/*
* scansbr.c -- routines to help scan along...
*
- * $Id$
- *
* This code is Copyright (c) 2002, by the authors of nmh. See the
* COPYRIGHT file in the root directory of the nmh distribution for
* complete copyright information.
#include <h/tws.h>
#include <h/utils.h>
-#ifdef _FSTDIO
-# define _ptr _p /* Gag */
-# define _cnt _w /* Wretch */
-#endif
-
-#ifdef SCO_5_STDIO
-# define _ptr __ptr
-# define _cnt __cnt
-# define _base __base
-# define _filbuf(fp) ((fp)->__cnt = 0, __filbuf(fp))
-#endif
-
#define MAXSCANL 256 /* longest possible scan line */
/*
#define SBUFSIZ 512
static struct format *fmt;
-#ifdef JLR
-static struct format *fmt_top;
-#endif /* JLR */
-
static struct comp *datecomp; /* pntr to "date" comp */
static struct comp *bodycomp; /* pntr to "body" pseudo-comp *
* (if referenced) */
/*
* prototypes
*/
-int sc_width (void); /* from termsbr.c */
static int mh_fputs(char *, FILE *);
+#ifdef MULTIBYTE_SUPPORT
+#define SCAN_CHARWIDTH MB_CUR_MAX
+#else
+#define SCAN_CHARWIDTH 1
+#endif
int
scan (FILE *inb, int innum, int outnum, char *nfs, int width, int curflg,
int unseen, char *folder, long size, int noisy)
{
int i, compnum, encrypted, state;
- char *cp, *tmpbuf, **nxtbuf;
- char *saved_c_text;
+ unsigned char *cp, *tmpbuf, *startbody;
+ char **nxtbuf;
+ char *saved_c_text = NULL;
struct comp *cptr;
struct comp **savecomp;
- char *scnmsg;
- FILE *scnout;
+ char *scnmsg = NULL;
+ FILE *scnout = NULL;
char name[NAMESZ];
static int rlwidth, slwidth;
-
-#ifdef RPATHS
- char returnpath[BUFSIZ];
- char deliverydate[BUFSIZ];
-#endif
+ static size_t scanl_size;
/* first-time only initialization */
if (!scanl) {
width = MAXSCANL;
}
dat[3] = slwidth = width;
- scanl = (char *) mh_xmalloc((size_t) (slwidth + 2) );
+ /* Arbitrarily allocate 20 * slwidth to provide room for lots
+ of escape sequences. */
+ scanl_size = SCAN_CHARWIDTH * (20 * slwidth + 2);
+ scanl = (char *) mh_xmalloc (scanl_size);
if (outnum)
umask(~m_gmprot());
/* Compile format string */
- ncomps = fmt_compile (nfs, &fmt) + 1;
-
-#ifdef JLR
- fmt_top = fmt;
-#endif /* JLR */
- FINDCOMP(bodycomp, "body");
- FINDCOMP(datecomp, "date");
- FINDCOMP(cptr, "folder");
+ ncomps = fmt_compile (nfs, &fmt, 1) + 2;
+
+ bodycomp = fmt_findcomp("body");
+ datecomp = fmt_findcomp("date");
+ cptr = fmt_findcomp("folder");
if (cptr && folder)
- cptr->c_text = folder;
- FINDCOMP(cptr, "encrypted");
- if (!cptr)
- if ((cptr = (struct comp *) calloc (1, sizeof(*cptr)))) {
- cptr->c_name = "encrypted";
- cptr->c_next = wantcomp[i = CHASH (cptr->c_name)];
- wantcomp[i] = cptr;
+ cptr->c_text = getcpy(folder);
+ if (fmt_addcompentry("encrypted")) {
ncomps++;
}
- FINDCOMP (cptr, "dtimenow");
+ cptr = fmt_findcomp("dtimenow");
if (cptr)
cptr->c_text = getcpy(dtimenow (0));
+
+ /*
+ * In other programs I got rid of this complicated buffer switching,
+ * but since scan reads lots of messages at once and this complicated
+ * memory management, I decided to keep it; otherwise there was
+ * the potential for a lot of malloc() and free()s, and I could
+ * see the malloc() pool really getting fragmented. Maybe it
+ * wouldn't be an issue in practice; perhaps this will get
+ * revisited someday.
+ *
+ * So, some notes for what's going on:
+ *
+ * nxtbuf is an array of pointers that contains malloc()'d buffers
+ * to hold our component text. used_buf is an array of struct comp
+ * pointers that holds pointers to component structures we found while
+ * processing a message.
+ *
+ * We read in the message with m_getfld(), using "tmpbuf" as our
+ * input buffer. tmpbuf is set at the start of message processing
+ * to the first buffer in our buffer pool (nxtbuf).
+ *
+ * Every time we find a component we care about, we set that component's
+ * text buffer to the current value of tmpbuf, and then switch tmpbuf
+ * to the next buffer in our pool. We also add that component to
+ * our used_buf pool.
+ *
+ * When we're done, we go back and zero out all of the component
+ * text buffer pointers that we saved in used_buf.
+ *
+ * Note that this means c_text memory is NOT owned by the fmt_module
+ * and it's our responsibility to free it.
+ */
+
nxtbuf = compbuffers = (char **) calloc((size_t) ncomps, sizeof(char *));
if (nxtbuf == NULL)
adios (NULL, "unable to allocate component buffers");
nxtbuf = compbuffers;
savecomp = used_buf;
tmpbuf = *nxtbuf++;
+ startbody = NULL;
dat[0] = innum ? innum : outnum;
dat[1] = curflg;
dat[4] = unseen;
}
if ((scnout = fopen (scnmsg, "w")) == NULL)
adios (scnmsg, "unable to write");
-#ifdef RPATHS
- /*
- * Add the Return-Path and Delivery-Date
- * header fields to message.
