The Devil's Doorknob BBS Capture (1996-2003)

home *** CD-ROM | disk | FTP | other *** search

/ The Devil's Doorknob BBS Capture (1996-2003) / devilsdoorknobbbscapture1996-2003.iso / Dloads / OTHERUTI / MASTER-2.ZIP / MCOMM551 / MCOMM551.LZH / UPDATE.DOC < prev next >

Wrap

Text File | 1991-11-28 | 10KB | 208 lines

MCOMM 5.50 UPDATE NOTES Version 5.50 adds RTS flow control for the receive side. Previous versions supported monitoring CTS to insure the device being sent to was not overran but did not lower RTS if the PC was being overran. The function 'async_msrflow' has been modified to support this option. For more information, see the updated documentation on 'async_msrflow' in ASYNC.DOC. This change also necessitated a change in the ASYNC structure making it one byte longer. BE SURE TO USE THE RIGHT HEADER FILE AND RECOMPILE ANY CODE THAT USES COMM.H or COMM.HPP. Added a Zmodem demo program called TXZM. The shell code and linkable libraries are not part of the shareware version of MCOMM. Complete source comes with the registered version. This is primarily a demo with limited documentation and limited support. It also happens to be very fast and contains a lot of extras. It MAY NOT be distributed except as stated in its documentation. MCOMM 5.41 UPDATE NOTES Version 5.41 contains a few minor changes and bug fixes. These are: 1) The 'volatile' keyword was added to ASYNC structure members that are modified by the interrupt handler. This corrects problems with optimizing compilers (such as MSC 6.0) that load variables into registers. Many of the macro implemented functions test variables that are modified by the interrupt handler. If the compiler, in optimizing a loop, loads a variable into a register that is modified by the interrupt handler, it will never see the variable change. The 'volatile' key word instructs the compiler to not assume the variable remains constant. Thanks to Chris Fanning for pointing this out to me. 2) The ANSI cursor position sequence did not recognize the window relative ('v_wrel') mode of operation. If an ANSI cursor sequence was received, it was always treated relative to 0,0. It now treats the current window's top left corner as 0,0 if 'v_wrel' is non-zero. Thanks to David Speer for this information. 3) When a line was exactly the width of the screen, any following CR/LFs were ignored until a non CR/LF character was received. This is now fixed. 4) The definition of SHFT_TAB in KEYS.H has been wrong for the last several versions. Thanks to Cleo Yarber for pointing this out to me several times. MCOMM 5.40 UPDATE NOTES The 5.40 version of MCOMM corrects two problems: 1) Some UARTs miss transmit interrupts when they are ran in full duplex mode. Code was added to detect missed interrupts and correct the situation. This condition would cause previous versions of MCOMM to quit transmitting until the transmit buffer was flushed or the port closed and reopened. This bug only showed up when operating in full duplex mode and only with some UARTs. Version 5.40 has been tested by running several megabytes of data through 8250, 16450, and 16550 UARTs in full duplex mode and never fails to detect a missed interrupt and correct for it. 2) The loop that loads bytes into the transmit FIFOs (or the transmit hold register if the UART isn't a 16550) updates the transmit buffer output pointer in a register. The loop is in the transmit interrupt handler and in previous versions of MCOMM5 the transmit interrupt handler code was called by the routine that started up the transmitter. If a transmit interrupt occurred while the transmit start up code was within this loop, the buffer output pointer could be corrupted. The result was scrambled output. This only happened if the transmitter was sent a character, allowed to empty, and then immediately sent another character. If only one character was sent at a time slowly, as in typing, or if the buffer was loaded up rapidly as in a program loop, everything was OK. If characters were sent in a program loop that put a "small" delay between characters (character pacing), it often failed. This was corrected by changing the transmit initiate routine so that it toggles the transmit interrupt enable bit in the UART's interrupt enable register to initiate transmit interrupts rather than calling the interrupt handler directly. ADDITIONS: An option was added to the async_open function to force FIFO mode off. Previously you could OR the ComBase parameter with 0x8000 and the code skipped over the check for a 16550 UART. If it was there and on, it stayed on. If it was there and off, it stayed off. You can now OR the ComBase parameter with 0x4000 to force the FIFOs to be off. If you don't OR anything with the ComBase parameter the FIFOs will be enabled if a 16550 is present. To deal with problems with the Western Digital 16550, the ASYNC port structure is now one element longer. The new element is called 'IERVal'. Also options were added to allow you to open a port without enabling MSR interrupts, LSR interrupts, or both. See the notes on the WD16550 in ASYNC.DOC for more information. Conversion LIBs and a batch file were added to support Zortech C / C++. C++ header files and a C++ demo program for either Turbo C++ or Zortech C++ were added. Two new macros are now defined in COMM.H. They are: async_txcnt(ASYNC *port) -- returns number of bytes left in the transmit buffer still to be transmitted. Use async_txempty if you are waiting for 0 byte condition -- it's more efficient. async_rxfree(ASYNC *port) - returns bytes of free space left in the receive buffer. VIDEO FUNCTIONS: One minor bug was fixed. The rd_scrnd function returned an incorrect pointer if the function was called with a request to read 0 bytes. The result string was correct but the return value was wrong. It now returns the correct pointer. A new global variable was added to control whether a displayed backspace character was destructive or non-destructive. Previously it was always destructive and that still is the default. The variable is v_bksp. Setting v_bksp to 0 will cause the backspace character to be non-destructive. MCOMM 5.20 UPGRADE NOTES Major changes in this version of MCOMM are support for the 16550 UART's FIFO mode of operation, changes in the header files, new time-out functions, and updated documentation. 16550 UART SUPPORT: The async code now supports the 16550 UART's FIFO mode of operation. Key features are: 1) The async_open function automatically detects a 16550 and enables its FIFOs (can be optionally defeated). 2) Functions are provided to set the receiver FIFO trigger level and to control the maximum number of bytes to load into the transmit FIFO at one time. 3) The interrupt handler has been re-written to make most efficient use of the FIFO registers (they process multiple bytes per interrupt). ASYNC.DOC contains full descriptions of the new functions and changes made to the old functions. It also has some information on how to set the receiver FIFO trigger level and the 'maximum number of bytes to transmit to the FIFOs at one time' level to achieve the best results. HEADER FILE CHANGES: Due to stricter type checking of newer compilers, several functions that formerly were declared as taking 'char' argu- ments are now declared as taking 'int' arguments. This prevents getting an excessive number of warnings when pass- ing constants to these functions. The 'cdecl' keyword was added to the function prototypes to prevent problems when compiling with Microsoft's 'fastcall' option enabled. This keyword, as well as the 'near' and 'far' keywords, may not be compatible with your compiler. See MUSTREAD.DOC for more information. All the video functions are now prototyped in ANSIDRV.H. Before half of them were in ANSIDRV.H and the rest were in EXTRA.H. TIME OUT FUNCTIONS: The time-out functions have been changed to a new version that takes a pointer to a 'long' rather than a pointer to a TIMEOUT structure. The new functions are faster and work with longer timeouts than the old versions. Converting to the new style consists mainly of changing all occurrences of 'struct TIMEOUT' in your code to 'long'.