readme.fasi

一个通讯程序源码

FAS/i is a special-purpose, unsupported version of FAS 2.08 for
those who need to have non-portable, but extended access to their tty driver.

Reasons to use:

1. You get cumulative statistics on such things as receievd and
transmitted characters, modem signals and device errors.  As an
example of the information available, the ecu interactive 'fasi'
command produces:

base address: 0218 irq=3 device is 16550
MSR=*CTS*DSR*   MCR=*DTR*RTS*OUT2*
LCR=*8db*1sb*NOPAR*   IER=*RDAV*TBMT*LS*MS*
recv ring cnt=0  xmit ring cnt=0  xmit fifo size=16
characters received    =        3097
characters transmitted =       22407
modem status events    =         137
overrun errors=0  framing errors=0  parity errors=0
rings detected=0  breaks detected=0
xmtr flow off XON/XOFF=0  RTS/CTS=31
rcvr flow off XON/XOFF=0  RTS/CTS=0
driver:  'FAS/i 2.08.0'
space.c: 'FAS/i 2.08:{1,4,03f8-03ff,COM1},{8,3,0210-024f,DIGI-PC8}'

2. There are no other reasons to use.

Reasons NOT to use:

1.  It is not supported. It is released configured for one COM1 (irq 4)
and one 8-port Digiboard PC-8.  I will help with any issue that I can,
but I will be very uninterested in answering questions like "How do I
get FAS/i to work with my PacificRim ModemBlaster 4800?"

2. It is less efficient than FAS.  Statistics take CPU cycles to accumulate.
Also, Uwe has done an indescribably superb job of optimizing the driver
for efficiency.  I have not analyzed the effect my changes have made to
the micromanagment of emitted code Uwe did, but it cannot have but harmed.

3. The driver is non-standard.  It barks in the face of most of
what I look for in well-produced software.

4. Uwe will continue to work magic and this driver is unlikely to
inherit it.

5. FAS nor FAS/i appear to support DOS access to communications
devices through MERGE.

Now, you say, why does that Tucker kid want to turn my tty driver
into a newt ("Well, [it] got better.")?  Because it can be very useful
if you are developing asychronous communications systems.  I find
it very useful to know how many times CTS fluctuated during a test
session.  Like the 'ecufriend,' it isn't for everyone, but if you
need it, there it is.

>Message-Id: <[e-mail]>
>From: emory!geminix.in-berlin.de!gemini (Uwe Doering)
>Subject: Re: [Request permission to distribute FAS 2.08 instrumented version]
>To: [e-mail] (Warren Tucker)
>Date: Mon, 29 Apr 91 17:43:51 MES
>
>Hello Warren,
>
>>...I am writing is to ask your permission ...
>>to include with ecu a modified FAS 2.08 I am calling FAS/i (for
>>instrumentation) so you can have access to statistics 
>> [some stuff deleted]
>
>You sent me the necessary patches some time ago. Since then I tried to
>make up my mind about this issue. I decided now that I won't have the
>patches in the official FAS release. There is a reason for that. I want
>to keep FAS as clean as possible from the application program standpoint.
>
[some stuff deleted]
>
>You have my permission to release your special FAS version, but please
>make it clear in the docs that _you_ do the support for it, and
>that it is no official FAS release.
>
>      Uwe
>-- 
>Uwe Doering  |  INET : [e-mail]
>Berlin       |----------------------------------------------------------------
>Germany      |  UUCP : ...!unido!fub!geminix.in-berlin.de!gemini

The strength of my earlier comments is driven in part by Uwe's comments.
I will be very disasppointed and red-faced if Uwe gets ONE query or
request in regard to this hack.  I will be similarly dismayed if he
gets one comment pro or con about folding these features into the
official FAS.

This is a dead-end, special-purpose junkbox addition that just happens
to be potentially useful.

Now, with all that out of the way, here are a few useful tidbits:

1. Configuration and installation are for the most part similar to
the standard FAS as of this writing.

2. You will need to manually create a /usr/include/local directory
before you begin any makes.

3. The original FAS 2.08 functionality may be restored by turning
off #define FASI in several places.

4. It has been used only on SCO.

5. To use with ECU, you'll need to hack -DFASI_IN_USE into the
ecu Makefile.  The other programs in ecu don't need to know about it.
(Hint: ecufriends can make good use of the features.)

6. If you turn on FAS/i support in ecu, you get the undocumented
fasi interactive and procedure commands and these %functions:

Integer functions:
%fasi       defined for all ecu versions; returns 1 if FAS/i
            support included, else 0 if not. The other functions
            will cause procedure termination with undefined function
            errors "ifi %fasi==0".
%msr        MSR current value
%lnerr      accumulated FE+OE+PE count
%ridet      accumulated RI count
%brdet      accumulated BREAK count

String functions:
%msrtext    MSR current value in string form for easier (less efficient)
            MSR inspecition.  You can do something like

     $s20 = %msrtext
     ifi %instr($s20,'CTS')
         echo 'CTS present'
     ifi %instr($s0,'RING')
         echo 'We are receiving a RING this very instant')
The returned string is one or more substrings separated by asterisks.
So, you might get 'CTS*DSR*RING*'.  The list of substrings, one for each
bit in the canonized 8250 MSR:

dCTS        delta CTS   <---+
dDSR        delta DSR       | you are unlikely to see these
dRI         delta RI        | since the driver catches interrupts
dDCD        delta DCD   <----
CTS     
DSR     
RING        
DCD     

7. The fasiintf.c modules contains examples of each FAS/i-specific
ioctl.  These are also the only 'documentation' ever likely to be
produced for them other than this list:

FASIC_SIP                  get entire fas_info struct
FASIC_MSR                  get various registers
FASIC_LCR     
FASIC_IER    
FASIC_MCR   
FASIC_DVR_IDENT            get driver revision
FASIC_SPACE_IDENT          get space.c revision
FASIC_RESET_STAT           reset statistics

8. This hacked 'release' is in the style of a purpose-specific
driver.  If you have an SCO UNIX 3.2.x system with a standard COM1
port and a Digiboard PC-8 on COM2, man are you in luck.  Otherwise,

   while(!bored && !fed_up && !success)
   {
     adopt();
     adapt();
     improve();
   } /* "... I always say." -- thanks to John Cleese */
   exit(!success);