++) - if (str > es) - return 0; - if (str > es+1) - return 0; - sp = str; pp = pat; - while (pp < ep && *sp++ == *pp) - pp++; - if (pp >= ep) - return --str; - } -}