- */
- if (get_returnpath (returnpath, sizeof(returnpath),
- deliverydate, sizeof(deliverydate))) {
- FPUTS ("Return-Path: ");
- FPUTS (returnpath);
- FPUTS ("Delivery-Date: ");
- FPUTS (deliverydate);
- }
-#endif /* RPATHS */
}
/* scan - main loop */
* buffer as the component temp buffer (buffer switching
* saves an extra copy of the component text).
*/
- if ((cptr = wantcomp[CHASH(name)])) {
- do {
- if (!strcasecmp(name, cptr->c_name)) {
- if (! cptr->c_text) {
- cptr->c_text = tmpbuf;
- for (cp = tmpbuf + strlen (tmpbuf) - 1;
+ if ((cptr = fmt_findcasecomp(name))) {
+ if (! cptr->c_text) {
+ cptr->c_text = tmpbuf;
+ for (cp = tmpbuf + strlen (tmpbuf) - 1;
cp >= tmpbuf; cp--)
- if (isspace (*cp))
- *cp = 0;
- else
- break;
- *--savecomp = cptr;
- tmpbuf = *nxtbuf++;
- }
- break;
- }
- } while ((cptr = cptr->c_next));
+ if (isspace (*cp))
+ *cp = 0;
+ else
+ break;
+ *--savecomp = cptr;
+ tmpbuf = *nxtbuf++;
+ }
}
while (state == FLDPLUS) {
case BODY:
compnum = -1;
+ /*
+ * A slight hack ... if we have less than rlwidth characters
+ * in the buffer, call m_getfld again.
+ */
+
+ if ((i = strlen(tmpbuf)) < rlwidth) {
+ state = m_getfld (state, name, tmpbuf + i,
+ rlwidth - i, inb);
+ }
+
if (! outnum) {
state = FILEEOF; /* stop now if scan cmd */
+ if (bodycomp && startbody == NULL)
+ startbody = tmpbuf;
goto finished;
}
if (putc ('\n', scnout) == EOF) DIEWRERR();
FPUTS (tmpbuf);
/*
- * performance hack: some people like to run "inc" on
- * things like net.sources or large digests. We do a
- * copy directly into the output buffer rather than
- * going through an intermediate buffer.
+ * The previous code here used to call m_getfld() using
+ * pointers to the underlying output stdio buffers to
+ * avoid the extra copy. Tests by Markus Schnalke show
+ * no noticable performance loss on larger mailboxes
+ * if we incur an extra copy, and messing around with
+ * internal stdio buffers is becoming more and more
+ * unportable as times go on. So from now on just deal
+ * with the overhead of an extra copy.
*
- * We need the amount of data m_getfld found & don't
- * want to do a strlen on the long buffer so there's
- * a hack in m_getfld to save the amount of data it
- * returned in the global "msg_count".
+ * Subtle change - with the previous code tmpbuf wasn't
+ * used, so we could reuse it for the {body} component.
+ * Now since we're using tmpbuf as our read buffer we
+ * need to save the beginning of the body for later.
+ * See the above (and below) use of startbody.
*/
body:;
+ if (bodycomp && startbody == NULL) {
+ startbody = tmpbuf;
+ tmpbuf = *nxtbuf++;
+ }
+
while (state == BODY) {
-#ifdef LINUX_STDIO
- if (scnout->_IO_write_ptr == scnout->_IO_write_end) {
-#else
- if (scnout->_cnt <= 0) {
-#endif
- if (fflush(scnout) == EOF)
- DIEWRERR ();
- }
-#ifdef LINUX_STDIO
- state = m_getfld(state, name, scnout->_IO_write_ptr,
- (long)scnout->_IO_write_ptr-(long)scnout->_IO_write_end , inb);
- scnout->_IO_write_ptr += msg_count;
-#else
- state = m_getfld( state, name, scnout->_ptr, -(scnout->_cnt), inb );
- scnout->_cnt -= msg_count;
- scnout->_ptr += msg_count;
-#endif
+ state = m_getfld(state, name, tmpbuf, rlwidth, inb);
+ FPUTS(tmpbuf);
}
goto finished;
/* Save and restore buffer so we don't trash our dynamic pool! */
if (bodycomp) {
saved_c_text = bodycomp->c_text;
- bodycomp->c_text = tmpbuf;
+ bodycomp->c_text = startbody;
}
if (size)
}
}
- fmt_scan (fmt, scanl, slwidth, dat);
-
-#if 0
- fmt = fmt_scan (fmt, scanl, slwidth, dat);
- if (!fmt)
- fmt = fmt_top; /* reset for old format files */
-#endif
+ fmt_scan (fmt, scanl, scanl_size, slwidth, dat);
if (bodycomp)
bodycomp->c_text = saved_c_text;
if (noisy)
fputs (scanl, stdout);
- FINDCOMP (cptr, "encrypted");
+ cptr = fmt_findcomp ("encrypted");
encrypted = cptr && cptr->c_text;
/* return dynamically allocated buffers to pool */
}
-/*
- * Cheat: we are loaded with adrparse, which wants a routine called
- * OfficialName(). We call adrparse:getm() with the correct arguments
- * to prevent OfficialName() from being called. Hence, the following
- * is to keep the loader happy.
- */
-char *
-OfficialName (char *name)
-{
- return name;
-}
-
-
static int
mh_fputs(char *s, FILE *stream)
{
projects / mmh / blobdiff