/* Copyright (C) 2005 J.F.Dockes
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the
* Free Software Foundation, Inc.,
* 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*/
#ifndef TEST_MH_MBOX
#include "autoconfig.h"
#include <stdio.h>
#include <fcntl.h>
#include <errno.h>
#include <unistd.h>
#include <time.h>
#include <regex.h>
#include <sys/stat.h>
#include <errno.h>
#include <cstring>
#include <map>
#include "cstr.h"
#include "mimehandler.h"
#include "debuglog.h"
#include "readfile.h"
#include "mh_mbox.h"
#include "smallut.h"
#include "rclconfig.h"
#include "md5ut.h"
#include "conftree.h"
#include "ptmutex.h"
using namespace std;
// Define maximum message size for safety. 100MB would seem reasonable
static const unsigned int max_mbox_member_size = 100 * 1024 * 1024;
class FpKeeper {
public:
FpKeeper(FILE **fpp) : m_fpp(fpp) {}
~FpKeeper()
{
if (m_fpp && *m_fpp) {
fclose(*m_fpp);
*m_fpp = 0;
}
}
private: FILE **m_fpp;
};
static PTMutexInit o_mcache_mutex;
/**
* Handles a cache for message numbers to offset translations. Permits direct
* accesses inside big folders instead of having to scan up to the right place
*
* Message offsets are saved to files stored under cfg(mboxcachedir), default
* confdir/mboxcache. Mbox files smaller than cfg(mboxcacheminmbs) are not
* cached.
* Cache files are named as the md5 of the file UDI, which is kept in
* the first block for possible collision detection. The 64 bits
* offsets for all message "From_" lines follow. The format is purely
* binary, values are not even byte-swapped to be proc-idependant.
*/
class MboxCache {
public:
typedef MimeHandlerMbox::mbhoff_type mbhoff_type;
MboxCache()
: m_ok(false), m_minfsize(0)
{
// Can't access rclconfig here, we're a static object, would
// have to make sure it's initialized.
}
~MboxCache() {}
mbhoff_type get_offset(RclConfig *config, const string& udi, int msgnum)
{
LOGDEB0(("MboxCache::get_offsets: udi [%s] msgnum %d\n", udi.c_str(),
msgnum));
if (!ok(config)) {
LOGDEB0(("MboxCache::get_offsets: init failed\n"));
return -1;
}
PTMutexLocker locker(o_mcache_mutex);
string fn = makefilename(udi);
FILE *fp = 0;
if ((fp = fopen(fn.c_str(), "r")) == 0) {
LOGDEB(("MboxCache::get_offsets: open failed, errno %d\n", errno));
return -1;
}
FpKeeper keeper(&fp);
char blk1[o_b1size];
if (fread(blk1, 1, o_b1size, fp) != o_b1size) {
LOGDEB0(("MboxCache::get_offsets: read blk1 errno %d\n", errno));
return -1;
}
ConfSimple cf(string(blk1, o_b1size));
string fudi;
if (!cf.get("udi", fudi) || fudi.compare(udi)) {
LOGINFO(("MboxCache::get_offset:badudi fn %s udi [%s], fudi [%s]\n",
fn.c_str(), udi.c_str(), fudi.c_str()));
return -1;
}
if (fseeko(fp, cacheoffset(msgnum), SEEK_SET) != 0) {
LOGDEB0(("MboxCache::get_offsets: seek %lld errno %d\n",
cacheoffset(msgnum), errno));
return -1;
}
mbhoff_type offset = -1;
int ret;
if ((ret = fread(&offset, 1, sizeof(mbhoff_type), fp))
!= sizeof(mbhoff_type)) {
LOGDEB0(("MboxCache::get_offsets: read ret %d errno %d\n",
ret, errno));
return -1;
}
LOGDEB0(("MboxCache::get_offsets: ret %lld\n", (long long)offset));
return offset;
}
// Save array of offsets for a given file, designated by Udi
void put_offsets(RclConfig *config, const string& udi, mbhoff_type fsize,
vector<mbhoff_type>& offs)
{
LOGDEB0(("MboxCache::put_offsets: %u offsets\n", offs.size()));
if (!ok(config) || !maybemakedir())
return;
if (fsize < m_minfsize)
return;
PTMutexLocker locker(o_mcache_mutex);
string fn = makefilename(udi);
FILE *fp;
if ((fp = fopen(fn.c_str(), "w")) == 0) {
LOGDEB(("MboxCache::put_offsets: fopen errno %d\n", errno));
return;
}
FpKeeper keeper(&fp);
