Beijing Paradise BBS Backup

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

/ Beijing Paradise BBS Backup / PARADISE.ISO / software / BBSDOORW / BNU189G.ZIP / BNU.REV < prev next >

Wrap

Text File | 1993-07-07 | 22.7 KB | 533 lines

-------------------------- Changes since 1.70 release -------------------------- ** MAKE SURE YOU REMOVE ALL REFERENCES TO THE /M SWITCH IN ALL YOUR EXISTING COMMAND LINES - IT IS NO LONGER SUPPORTED! BNU NO LONGER ISSUES A MESSAGE WHEN IT IS INITIALISED ---- v1.89g 09 Oct 1990 - Fixed a stupid bug in BNU's setup code on installation which crashed some systems when the FOSSIL was initialised... ---- v1.89f 08 Oct 1990 - An incorrect message (no ports available) will no longer display if an attempt is made to init the driver when it isn't resident - Corrected problem with the INS8250-B UART within the interrupt service routine; apparently the line status register is not reliable on these UARTS. - Fixed bug in baud rate setting on install when FOSSIL is locked on a port (it would skip setting the baud rate immediately) ---- v1.89e 19 Aug 1990 - Fixed a typo in the buffer init routine which caused a change buffer size request (/T /R) to hang - Fixed a problem in PC-MOS/386 reboot function ---- v1.89d 15 Aug 1990 - Fixed problems (I hope) in transmitter restart - Boot under PC-MOS/386 with /B switch fixed - Fixed watchdog call; was previously not installing itself correctly within the timer tick - A *lot* more work done on the transmitter logic; this should solve problems with the transmitter stop; BNU now disables TX interrupts when the transmitter is idle (thanks to Dave Hatch :-)) The ISR has undergone a significant overhaul. - Removed a number of kludges I had previously used to work around the stuck transmitter problem - Fixed a subtle problem in the flow control call which I _think_ was causing problems with the chat function under Maximus on some hardware - Removed the "disable init message" switch from BNU.COM/.SYS and BNUHLP. BNU no longer outputs a message when initialised - that code has been moved to the non-resident portion, since I think it's more beneficial to consume less memory than output copyright messages :-) Init message status removed from /S display. - Debug panel display now includes interrupt enable register readings - Changed meaning of some values in the debug panel in line with changes to the transmitter logic; hopefully I'll get time to do a developer's doc for the next release which will explain them ** Please watch this one carefully: some significant things have changed and it requires careful testing - especially under BBS software which I haven't tested extensively at all. ---- v1.89c 05 Aug 1990 - Added PORTTEST (COM port diagnostic) to the package - Fixed a possible transmitter overrun problem with 16550AFN's by shortening the maximum transmitter FIFO queue by 1. - Transmitter queue routine changed slightly; should be faster now (though I doubt if you'll see any real difference in transfer rates!) - TM-DD (hopefully) fixed. - Added /B1 switch to disable DOS busy detection. If you found the last TM a little slower than previously, and you don't run programs which call FOSSIL in background, then try this to see if it improves things. I'm curious. :-) - Couple of other minor mods not worth mentioning. ---- v1.89ß 22 Jul 1990 - Allowed dynamic changing of watchdog delay period without having to disable it first; this has the effect of BNU not considering multiple trys at watchdog enable an error (ie. no big deal..) - Fixed (again! argh!) reentrancy problem in FOSSIL init call; this produced strange problems when two ports initialised at once - Fixed baud rate lock at 9600 in the .COM version; was actually locking at 75 baud in error (which is pretty handy...) - BNU now returns 0xffff when an errant call is made; should be a temporary workaround until something in Opus/CHAT is fixed (PLEASE TEST!) - BNU again releases its environment; let me know of any problems on this one - Calling port status which returns the static image of the line status register now masks out receiver overrun errors for any subsequent calls until it again hits that condition (character received when FOSSIL RX buffer full) - Added more CPU register reporting to the diags window to help diagnose errors; particularly errant calls - Several