When performing floating point multiplication, the prior code overwrote
an additional word of the floating point fraction with zeros. This is
harmless with standard FP, as the floating variables always have space
for EP-length vars. When doing an EP multiply, this causes a word on
the stack to be zeroed. For the latest Raspbian release, this causes a
segfault as there's no padding past that var on the stack.
This fix, which has been tested against the original crashing code plus
validated using the FPP-8 diagnostics, avoids the overwrite.
Devices that do single character I/O could be attached to non seekable
host OS devices (tty, pipes, etc.) and thus shouldn't count on fseek()
and ftell(). These DEVICEs on these simulators do single character I/O
and easily can update their POS REGisters to reflect how much data has
been emitted. Changing such a REGister will have no useful effect
when attached to a non seekable file.
Historically this functionality was reimplemented within each
DEVICE simulator often with slightly different implementations
and inconsistencies. Solving this globally within SCP required
changes in many places, but should henceforth be reasonably
managed.
As discussed in #1034
- Initialize local state variables to 0. Likely non functional changes due
to lack of depth in static analysis scan. Coverity detects real problems
like this.
- Fix inconsistent statement indentation.
This avoids a potential invalid pointer dereference when formatting
the return value from sim_instr() if it is < SCPE_BASE but greater
than the previously defined static array size.sizeof
Update simh.doc to reflect this generic change.
This implements the principle of "least surprise", in that users won't
normally expect to start overwriting an existing file on these devices.
Real hardware didn't behave that way. A new (empty) file can always
be created with the -N switch on the ATTACH.
Although these devices interpret a -A switch at attach time, they are not
sequential output devices and thus don't have an ambiguous interpretation
of the switch.
Incorrectly mentioned in #821
The -A attach time switch implicitly means to open files in append
mode.
Some devices used -A to indicate ASCII data or AUTOSIZE of disk or
fixed size tapes - DECtape. These devices are either read only devices
(Paper tape readers or card readers), so explicit forcing of the attach
to be -R avoids unexpected positioning errors. The random access
Disk and fixed sized tapes buffer the container contents in memory
so append mode isn't relevant, but care must be taken to assure that
the buffer reading starts at the beginning of the file without regard
to the explicit open mode.
As reported in #821
Most history routines defined a local sim_eval of the proper length, but the
erroneous ones were fixed length machines that defined an integer variable
instead of an integer array of length 1. The VAX used the global sim_eval.
The changes follow the VAX practice.
The code is a bit difficult to understand, but it represents the 'normal'
path for processing a DECtape word. The code always flows all the
way to the break.
For start of block, there's extra code to check for a timing error; but
then the first word is processed (case 0).
For a normal word, a 3-cycle data break is done - increment word
count and current address, check for word count overflow, put the
word in the buffer; but then check for end of block (case DTO_WCO).
If the word count has already overflowed, just check for end of block.
So yes, the end of case 0 should be labeled 'fall through' as well.
FNC_WRIT has the same structure and needs the same comment.
For the PDP11, this includes deferring the setting of DONE on an error, allowing RSTS V4 to work correctly.
There was a minor (possibly only theoretical) bug in dt_seterr as well.
dt_seterr is supposed to stop the drive. If the drive is accelerating, at
speed, or stopped, it works fine. But if the drive is decelerating, it did
not clear out the "successor" states. In theory, the drive could be in the
middle of reversing - that is, there are successor states of accelerating
(in reverse) and at speed. So the successor states need to be cleared,
or the drive won't actually stop.
The fifth DECtape controller (the PDP8's TD8E) is much simpler and
doesn't have the problem.
It was defined in lowercase and has now been made upper case for consistency.
The code which performs the match also has been change to do a case
insensitive compare.
These changes facilitate more robust parameter type checking and helps
to identify unexpected coding errors.
Most simulators can now also be compiled with a C++ compiler without
warnings.
Additionally, these changes have also been configured to facilitate easier
backporting of simulator and device simulation modules to run under the
simh v3.9+ SCP framework.