string blk1;
blk1.append("udi=");
blk1.append(udi);
blk1.append(cstr_newline);
blk1.resize(o_b1size, 0);
if (fwrite(blk1.c_str(), 1, o_b1size, fp) != o_b1size) {
LOGDEB(("MboxCache::put_offsets: fwrite errno %d\n", errno));
return;
}
for (vector<mbhoff_type>::const_iterator it = offs.begin();
it != offs.end(); it++) {
mbhoff_type off = *it;
if (fwrite((char*)&off, 1, sizeof(mbhoff_type), fp) !=
sizeof(mbhoff_type)) {
return;
}
}
}
// Check state, possibly initialize
bool ok(RclConfig *config) {
PTMutexLocker locker(o_mcache_mutex);
if (m_minfsize == -1)
return false;
if (!m_ok) {
int minmbs = 5;
config->getConfParam("mboxcacheminmbs", &minmbs);
if (minmbs < 0) {
// minmbs set to negative to disable cache
m_minfsize = -1;
return false;
}
m_minfsize = minmbs * 1000 * 1000;
config->getConfParam("mboxcachedir", m_dir);
if (m_dir.empty())
m_dir = "mboxcache";
m_dir = path_tildexpand(m_dir);
// If not an absolute path, compute relative to config dir
if (m_dir.at(0) != '/')
m_dir = path_cat(config->getConfDir(), m_dir);
m_ok = true;
}
return m_ok;
}
private:
bool m_ok;
// Place where we store things
string m_dir;
// Don't cache smaller files. If -1, don't do anything.
mbhoff_type m_minfsize;
static const size_t o_b1size;
// Create the cache directory if it does not exist
bool maybemakedir()
{
struct stat st;
if (stat(m_dir.c_str(), &st) != 0 && mkdir(m_dir.c_str(), 0700) != 0) {
return false;
}
return true;
}
// Compute file name from udi
string makefilename(const string& udi)
{
string digest, xdigest;
MD5String(udi, digest);
MD5HexPrint(digest, xdigest);
return path_cat(m_dir, xdigest);
}
// Compute offset in cache file for the mbox offset of msgnum
mbhoff_type cacheoffset(int msgnum)
{// Msgnums are from 1
return o_b1size + (msgnum-1) * sizeof(mbhoff_type);
}
};
const size_t MboxCache::o_b1size = 1024;
static class MboxCache o_mcache;
static const string cstr_keyquirks("mhmboxquirks");
MimeHandlerMbox::~MimeHandlerMbox()
{
clear();
}
void MimeHandlerMbox::clear()
{
m_fn.erase();
if (m_vfp) {
fclose((FILE *)m_vfp);
m_vfp = 0;
}
m_msgnum = m_lineno = 0;
m_ipath.erase();
m_offsets.clear();
RecollFilter::clear();
}
bool MimeHandlerMbox::set_document_file(const string& mt, const string &fn)
{
LOGDEB(("MimeHandlerMbox::set_document_file(%s)\n", fn.c_str()));
RecollFilter::set_document_file(mt, fn);
m_fn = fn;
if (m_vfp) {
fclose((FILE *)m_vfp);
m_vfp = 0;
}
m_vfp = fopen(fn.c_str(), "r");
if (m_vfp == 0) {
LOGERR(("MimeHandlerMail::set_document_file: error opening %s\n",
fn.c_str()));
return false;
}
#if defined O_NOATIME && O_NOATIME != 0
if (fcntl(fileno((FILE *)m_vfp), F_SETFL, O_NOATIME) < 0) {
// perror("fcntl");
}
#endif
fseek((FILE *)m_vfp, 0, SEEK_END);
m_fsize = ftell((FILE*)m_vfp);
fseek((FILE*)m_vfp, 0, SEEK_SET);
m_havedoc = true;
m_offsets.clear();
m_quirks = 0;
// Check for location-based quirks:
string quirks;
if (m_config && m_config->getConfParam(cstr_keyquirks, quirks)) {
if (quirks == "tbird") {
LOGDEB(("MimeHandlerMbox: setting quirks TBIRD\n"));
m_quirks |= MBOXQUIRK_TBIRD;
}
}
// And double check for thunderbird
string tbirdmsf = fn + ".msf";
if ((m_quirks&MBOXQUIRK_TBIRD) == 0 && access(tbirdmsf.c_str(), 0) == 0) {
LOGDEB(("MimeHandlerMbox: detected unconfigured tbird mbox in %s\n",
fn.c_str()));
m_quirks |= MBOXQUIRK_TBIRD;
}
return true;
}
#define LL 1024
typedef char line_type[LL+10];
static inline void stripendnl(line_type& line, int& ll)
{
ll = strlen(line);
while (ll > 0) {
if (line[ll-1] == '\n' || line[ll-1] == '\r') {
line[ll-1] = 0;
ll--;
} else
break;
}
}
// The mbox format uses lines beginning with 'From ' as separator.
