X-Git-Url: http://git.marmaro.de/?a=blobdiff_plain;f=sbr%2Fm_getfld.c;h=108af13826a0bcce5a2986babaac6a664c8d9d72;hb=976303d04d3bc2cad0afd5e3d364264783da56c2;hp=548b3eaf2c4f7c049da7e6f339e811a10dcd24dd;hpb=5e8e1811bd39ba43dda7a3446694474ab16ab960;p=mmh diff --git a/sbr/m_getfld.c b/sbr/m_getfld.c index 548b3ea..108af13 100644 --- a/sbr/m_getfld.c +++ b/sbr/m_getfld.c @@ -8,53 +8,22 @@ #include #include +#include +#include /* -** This module has a long and checkered history. First, it didn't burst -** maildrops correctly because it considered two CTRL-A:s in a row to be -** an inter-message delimiter. It really is four CTRL-A:s followed by a -** newline. Unfortunately, MMDF will convert this delimiter *inside* a -** message to a CTRL-B followed by three CTRL-A:s and a newline. This -** caused the old version of m_getfld() to declare eom prematurely. The -** fix was a lot slower than +** This module has a long and checkered history. ** -** c == '\001' && peekc (iob) == '\001' +** [ Here had been some history of delimiter problems in MMDF maildrops ... ] ** -** but it worked, and to increase generality, MBOX style maildrops could -** be parsed as well. Unfortunately the speed issue finally caught up with -** us since this routine is at the very heart of MH. +** 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 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. ** -** To speed things up considerably, the routine Eom() was made an auxilary -** 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. +** [ ... and here had been some more of it. ] ** -** After mhl was made a built-in in msh, m_getfld() worked just fine -** (using m_unknown() at startup). Until one day: a message which was -** the result of a bursting was shown. Then, since the burst boundaries -** aren't CTRL-A:s, m_getfld() would blinding plunge on past the boundary. -** Very sad. The solution: introduce m_eomsbr(). This hook gets called -** after the end of each line (since testing for eom involves an fseek()). -** This worked fine, until one day: a message with no body portion arrived. -** Then the -** -** while (eom(c = getc(iob), iob)) -** continue; -** -** loop caused m_getfld() to return FMTERR. So, that logic was changed to -** check for (*eom_action) and act accordingly. -** -** [ Note by meillo 2011-10: -** as msh was removed from mmh, m_eomsbr() became irrelevant. ] -** -** This worked fine, until one day: someone didn't use four CTRL:A's as -** their delimiters. So, the bullet got bit and we read mts.h and -** continue to struggle on. It's not that bad though, since the only time -** the code gets executed is when inc (or msh) calls it, and both of these -** have already called mts_init(). -** -** [ Note by meillo 2012-02: -** MMDF-style maildrops (4x ^A) and mts_init() were removed. ] ** ** ------------------------ ** (Written by Van Jacobson for the mh6 m_getfld, January, 1986): @@ -76,9 +45,8 @@ ** If you hack on this and slow it down, I, my children and my ** children's children will curse you. ** -** This routine gets used on three different types of files: normal, -** single msg files, "packed" unix or mmdf mailboxs (when used by inc) -** and packed, directoried bulletin board files (when used by msh). +** This routine gets used on two different types of files: normal, +** single msg files and "packed" unix mailboxs (when used by inc). ** The biggest impact of different file types is in "eom" testing. The ** code has been carefully organized to test for eom at appropriate ** times