argument.

smb_hallochdr() and smb_hallocdat() for calculating the required padding for
proper block alignment.

Using BOOL macros instead of int's and numeric constants.

message (in memory) as already hashed with smb_hashmsg(), preventing redundant
hashing by smb_addmsghdr().
Using SMB_SUCCESS instead of 0.
Fixed a case where smb_addmsghdr() could return without calling
smb_unlocksmbhdr() if the header was an invalid header length (too large).
smb_addmsghdr() now hashes the header fields if the message has not already
been hashed with smb_hashmsg() before allocating header blocks and
will return a failure (SMB_DUPE_MSG) with detail in smb.last_error if a
duplicate hash was found - new feature of smb_hashmsg().

smb_addmsghdr() calls smb_hashmsg() again (with update=TRUE).

Created convenience functions and macros for fast message index or header
look-ups based on hashed header fields (i.e. message-IDs).

(mark) to indicate whether to mark all found hashes (or stop on first found).

mark found hashes and only write the not-found hashes to disk, rather than not
writing *any* hashes (when any match) because the message is a "dupe".
When update is false, the behavior is the same as before.

(using smb_close_fp).

Created smb_hashstr() to create hashes of ASCIIZ data.

New smb.hash file uses multiple hashing methods (CRC-16, CRC-32 and MD5) to
calculate hashes of various "distinguishing marks" (e.g. message ID and body
text) to create a permanent history of message "fingerprints" that can be used
for duplicate message detection or fast location of messages based on one
of these "distinguishing marks". For example, reading each message header
in a message base to find a message with a specific message-ID (i.e.
how get_msg_by_id() currently works) significantly slows down message
importing with auto-thread linkage (a newly added feature). This hash file
will be used to speed this up considerably as well as improve duplicate
message detection.

This feature is supported with the following new smblib functions:
smb_findhash, smb_hashmsg, smb_hash, smb_msghashes, and smb_addhashes.

smb_addmsghdr() currently calls smb_hashmsg(), but this will probably change
with yet another new function call (smb_addmsg?) which will have access to the
body text (a very useful hash for duplicate message detection).

Existing SMB message bases will not have an existing .hash file (chksmb will
"missing hash" errors), but fixsmb can be used to create the .hash file for
existing message bases. Other than the chksmb errors, there are no real
problems with missing hashes, accept for the potential for failed auto-thread
linkage when importing replies to existing messages. Over time, this will fix
itself as new messages are imported and their hashes are stored automatically.

Projects using smblib now must link with crc16.c, crc32.c, and md5.c.

Other changes:
Created GETMSGTXT_TAIL_ONLY convenience macro.
smb_open/close_ha/da functions are now just macros to the new smb_open_fp and
smb_close_fp functions. These functions were 99% redundant.
New macros smb_open/close_hash also point to smb_open/close_fp.
More use of SAFEPRINTF macros and safe_snprintf().

header.

Increased smblib version to 2.31.

fread/fwrite does not set errno, use ferror() to get error value instead.
Using sizeof() instead of numeric constant (2) in several places.
More detail in some last_error strings.

since this value is needed by smb_putmsghdr() to detect an illegal header
length increase (exceeds allocated space for header) - the header length is now
calculated each time for each call to smb_hfield/hfield_append().

for additional tracking information.
Created DLL-exported msg_client_hfields() in postmsg.cpp as a common function
to add these client tracking header fields.

used to trace message origins for security reasons.

smb_read() will now retry the read operation if the read fails with an errno of
DEADLOCK (occurs on Unix if a Win32 system has the bytes already locked).

Created smb_getmsgtxtlen() to calculate total message body text length,
including comment header fields.

offset (which it is not), causing most message index "gets" by offset to fail.

smb_hfield_append_str() used for creating and appending ASCIIZ header fields.
These functions handle the string-length calculation automatically.

hfield type specified in string (or UNKNOWN, if not found). Returns parsed
number if string contains a number.
smb_hfieldtype() now returns mixed-case (easier to read) versions of hfield
type names.

Created new functions: smb_lock, smb_unlock, and smb_islocked - used to lock
a message base during maintenance/packing.
New smb error constant: SMB_ERR_DELETE.
Created anonymous union members in smbmsg_t for future filebase support.
Increased SMBLIB_VERSION to 2.30.

smb_incmsg() is now smb_incmsg_dfields(), which now accepts a number of header
references argument (refs).
smb_incmsg() is now just a macro that calls smb_incmsg_dfields() with 1
for the number of references.

Created SMB_ALL_REFS macro, used to free all references to data block.

smb_getmsgidx() (no longer tries to read past EOF).

the length is 0.
smb_addcrc() now verifies the file is evenly divisible by 4 bytes (the size of
a 32-bit CRC record).

required header fields (to/from/subj) is missing. Instead, it just points the
convenience pointer(s) for the missing field(s) to a static null string.
Increased version to 2.22.

more error checking.
smb_*allocdat() routines now verify that the return offset will not be >2GB.