1
2 09/23/87 HASP
3 Known errors in the current release of HASP.
4 # Associated TR's
5 Description
6
7 0017 phx19286
8 The calls to iodd_msg_ in hasp_ws_sim_driver_ which report errors
9 returned by receive_file_ have the code and program name arguments
10 inverted.
11
12 0016 phx19004
13 If tty_area runs out of room while setting up a HASP multiplexer,
14 the system will crash as priv_hasp_mpx's area handler will fault.
15
16 0014 phx17682
17 If Multics is sending output print/punch files to a workstation at
18 full speed i.e. data is always ready to transmit whenever the
19 workstation asks for the next block of data and the workstation
20 wishes to send an input file to Multics, the workstation will not
21 start sending data until after any output from Multics has ceased
22 i.e. at the end of the print/punch file. However, if the
23 workstation has already started sending input before Multics starts
24 sending its output, data will flow in both directions simultaneously.
25
26 This problem is caused by the algorithm used to build output blocks in
27 Multics. When the workstation is ready to send an input file, it will
28 first send a special Request-To-Send RTS record to Multics. The
29 workstation will then wait for an acknowledgement to the RTS record
30 before starting to send the file. Multics, however, will not send the
31 RTS acknowledgement unless there is no data waiting to be output.
32 Multics does not include the RTS acknowledgement record in a block
33 with other data because there are workstations which will not
34 recognize this record unless it is the only record in a data block.
35
36 Multics should be changed to give priority to RTS acknowledgement
37 records over normal data records.
38
39 0011
40 The ++PASSWORD control record should be supported to provide some
41 small level of security on returned data files. The ability to use
42 a blank password should also be allowed.
43
44 0009 phx13648
45 Failure to load automatically or in response to load_mpx should
46 issue a "Load failed" error message rather than "Crash signalled".
47 The error message documentation for this new message could then
48 explain the various interactions with configuration parameters eg:
49 connect_timeout that can cause load to fail.
50
51 0006
52 Ring-0 will occasionaly send the end of one block and the beginning of
53 the next block in response to a SEND_OUTPUT from the FNP. This will
54 cause the FNP to report a malformatted BISYNC block to ring-0 which
55 then crashes the multiplexer. This is an extremely rare occurence.
56
57 0005
58 The multiplexer generates a spurious INPUT_AVAILABLE interrupt for a
59 subchannel when there is actually no input present. This behavior can
60 cause applications such as IMFT to block forever on input or timeout
61 at inappropriate times.
62
63 0004 phx16125
64 Blocks can be transmitted out of sequence. Ie: the system will send
65 block N+1 when the remote host or workstation is expecting block N.