and at no other times (since the check is quite expensive). @@ -171,7 +139,7 @@ int ismbox = FALSE; ** is used in m_Eom because the first character of the string ** has been read and matched before m_Eom is called. */ -char *msg_delim = ""; +static char *msg_delim = ""; static unsigned char *fdelim; static unsigned char *delimend; @@ -181,34 +149,85 @@ static int edelimlen; static int (*eom_action)(int) = NULL; -#ifdef _FSTDIO -# define _ptr _p /* Gag */ -# define _cnt _r /* Retch */ -# define _filbuf __srget /* Puke */ -# define DEFINED__FILBUF_TO_SOMETHING_SPECIFIC -#endif +/* +** This replaces the old approach, which included direct access to +** stdio internals. It uses one fread() to load a buffer that we +** manage. +*/ +#define MSG_INPUT_SIZE 8192 +static struct m_getfld_buffer { + unsigned char msg_buf[2 * MSG_INPUT_SIZE]; + unsigned char *readpos; + unsigned char *end; /* One past the last character read in. */ +} m; + +static void +setup_buffer(FILE *iob, struct m_getfld_buffer *m) +{ + /* + ** Rely on Restrictions that m_getfld() calls on different file + ** streams are not interleaved, and no other file stream read + ** methods are used. And, the first call to m_getfld (), etc., on + ** a stream always reads at least 1 byte. + ** I don't think it's necessary to use ftello() because we just + ** need to determine whether the current offset is 0 or not. + */ + if (ftell(iob) == 0) { + /* A new file stream, so reset the buffer state. */ + m->readpos = m->end = m->msg_buf; + } +} + +static size_t +read_more(struct m_getfld_buffer *m, FILE *iob) +{ + size_t num_read; + + /* Move any leftover at the end of buf to the beginning. */ + if (m->end > m->readpos) { + memmove(m->msg_buf, m->readpos, m->end - m->readpos); + } + m->readpos = m->msg_buf + (m->end - m->readpos); + num_read = fread(m->readpos, 1, MSG_INPUT_SIZE, iob); + m->end = m->readpos + num_read; + + return num_read; +} + +static int +Getc(FILE *iob) +{ + if (m.end - m.readpos < 1) { + if (read_more(&m, iob) == 0) { + /* + ** Pretend that we read a character. + ** That's what stdio does. + */ + ++m.readpos; + return EOF; + } + } + return (m.readpos < m.end) ? *m.readpos++ : EOF; +} -#ifdef SCO_5_STDIO -# define _ptr __ptr -# define _cnt __cnt -# define _base __base -# define _filbuf(fp) ((fp)->__cnt = 0, __filbuf(fp)) -# define DEFINED__FILBUF_TO_SOMETHING_SPECIFIC -#endif +static int +Ungetc(int c, FILE *iob) +{ + return (m.readpos == m.msg_buf) ? EOF : (*--m.readpos = c); +} -#ifndef DEFINED__FILBUF_TO_SOMETHING_SPECIFIC -extern int _filbuf(FILE*); -#endif int m_getfld(int state, unsigned char *name, unsigned char *buf, int bufsz, FILE *iob) { - register unsigned char *bp, *cp, *ep, *sp; - register int cnt, c, i, j; + unsigned char *bp, *cp, *ep, *sp; + int cnt, c, i, j, k; + + setup_buffer(iob, &m); - if ((c = getc(iob)) < 0) { + if ((c = Getc(iob)) < 0) { msg_count = 0; *buf = 0; return FILEEOF; @@ -216,10 +235,10 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, if (eom(c, iob)) { if (! eom_action) { /* flush null messages */ - while ((c = getc(iob)) >= 0 && eom(c, iob)) + while ((c = Getc(iob)) >= 0 && eom(c, iob)) ; if (c >= 0) - ungetc(c, iob); + Ungetc(c, iob); } msg_count = 0; *buf = 0; @@ -232,16 +251,16 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, case FLD: if (c == '\n' || c == '-') { /* we hit the header/body separator */ - while (c != '\n' && (c = getc(iob)) >= 0) + while (c != '\n' && (c = Getc(iob)) >= 0) ; - if (c < 0 || (c = getc(iob)) < 0 || eom(c, iob)) { + if (c < 0 || (c = Getc(iob)) < 0 || eom(c, iob)) { if (!eom_action) { /* flush null messages */ - while ((c = getc(iob)) >= 0 && eom(c, iob)) + while ((c = Getc(iob)) >= 0 && eom(c, iob)) ; if (c >= 0) - ungetc(c, iob); + Ungetc(c, iob); } msg_count = 0; *buf = 0; @@ -258,47 +277,28 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, cp = name; i = NAMESZ - 1; for (;;) { -#ifdef LINUX_STDIO - bp = sp = (unsigned char *) iob->_IO_read_ptr - 1; - j = (cnt = ((long) iob->_IO_read_end - - (long) iob->_IO_read_ptr) + 1) < i ? cnt : i; -#elif defined(__DragonFly__) - bp = sp = (unsigned char *) ((struct __FILE_public *)iob)->_p - 1; - j = (cnt = ((struct __FILE_public *)iob)->_r+1) < i ? cnt : i; -#else - bp = sp = (unsigned char *) iob->_ptr - 1; - j = (cnt = iob->_cnt+1) < i ? cnt : i; -#endif + /* Store current pos, ungetting the last char. */ + bp = sp = (unsigned char *) m.readpos - 1; + j = ((cnt = m.end - m.readpos + 1) < i) ? cnt : i; + while (--j >= 0 && (c = *bp++) != ':' && c != '\n') *cp++ = c; j = bp - sp; if ((cnt -= j) <= 0) { -#ifdef LINUX_STDIO - iob->_IO_read_ptr = iob->_IO_read_end; - if (__underflow(iob) == EOF) { -#elif defined(__DragonFly__) - if (__srget(iob) == EOF) { -#else - if (_filbuf(iob) == EOF) { -#endif + /* + ** Used to explicitly force refill of the + ** buffer here, but Getc() will do that + ** if necessary. + */ + if (Getc (iob) == EOF) { *cp = *buf = 0; advise(NULL, "eof encountered in field \"%s\"", name); return FMTERR; } -#ifdef LINUX_STDIO - iob->_IO_read_ptr++; /* NOT automatic in __underflow()! */ -#endif } else { -#ifdef LINUX_STDIO - iob->_IO_read_ptr = bp + 1; -#elif defined(__DragonFly__) - ((struct __FILE_public *)iob)->_p = bp + 1; - ((struct __FILE_public *)iob)->_r = cnt - 1; -#else - iob->_ptr = bp + 1; - iob->_cnt = cnt - 1; -#endif + /* Restore the current offset. */ + m.readpos = bp + 1; } if (c == ':') break; @@ -352,7 +352,13 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, ** that should be harmless enough, right? ** This is a corrupt message anyway. */ - fseek(iob, ftell(iob) - 2, SEEK_SET); + /* emulates: fseek(iob, ftell(iob) -(-2 + cnt + 1), SEEK_SET) */ + m.readpos += cnt - 1; + /* + ** Reset file stream position so caller, + ** e.g., get_content, can use ftell(), etc. + */ + fseek(iob, -cnt - 1, SEEK_CUR); return BODY; } if ((i -= j) <= 0) { @@ -376,16 +382,9 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, */ cp = buf; i = bufsz-1; for (;;) { -#ifdef LINUX_STDIO - cnt = (long) iob->_IO_read_end - (long) iob->_IO_read_ptr; - bp = (unsigned char *) --iob->_IO_read_ptr; -#elif defined(__DragonFly__) - cnt = ((struct __FILE_public *)iob)->_r++; - bp = (unsigned char *) --((struct __FILE_public *)iob)->_p; -#else - cnt = iob->_cnt++; - bp = (unsigned char *) --iob->_ptr; -#endif + /* Set and save the current pos and update cnt. */ + cnt = m.end - m.readpos; + bp = --m.readpos; c = cnt < i ? cnt : i; while ((ep = locc( c, bp, '\n' ))) { /* @@ -393,21 +392,13 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, ** return. */ if ((j = *++ep) != ' ' && j != '\t') { -#ifdef LINUX_STDIO - j = ep - (unsigned char *) iob->_IO_read_ptr; - memcpy(cp, iob->_IO_read_ptr, j); - iob->_IO_read_ptr = ep; -#elif defined(__DragonFly__) - j = ep - (unsigned char *) ((struct __FILE_public *)iob)->_p; - memcpy(cp, ((struct __FILE_public *)iob)->_p, j); - ((struct __FILE_public *)iob)->_p = ep; - ((struct __FILE_public *)iob)->_r -= j; -#else - j = ep - (unsigned char *) iob->_ptr; - memcpy(cp, iob->_ptr, j); - iob->_ptr = ep; - iob->_cnt -= j; -#endif + /* + ** Save the text and update the + ** current position. + */ + j = ep - m.readpos; + memcpy (cp, m.readpos, j); + m.readpos = ep; cp += j; state = FLD; goto finish; @@ -419,64 +410,34 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, ** end of input or dest buffer - copy what ** we've found. */ -#ifdef LINUX_STDIO - c += bp - (unsigned char *) iob->_IO_read_ptr; - memcpy(cp, iob->_IO_read_ptr, c); -#elif defined(__DragonFly__) - c += bp - (unsigned char *) ((struct __FILE_public *)iob)->_p; - memcpy(cp, ((struct __FILE_public *)iob)->_p, c); -#else - c += bp - (unsigned char *) iob->_ptr; - memcpy(cp, iob->_ptr, c); -#endif - i -= c; - cp += c; + c += bp - m.readpos; + for (k = 0; k < c; ++k, --i) { + *cp++ = Getc (iob); + } if (i <= 0) { /* the dest buffer is full */ -#ifdef LINUX_STDIO - iob->_IO_read_ptr += c; -#elif defined(__DragonFly__) - ((struct __FILE_public *)iob)->_r -= c; - ((struct __FILE_public *)iob)->_p += c; -#else - iob->_cnt -= c; - iob->_ptr += c; -#endif state = FLDPLUS; break; } /* ** There's one character left in the input - ** buffer. Copy it & fill the buffer. - ** If the last char was a newline and the - ** next char is not whitespace, this is - ** the end of the field. Otherwise loop. + ** buffer. Copy it & fill the buffer (that + ** fill used to be explicit, but now Getc() + ** does it). If the last char was a newline + ** and the next char is not whitespace, this + ** is the end of the field. Otherwise loop. */ --i; -#ifdef LINUX_STDIO - *cp++ = j = *(iob->_IO_read_ptr + c); - iob->_IO_read_ptr = iob->_IO_read_end; - c = __underflow(iob); - iob->_IO_read_ptr++; /* NOT automatic! */ -#elif defined(__DragonFly__) - *cp++ =j = *(((struct __FILE_public *)iob)->_p + c); - c = __srget(iob); -#else - *cp++ = j = *(iob->_ptr + c); - c = _filbuf(iob); -#endif - if (c == EOF || - ((j == '\0' || j == '\n') && c != ' ' && c != '\t')) { + *cp++ = j = Getc(iob); + c = Getc(iob); + if (c == EOF || ((j == '\0' || j == '\n') + && c != ' ' && c != '\t')) { if (c != EOF) { -#ifdef LINUX_STDIO - --iob->_IO_read_ptr; -#elif defined(__DragonFly__) - --((struct __FILE_public *)iob)->_p; - ++((struct __FILE_public *)iob)->_r; -#else - --iob->_ptr; - ++iob->_cnt; -#endif + /* + ** Put the character back for + ** the next call. + */ + --m.readpos; } state = FLD; break; @@ -494,17 +455,10 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, ** don't add an eos. */ i = (bufsz < 0) ? -bufsz : bufsz-1; -#ifdef LINUX_STDIO - bp = (unsigned char *) --iob->_IO_read_ptr; - cnt = (long) iob->_IO_read_end - (long) iob->_IO_read_ptr; -#elif defined(__DragonFly__) - bp = (unsigned char *) --((struct __FILE_public *)iob)->_p; - cnt = ++((struct __FILE_public *)iob)->_r; -#else - bp = (unsigned char *) --iob->_ptr; - cnt = ++iob->_cnt; -#endif - c = (cnt < i ? cnt : i); + /* Back up and store the current position and update cnt. */ + bp = --m.readpos; + cnt = m.end - m.readpos; + c = (cnt < i) ? cnt : i; if (ismbox && c > 1) { /* ** packed maildrop - only take up to the (possible) @@ -587,15 +541,8 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, } } memcpy( buf, bp, c ); -#ifdef LINUX_STDIO - iob->_IO_read_ptr += c; -#elif defined(__DragonFly__) - ((struct __FILE_public *)iob)->_r -= c; - ((struct __FILE_public *)iob)->_p += c; -#else - iob->_cnt -= c; - iob->_ptr += c; -#endif + /* Advance the current position to reflect the copy out. */ + m.readpos += c; if (bufsz < 0) { msg_count = c; return (state); @@ -604,7 +551,7 @@ m_getfld(int state, unsigned char *name, unsigned char *buf, break; default: - adios(NULL, "m_getfld() called with bogus state of %d", state); + adios(EX_SOFTWARE, NULL, "m_getfld() called with bogus state of %d", state); } finish: *cp = 0; @@ -612,16 +559,22 @@ finish: return (state); } -static char fromline[BUFSIZ] = ""; void -m_unknown(FILE *iob) +thisisanmbox(FILE *iob) { - register int c; - register long pos; + int c; char text[10]; - register char *cp; - register char *delimstr; + char *cp; + char *delimstr; + + setup_buffer(iob, &m); + + c = Getc(iob); + if (feof(iob)) { + return; + } + Ungetc(c, iob); /* ** Figure out what the message delimitter string is for this @@ -634,21 +587,22 @@ m_unknown(FILE *iob) ** say the style is MBOX and eat the rest of the line. Otherwise ** abort. */ - - pos = ftell(iob); - if (fread(text, sizeof(*text), 5, iob) != 5) { - adios(NULL, "Read error"); + for (c=0, cp=text; c<5; ++c, ++cp) { + if ((*cp = Getc(iob)) == EOF) { + break; + } + } + if (c != 5) { + adios(EX_IOERR, NULL, "Read error"); } if (strncmp(text, "From ", 5)!=0) { - adios(NULL, "No Unix style (mbox) maildrop."); + adios(EX_USAGE, NULL, "No Unix style (mbox) maildrop."); } ismbox = TRUE; delimstr = "\nFrom "; - cp = fromline; - while ((c = getc(iob)) != '\n' && cp - fromline < BUFSIZ - 1) - *cp++ = c; - *cp = '\0'; - + while ((c = Getc(iob)) != '\n' && c >= 0) { + continue; + } c = strlen(delimstr); fdelim = (unsigned char *) mh_xmalloc((size_t) (c + 3)); *fdelim++ = '\0'; @@ -659,8 +613,9 @@ m_unknown(FILE *iob) edelimlen = c - 1; strcpy(msg_delim, delimstr); delimend = (unsigned char *)msg_delim + edelimlen; - if (edelimlen <= 1) - adios(NULL, "maildrop delimiter must be at least 2 bytes"); + if (edelimlen <= 1) { + adios(EX_DATAERR, NULL, "maildrop delimiter must be at least 2 bytes"); + } /* ** 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 @@ -669,8 +624,9 @@ m_unknown(FILE *iob) */ pat_map = (unsigned char **) calloc(256, sizeof(unsigned char *)); - for (cp = (char *) fdelim + 1; cp < (char *) delimend; cp++ ) + for (cp = (char *) fdelim + 1; cp < (char *) delimend; cp++) { pat_map[(unsigned char)*cp] = (unsigned char *) cp; + } } @@ -681,72 +637,49 @@ m_unknown(FILE *iob) static int m_Eom(int c, FILE *iob) { - register long pos = 0L; - register int i; + unsigned char *pos; + int i; char text[10]; - register char *cp; + char *cp; - pos = ftell(iob); - if ((i = fread(text, sizeof *text, edelimlen, iob)) != edelimlen || + pos = m.readpos; /* ftell */ + for (i=0, cp=text; i= 0 && cp - fromline < BUFSIZ - 1) - *cp++ = c; - *cp = '\0'; + while ((c = Getc(iob)) != '\n' && c >= 0) { + continue; + } } return 1; } -/* -** Return the Return-Path and Delivery-Date header information. -** -** Currently, I'm assuming that the "From " line takes the following form: -** "From" sender@host date (sendmail delivery) -*/ -int -get_returnpath(char *rp, int rplen, char *dd, int ddlen) -{ - char *cp; - - if (!(cp = strchr(fromline, ' '))) { - return 0; - } - /* Extract the Return-Path. */ - snprintf(rp, rplen, "%.*s\n", (int)(cp - fromline), fromline); - /* Advance over the whitespace. */ - while (*cp == ' ' || *cp == '\t') { - cp++; - } - /* Extract the Delivery-Date. */ - snprintf(dd, ddlen, "%.*s\n", 24, cp); - *fromline = '\0'; - return 1; -} - - static unsigned 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++; + char *es = str + strln - patln; + char *sp; + char *pp; + char *ep = pat + patln; + char pc = *pat++; for(;;) { while (pc != *str++)