modifications to support FDCD in BNUHLP - Found the (argh!) bug in TM which wasn't checking the correct flag in DOS to see whether to time slice or not; Should be safe now to use TM with DoorWay and D'Bridge (without the QEMM exception errors) which make calls to FOSSIL calls in "background" (PLEASE TEST THIS - I NEED FEEDBACK TO MAKE SURE IT'S WORKING CORRECTLY!) - PC-MOS/386 "in-MOS" detection code reinstated (same problem as above) - Added TM-DD to supports time slicing under DoubleDOS - CHKCRC provides a check on all file CRC's. (see CRC.LST for distribution values) ---- v1.88ß 27 May 1990 - Removed shared-IRQ support :-( (temporary until 1.90 release) - Fixed bug in "lock baud rate" switch (/L) in BNUHLP which set incorrect bits - Reinstated modem error reporting support in diags window - 'Snoop' added to BNU package for aiding in diagnosis and allowing a 'snapshot' ---- v1.87α 19 Apr 1990 - More revisions to BNU's interrupt service; some fairly extreme optimisations added, some "debugging" options eliminated as a result (the debug window no longer logs modem status changes) - BNU now clears line status errors on each status call (solves the problem with Opus/framing errors). - Fixed a problem in "transmit flush" which caused apparent 'lockups' with some UART configurations ---- v1.86ß 17 Apr 1990 - Removed requirement for / or - preceeding options in TM, to be consistent with BNU and BNUHLP - Significant improvement to ISR; BNU should now work *much* better in 'tough' conditions; ie. high baud rate and multi line. - Fixed a flow control bug which caused the transmitter to stop during a file transfer whenever the receiver's buffer became full (usually restarted by timeout). - Modified the way the transmitter works, completely removing the requirement for disabling interrupts when working with 16550 FIFO's. ---- v1.85ß 08 Apr 1990 - Removed some debug code (might have been causing problems) - TM-PM revised; removed INMOS detection code, which caused task slicing not to work (apparently MOS switches to its internal stack on a software interrupt? Hmmm) - Corrected two problems in port parameter locking using the /L switch in BNUHLP - Corrected the status display of locked baud rates and port parameters. ---- v1.85α - Optimised some ISR code for better efficiency/speed, some further rearrangement to implement shared IRQ's - Fixed 9600 locked baud rate (was locking at 7200) - Implemented shared IRQ's - Semaphore used to ensure two ports not colliding in init function ---- v1.84ß 07 Mar 1990 - Remote CTS timeout strategy eliminated (this was a kludge around previously existing problems which have since been solved), so /Y is now defunct - Status /S switch now displays carrier watch status - Automatically adapts boot procedure for PC-MOS/386 when detected, reboot should also work more reliably under DESQview/386 (any feedback appreciated) - To maintain BNU's re-entrancy from within DOS, TM now checks that DOS is not busy prior doing find task number calls and time slicing (I don't think this is necessary, but not knowing for sure I thought I'd make sure it wouldn't cause any problems) - BNUHLP now supports status display; handshaking force on/off also added to status in both BNUHLP and BNU.COM - Bug fixed in .SYS version which could cause problems when BNUHLP was run (some variables were inadvertantly being overwritten) - Port locking/unlocking on-the-fly added to BNUHLP - The '/' or '-' before command line arguments is now optional (earth shattering, huh?) - Added 'all' option to handshaking mask control switch /H, and squashed a wee bug there in BNUHLP's version. You can now specify something like: /HF0+,A This allows switching on/off ALL flow control flags (XON/XOFF local, CTS/RTS local and control remote using XON/XOFF), and is equivalent to /HF0+,CRX but is a little easier to understand - Additional test for receiver ready added to ISR, which will hopefully further ensure reduction in the incidence of receive character loss - Up to 16 ports are now supported for concurrent use; this is BNU's maximum port configuration anyway. - New BNUPORT added to this package - earlier versions will not work and are incompatible; patch table files from earlier versions will be read by this version and automagically converted to the new more compact format - And especially for those