// Mailers are supposed to quote any other lines beginning with
// 'From ', turning it into '>From '. This should make it easy to detect
// message boundaries by matching a '^From ' regular expression
// Unfortunately this quoting is quite often incorrect in the real world.
//
// The rest of the format for the line is somewhat variable, but there will
// be a 4 digit year somewhere...
// The canonic format is the following, with a 24 characters date:
// From toto@tutu.com Sat Sep 30 16:44:06 2000
// This resulted into the pattern for versions up to 1.9.0:
// "^From .* [1-2][0-9][0-9][0-9]$"
//
// Some mailers add a time zone to the date, this is non-"standard",
// but happens, like in:
// From toto@truc.com Sat Sep 30 16:44:06 2000 -0400
//
// This is taken into account in the new regexp, which also matches more
// of the date format, to catch a few actual issues like
// From http://www.itu.int/newsroom/press/releases/1998/NP-2.html:
// Note that this *should* have been quoted.
//
// http://www.qmail.org/man/man5/mbox.html seems to indicate that the
// fact that From_ is normally preceded by a blank line should not be
// used, but we do it anyway (for now).
// The same source indicates that arbitrary data can follow the date field
//
// A variety of pathologic From_ lines:
// Bad date format:
// From uucp Wed May 22 11:28 GMT 1996
// Added timezone at the end (ok, part of the "any data" after the date)
// From qian2@fas.harvard.edu Sat Sep 30 16:44:06 2000 -0400
// Emacs VM botch ? Adds tz between hour and year
// From dockes Wed Feb 23 10:31:20 +0100 2005
// From dockes Fri Dec 1 20:36:39 +0100 2006
// The modified regexp gives the exact same results on the ietf mail archive
// and my own's.
// Update, 2008-08-29: some old? Thunderbird versions apparently use a date
// in "Date: " header format, like: From - Mon, 8 May 2006 10:57:32
// This was added as an alternative format. By the way it also fools "mail" and
// emacs-vm, Recoll is not alone
// Update: 2009-11-27: word after From may be quoted string: From "john bull"
static const char *frompat =
"^From[ ]+([^ ]+|\"[^\"]+\")[ ]+" // 'From (toto@tutu|"john bull") '
"[[:alpha:]]{3}[ ]+[[:alpha:]]{3}[ ]+[0-3 ][0-9][ ]+" // Fri Oct 26
"[0-2][0-9]:[0-5][0-9](:[0-5][0-9])?[ ]+" // Time, seconds optional
"([^ ]+[ ]+)?" // Optional tz
"[12][0-9][0-9][0-9]" // Year, unanchored, more data may follow
"|" // Or standard mail Date: header format
"^From[ ]+[^ ]+[ ]+" // From toto@tutu
"[[:alpha:]]{3},[ ]+[0-3]?[0-9][ ]+[[:alpha:]]{3}[ ]+" // Mon, 8 May
"[12][0-9][0-9][0-9][ ]+" // Year
"[0-2][0-9]:[0-5][0-9](:[0-5][0-9])?" // Time, secs optional
;
// Extreme thunderbird brokiness. Will sometimes use From lines
// exactly like: From ^M (From followed by space and eol). We only
// test for this if QUIRKS_TBIRD is set
static const char *miniTbirdFrom = "^From $";
static regex_t fromregex;
static regex_t minifromregex;
static bool regcompiled;
static PTMutexInit o_regex_mutex;
static void compileregexes()
{
PTMutexLocker locker(o_regex_mutex);
// As the initial test of regcompiled is unprotected the value may
// have changed while we were waiting for the lock. Test again now
// that we are alone.