who don't read release notes <grin> the /M option is now disabled in all beta versions so that the init message WILL be displayed with the new "THIS IS A BETA" message. If you ask about it, you're in for some public embarassment about not RTFM'ing! ;-) ---- v1.83α 14 Feb 1990 - Experimental version(s) only - /? is now the ONLY help switch; /H reserved for ... - Added setting handshake mask ON and OFF to force or mask out local and remote handshaking. Format of this switch is: /H{F|N}<port>{+|-},[C|X|R] Examples are: /HF0+,XR Forces ON, xon/xoff handshaking for both remote and local so that despite what the application sets, XON/XOFF is always enabled. /HF0-,X Switchs off forcing on xon/xoff handshaking when XON/XOFF is received from the remote system; this resets any switch which forced 'X' on (ie. reverses /FF0+,X). /HN0+,C Forces CTS/RTS handshaking off on COM1. /HN1-,CX Switches forcing both CTS/RTS and XON/XOFF handshaking off for COM2, reversing a previous /FN1+,CX. This switch should ONLY be used in special circumstances, where you wish to achieve a SPECIFIC PURPOSE. WARNING: DO NOT USE THIS SWITCH UNLESS YOU UNDERSTAND IT AND KNOW EXACTLY THE EFFECT YOU WISH TO ACHIEVE!! Some examples of use are where your modem does not do true CTS/RTS handshaking or does it incorrectly. For example, making CTS follow carrier detect, stopping any commands being sent to the modem after carrier loss (and the application does not provide an option to not use CTS/RTS flow control); similarly where you use a three-wire direct connect wire between machines without a CTS line; where your modem cable does not support a CTS wire; where a BBS uses block read/writes but does not enable XON/XOFF flow control to allow ^S and ^Q to control screen start and stop. - A few bug fixes/corrections: . Setting baud rate when FOSSIL was not initialised . Fixed serious problem in transmit/receive block functions where interrupts would be disabled for too long due to a missing STI instruction in a common subroutine . Found error in ISR which prevented disabling of CTS/RTS flow control (CTS/RTS flow control was respected regardless) - Fixed reboot to work correctly under DESQview/386 ---- v1.82ß 23 Jan 1990 - Finally a released beta ... - CTS timeout now defaults to OFF, and must be enabled specifically with /Yn (where n is in ticks). Don't use it if you aren't having problems since it will definitely create some for you. ---- v1.81α 21 Jan 1990 - Inserted 'hooks' for an external data capture buffer and added 'reserved' function to integration for a program to be released - Modified 'break' handling to force a fixed interval break in transmit buffer flush, which can be switched on and off for each port individually. the new syntax is: /G<port>[+|-] where port is 0=COM1, 1=COM2 etc. and '+' switches this break handling on, and '-' switches it off - Modified (a little) the way communications interrupts are handled; SEOI rearms 8259A at commencement of IRQ handler to ensure no loss of interrupts is possible - Up to 8 ports are now supported concurrently (was 4) - Prepared for chaining of ports to the same IRQ (not supported in this version yet!) - Modified debug window for more concise and relevant info - Limited test releases ---- v1.80Γ 15 Jan 1990 - Corrected reboot code (typo) causing code not to work - Corrections to break signal logic - Further modifications to reboot code: it should work far more reliably now under DV/386 and PC-MOS/386. - Found suspicious error in maintenance of an internal status flag - Fixed error in the "BIOS compatible" code for setting baud rates: this indirectly affected locking baud rates as well - Limited release --- v1.79Σ 04 Jan 1990 * Emergency bug fix * - Fixed errors in new CTS signal recovery logic --- v1.80α 07 Jan 1990 - Modified .SYS version / BNUHLP interraction - BNUHLP now works in "non-verbose" mode when the /M- option has been set when installed or via BNUHLP itself - Fixed bug in BNUHLP not setting driver options correctly - Modified /O switch to default to '+' (enable compatibility mode) in .COM version and BNUHLP --- v1.80α 09 Jan 1990 - Removed (hopefully) braindamage from the CTS timeout routine, so now it should be more reliable and consistent - Note that /Y<ticks> allows setting of this timeout (this has been present, but 'undocumented' since adding this