if (regcompiled)
return;
regcomp(&fromregex, frompat, REG_NOSUB|REG_EXTENDED);
regcomp(&minifromregex, miniTbirdFrom, REG_NOSUB|REG_EXTENDED);
regcompiled = true;
}
bool MimeHandlerMbox::next_document()
{
if (m_vfp == 0) {
LOGERR(("MimeHandlerMbox::next_document: not open\n"));
return false;
}
if (!m_havedoc) {
return false;
}
FILE *fp = (FILE *)m_vfp;
int mtarg = 0;
if (!m_ipath.empty()) {
sscanf(m_ipath.c_str(), "%d", &mtarg);
} else if (m_forPreview) {
// Can't preview an mbox.
LOGDEB(("MimeHandlerMbox::next_document: can't preview folders!\n"));
return false;
}
LOGDEB0(("MimeHandlerMbox::next_document: fn %s, msgnum %d mtarg %d \n",
m_fn.c_str(), m_msgnum, mtarg));
if (mtarg == 0)
mtarg = -1;
if (!regcompiled) {
compileregexes();
}
// If we are called to retrieve a specific message, seek to bof
// (then scan up to the message). This is for the case where the
// same object is reused to fetch several messages (else the fp is
// just opened no need for a seek). We could also check if the
// current message number is lower than the requested one and
// avoid rereading the whole thing in this case. But I'm not sure
// we're ever used in this way (multiple retrieves on same
// object). So:
bool storeoffsets = true;
if (mtarg > 0) {
mbhoff_type off;
line_type line;
LOGDEB0(("MimeHandlerMbox::next_doc: mtarg %d m_udi[%s]\n",
mtarg, m_udi.c_str()));
if (!m_udi.empty() &&
(off = o_mcache.get_offset(m_config, m_udi, mtarg)) >= 0 &&
fseeko(fp, (off_t)off, SEEK_SET) >= 0 &&
fgets(line, LL, fp) &&
(!regexec(&fromregex, line, 0, 0, 0) ||
((m_quirks & MBOXQUIRK_TBIRD) &&
!regexec(&minifromregex, line, 0, 0, 0))) ) {
LOGDEB0(("MimeHandlerMbox: Cache: From_ Ok\n"));
fseeko(fp, (off_t)off, SEEK_SET);
m_msgnum = mtarg -1;
storeoffsets = false;
} else {
fseek(fp, 0, SEEK_SET);
m_msgnum = 0;
}
}
off_t message_end = 0;
bool iseof = false;
bool hademptyline = true;
string& msgtxt = m_metaData[cstr_dj_keycontent];
msgtxt.erase();
line_type line;
for (;;) {
message_end = ftello(fp);
if (!fgets(line, LL, fp)) {
LOGDEB2(("MimeHandlerMbox:next: eof\n"));
iseof = true;
m_msgnum++;
break;
}
m_lineno++;
int ll;
stripendnl(line, ll);
LOGDEB2(("mhmbox:next: hadempty %d lineno %d ll %d Line: [%s]\n",
hademptyline, m_lineno, ll, line));
if (hademptyline) {
if (ll > 0) {
// Non-empty line with empty line flag set, reset flag
// and check regex.