logic) - Internal version only --- v1.80ß 11 Jan 1990 - Support for flushing modem's transmit buffer on transmit purge call by sending break signal for a specified period: the time the break signal up is configurable (and this option is enabled) using the /G<ticks> switch (typical setting is /G5 for approx 0.25 seconds). This seems to provide faster response to 'hotkeyed' applications WARNING: You would be advised to test this carefully; it's use will normally require adjustment for your modem's settings to enable the feature to "purge modem buffer on receiving break from DTE (your computer)". THIS OPTION SHOULD ONLY BE USED WHERE YOUR MODEM HAS AN OPTION TO PREVENT SENDING 'BREAK' TO THE REMOTE: CURRENTLY - SO FAR AS I AM AWARE - ONLY HST'S OFFER THIS OPTION. MAKE SURE &Y0 IS ENABLED! - CTS/RTS override strategy now sends a break if the /G option is enabled to help the modem to recover from braindamage (use carefully) - Added /G to BNU.COM BNUHLP.COM and BNU.SYS. - Experimental version with limited release --- v1.79π 03 Jan 1990 - Changed some internals in the way locked baud rates work - Corrected problem in .sys version when parsing command line for a locked baud rate - Significant cleanup of .SYS version interface and facilities - removed /B switch (redundant) from .SYS version - added basic funtional BNUHLP.COM (see notes below) - added functional task manager driver for DESQview and PC-MOS/386 - added CTS signal low timeout to prevent blocked transmitter states when a modem is in error. The default timeout is 10 seconds (ie if CTS from the modem stays low for more than 10 seconds, BNU will ignore this until the next modem status change). --- v1.79Γ 01 Jan 1990 - Modified uninstall routine to allow for TM's install process and resident detection. - Small change to reboot code to cope with PC-MOS/386 v3.0 - Added hook for "get counter" from task manager (mainly for debugging), and modified diags window to indicate UART type in upper byte of FIFO size - Substantial modifications to port baud rate & parameters setting, fixing "bug" in the baud rate returned in "fossil data" call 1bh. This call will now return the ACTUAL baud rate/parameters (it previously returned incorrect data). - Small modifications to port init to force modem status change and kick interrupts... - Bios version of baud rate set no longer turns interrupts off, but preserves their state --- v1.79ß 23 Dec 1989 - Fixed bug in setting baud rate on install, port was not being immediately set if not COM1 - Forced watchdog to lower DTR prior reboot for both better security and to prevent autoanswer modems from answering until the application had fully restarted - CHANGED /B SWITCH *** NOTICE *** /Bn old syntax is no longer supported: BNU should now reboot correctly in all cases under DESQview/386 and on most PC-compatible hardware. The new syntax is now /Bn, where n = 0 is cold boot, n = 1 is warm boot - /I* will now deinit ALL active communications ports - Limited release to specifc sites to ensure specific problems have been solved... --- v1.78ß 16 Dec 1989 - Internal version only - Solved problem with locking baud rates on anything but COM1; this interferred with other communications software running on COM1, such as mouse drivers, etc. - Enabled switchable support for XOFF/ANY mode (any character received unblocks an XOFF'ed transmitter instead of just XON), the new command line switch is /X+ (or just /X) to enable XOFF/ANY and /X- to use the default XOFF/XON protocol - Added XOFF protocol setting to status display - Closed stdout & stderr on install to prevent DOS allocating a file handle on redirected stdout/stderr --- v1.77ß 03 Dec 1989 ** Back to beta versions again! - Improved transmitter logic handling in character IO functions - Changed and improved method of FOSSIL controlled transmitter block flag in FOSSIL function 10h - Opus (which uses this extensively) should be much improved and transmitter interrupt handling more efficient - Found and squashed "the" bug introduced in 1.73α causing hanging occasionally and memory allocation problems under PC-MOS/386 :-) - Autodetect for the older 16550 (no AN) now installed; /Z2 will be default for these machines unless a /Z switch is specified - /Z with no digit now restores BNU to it's default autodetect FIFO state - Modified /Z1 to be now 16TX by 08RX to optimise for the older style 16550 (this used to be BNU's default); this allows for interrupt latency on the receive side (the usual requirement for using the /Z switch) without sacrificing transmit buffer size!). - Corrected a problem in port lock command line parsing, where the specification of ,nps with the baud rate would not be effective or would not set the port correctly --- v1.76α 29 Nov 1989 - Internal test version only; testing TX state checking - eliminated most of the new logic since the problem it was meant to solve ended up being hardware related... :-( --- v1.75α 25 Nov 1989 - Placed more transmitter state checks in character TX, TX no wait and TX flush to increase TX turnaround and decrease chance of a 'stuck' transmitter. - This version was distributed to specific sites only. --- v1.74α 18 Nov 1989 - Placed further transmitter state checks in character TX, TX no wait and TX flush to increase TX turnaround and decrease chance of a 'stuck' transmitter. - Enabled a timeout delay before reboot using the carrier watchdog. The new command line parameter (compatible with old) is: BNU /W<port>+|-[,<ticks>] Example, to enable carrier watchdog on port 1 (COM2) to reboot approx 4 seconds after carrier loss, the following would be executed: BNU /W1+,72 The timeout value is expressed in "ticks", where each timer tick is equal to approx 18.2 per second. A 1 minute delay requires a delay factor of about 1100 ticks (there are approx 65520 ticks/hour). - Stopped BNU "complaining" if a /P command line switch duplicated the number of ports as previously set on install (saves having a separate install line if only one command line is ever used in a batch file). --- v1.73α 13 Nov 1989 - Changed to specific EOI in interrupt handler instead of non- specific EOI, - BNU now (again) supports the upper 8 interrupts on a 16 bit bus, and sends a specific EOI to the master 8259A on IRQ 2 if an upper (IRQ 8-15) interrupt is used, - Some redundant code has been eliminated from the ISR (saved a few more clock cycles, anyway), - BNU's internal CTS/RTS signal is now set on port init when CTS is found low (previously it was left clear and not set until the first MSI occured), - BNU now uses INT 1cH vector for timer interrupt functions, - The timer hook now re-enables interrupt processing and rearms the timer interrupt before calling 'hooked' functions. As a result, hooked functions are now more 'friendly' and forgiving on communications - for example the diags module should work better in all cases and interfere less (if at all) with comms, - Added view of the FIFO register to the diagnostics panel: this reads according to the current FIFO trigger value: C1 16550AN or better installed 81 16550 (old series) installed 01 FIFO disabled or 16550 not installed - BNU now aborts if no hardware COM ports are located - Forced port 'autosense' routine to use resident driver's version of port setup table (since it installs ok), - Status call now includes some code to check the state of the 'transmitter' to ensure that transmission is not temporarily blocked for some reason hopefully increasing TX turnaround. * NOTE: BNUPORT from earlier versions is incompatible with releases after 1.73α due to changes in EOI handling! Upgrade to BNUPORT version 1.80 (distributed via usual channels). --- v1.72ß 10 Nov 1989 - Minor changes to port detection routine to 'fix' BIOS data area correctly for invalid/unknown port addresses - Additions to status display: number of physical ports detected during diagnostics and default FIFO trigger level - Arrgh! Found some 'unfixed' code for the now variable 16550 FIFO sizes, RX/TX buffer purge was resetting the RX buffer to the old 'hard coded' 8 byte level - Fixed bug in error message display on the /Q option when FOSSIL vector required recapture. --- v1.71ß 04 Nov 1989 - Fixed 'tabs' problem with 'fast' ANSI option - Forced a single EOI on each call (experimental); Note: this is a temporary 'kludge' for experimental purposes, and means that IRQ's on the FIRST 8259A (that is IRQ 0-7) are currently supported - Forced additional loopback in ISR after servicing stray THRE's (TX holding register empty) caused during servicing of MSI's (modem status change) or received characters. This should ensure absolutely that no stray interrupts get lost while in the ISR. ~~~~~~~~~~~~~~~~~~~~~~~~~~