if (!(m_quirks & MBOXQUIRK_TBIRD)) {
// Tbird sometimes omits the empty line, so avoid
// resetting state (initially true) and hope for
// the best
hademptyline = false;
}
/* The 'F' compare is redundant but it improves performance
A LOT */
if (line[0] == 'F' && (
!regexec(&fromregex, line, 0, 0, 0) ||
((m_quirks & MBOXQUIRK_TBIRD) &&
!regexec(&minifromregex, line, 0, 0, 0)))
) {
LOGDEB1(("MimeHandlerMbox: msgnum %d, "
"From_ at line %d: [%s]\n", m_msgnum, m_lineno, line));
if (storeoffsets)
m_offsets.push_back(message_end);
m_msgnum++;
if ((mtarg <= 0 && m_msgnum > 1) ||
(mtarg > 0 && m_msgnum > mtarg)) {
// Got message, go do something with it
break;
}
// From_ lines are not part of messages
continue;
}
}
} else if (ll <= 0) {
hademptyline = true;
}
if (mtarg <= 0 || m_msgnum == mtarg) {
// Accumulate message lines
line[ll] = '\n';
line[ll+1] = 0;
msgtxt += line;
if (msgtxt.size() > max_mbox_member_size) {
LOGERR(("mh_mbox: huge message (more than %u MB) inside %s,"
" giving up\n", max_mbox_member_size/(1024*1024),
m_fn.c_str()));
return false;
}
}
}
LOGDEB2(("Message text length %d\n", msgtxt.size()));
LOGDEB2(("Message text: [%s]\n", msgtxt.c_str()));
char buf[20];
// m_msgnum was incremented when hitting the next From_ or eof, so the data
// is for m_msgnum - 1
sprintf(buf, "%d", m_msgnum - 1);
m_metaData[cstr_dj_keyipath] = buf;
m_metaData[cstr_dj_keymt] = "message/rfc822";
if (iseof) {
LOGDEB2(("MimeHandlerMbox::next: eof hit\n"));
m_havedoc = false;
if (!m_udi.empty() && storeoffsets) {
o_mcache.put_offsets(m_config, m_udi, m_fsize, m_offsets);
}
}
return msgtxt.empty() ? false : true;
}
#else // Test driver ->
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <string.h>
#include <iostream>
#include <string>
using namespace std;
#include "rclconfig.h"
#include "rclinit.h"
#include "cstr.h"
#include "mh_mbox.h"
static char *thisprog;
static char usage [] =
"Test driver for mbox walking function\n"
"mh_mbox [-m num] mboxfile\n"
" \n\n"
;
static void
Usage(void)
{
fprintf(stderr, "%s: usage:\n%s", thisprog, usage);
exit(1);
}
static RclConfig *config;
static int op_flags;
#define OPT_MOINS 0x1
#define OPT_m 0x2
//#define OPT_t 0x4
int main(int argc, char **argv)
{
string msgnum;
thisprog = argv[0];
argc--; argv++;
while (argc > 0 && **argv == '-') {
(*argv)++;
if (!(**argv))
/* Cas du "adb - core" */
Usage();
while (**argv)
switch (*(*argv)++) {
case 'm': op_flags |= OPT_m; if (argc < 2) Usage();
msgnum = *(++argv);
argc--;
goto b1;
// case 't': op_flags |= OPT_t;break;
default: Usage(); break;
}
b1: argc--; argv++;
}
if (argc != 1)
Usage();
string filename = *argv++;argc--;
string reason;
config = recollinit(RclInitFlags(0), 0, 0, reason, 0);
if (config == 0) {
cerr << "init failed " << reason << endl;
exit(1);
}
config->setKeyDir(path_getfather(filename));
MimeHandlerMbox mh(config, "some_id");
if (!mh.set_document_file("text/x-mail", filename)) {
cerr << "set_document_file failed" << endl;
exit(1);
}
if (!msgnum.empty()) {
mh.skip_to_document(msgnum);
if (!mh.next_document()) {
cerr << "next_document failed after skipping to " << msgnum << endl;
exit(1);
}
map<string, string>::const_iterator it =
mh.get_meta_data().find(cstr_dj_keycontent);
int size;
if (it == mh.get_meta_data().end()) {
size = -1;
cerr << "No content!!" << endl;
exit(1);
}
cout << "Doc " << msgnum << ":" << endl;
cout << it->second << endl;
exit(0);
}
int docnt = 0;
while (mh.has_documents()) {
if (!mh.next_document()) {
cerr << "next_document failed" << endl;
exit(1);
}
docnt++;
map<string, string>::const_iterator it =
mh.get_meta_data().find(cstr_dj_keycontent);
int size;
if (it == mh.get_meta_data().end()) {
size = -1;
} else {
size = it->second.length();
}
cout << "Doc " << docnt << " size " << size << endl;
}
cout << docnt << " documents found in " << filename << endl;
exit(0);
}
#endif // TEST_MH_MBOX