dinorefurb-dosbox-session
Owned emulator debugger sessions for a restoration's research tooling: the original program runs under DOSBox-X's structured debugger or under rust-dos's debug server, and the tooling reads registers and memory, stops at breakpoints, observes operations, makes checked writes to stopped guest memory and keeps an event log of the run beside the static evidence. The design is ADR 0026 for DOSBox-X and ADR 0040 for rust-dos, which has its own section. Both backends take the same run lock and keep the same event log.
The package owns the parts that decide whether a recorded run can be trusted and that carry no game knowledge: the emulator process, the machine-wide run lock, the guest drives, muted host audio, the session record, request IDs, operation observation, guarded writes, gated breakpoints and the event log. What a run means (executable fingerprints, address maps, state layouts and which fields may be written, input, screens, event kinds and what an outcome must say) stays in the restoration. A restored game never depends on this package.
Windows only. On another platform a session refuses to start and says so.
The DOSBox-X client is yours to import
DOSBox-X's structured debugger ("Agent") and its Python client dosbox_agent are in the
DOSBox-X source tree under GPL-2.0, and are not on PyPI. This package is MIT and never imports,
vendors or depends on the client. Your tooling imports it from your own checkout and passes a
factory that builds a client for the session's endpoint:
import sys
from pathlib import Path
from dinorefurb_dosbox_session import DosboxSession, SessionSettings, Target, verify_checkout
checkout = verify_checkout(Path("artifacts/dosbox-x")) # refuses another revision or local changes
sys.path.insert(0, str(checkout.path / "client" / "python"))
from dosbox_agent import AgentClient
settings = SessionSettings(
checkout=checkout.path,
emulator=checkout.path / "bin/x64/Agent Debug SDL2/dosbox-x.exe",
run_directory=Path("artifacts/runs/2026-10-10-startup"), # new or empty for each run
target=Target("GAME.EXE"),
client_factory=lambda endpoint: AgentClient.from_config(endpoint.agent_config),
prepare_drive=lambda drive_c: ..., # put the target's files on the new, empty C:
)
with DosboxSession(settings) as session:
registers = session.client.get_registers(session.session_id)
operation = session.continue_()
observation = session.observe(operation, timeout=10)
while observation.pending:
observation = session.observe(operation, timeout=10)
The checkout must be at the revision the package was verified against,
b6abbd5980a885f5f310a4088c59a8688d1b116c (tag dosbox-x-v2026.10.01, PINNED_REVISION), with
no modified, staged, deleted or untracked files. The __pycache__ directories that importing the
client writes are not counted as changes. Moving to a new revision is a package release.
What a session does
Entering DosboxSession (or calling start()):
- Refuses a platform other than Windows, a checkout that fails the check above, a missing
emulator, and a run directory that is not empty (a
drive-cfrom an earlier run included). Withevent_logset, hashes the named modules (refusing one that is not imported). - Takes the run lock (below), or refuses with a report of the recorded owner and processes.
With
event_logset, then writes the log header. A refusal up to here leaves the run directory empty, so the same directory can be used again. - Creates
drive-cempty in the run directory and callsprepare_driveon it. - Writes
dosbox.confandagent.envand launches the emulator with a native console that is created and hidden. Redirecting the console,-noconsoleandCREATE_NO_WINDOWeach broke debugger entry at the pinned revision. The emulator getsemulator_argumentsfirst, then-confand--agent-config. It is created inside a Windows job object that only this process holds, so Windows ends it when this process ends for any reason, a hard kill included, even before the run lock records it (ADR 0030). When Windows refuses the launch (a file that is not an executable, or a Python process inside a job that does not allow a nested one), the session raisesEmulatorLaunchFailed. - Waits for the readiness marker the guest's
[autoexec]writes toC:\DRREADY.TXTafter its drives are mounted. An answering debugger server is not readiness. Fails when the emulator exits first or the marker does not appear withinreadiness_timeoutseconds. - Builds the first client through the factory, reads the server's capabilities, and starts the
target stopped at its entry. Fails unless the session stops with reason
startup.
Any failure cleans up as leaving does, then raises. A failure after the log header was written ends the log with an outcome that records it.
Leaving (or close()) stops the debugger session, closes every client, terminates the owned
emulator and releases the run lock. If the emulator is still running afterwards, it writes
cleanup-diagnostic.txt, keeps the lock and raises CleanupFailed; closing again after the
process has exited releases the lock. If your program ends first, Windows ends the emulator with
it, and the stale-lock command then removes the lock. It never stops or removes anything the
session did not start.
The generated configuration
EmulatorConfig takes the media, extra dosbox.conf sections (such as
{"cpu": {"cycles": "fixed 10000"}}), keep_host_sound and the Agent limits.
- C: is the run's own
drive-c, writable. It is created empty for each run and never reused. - Each
Mediais mounted read-only on a letter from D to Z: anisowithimgmount -t iso, adirectorywithmount -ro. - Host audio is muted by default:
mixer master 0:0 /noshowin[autoexec]and[midi] mididevice=none. The emulated sound devices stay configured.keep_host_sound=Trueleaves both alone.nosoundset to anything DOSBox-X does not read as false is refused, becausenosound=truebroke structured readiness at the pinned revision. - The session writes
[autoexec]itself, so a section by that name is refused, as is a section name, key or value with a line break in it.
The run lock
One lock file per machine keeps two probes from sharing one machine's emulator, input or timing.
The path is C:\ProgramData\refurbished-dinosaurs\run.lock unless the environment variable
REFURBISHED_DINOSAURS_RUN_LOCK names another. That variable is a machine setting: two programs
that resolve the lock path differently do not exclude each other, so set it in your user
environment or not at all. SessionSettings.lock_path overrides both, for tests.
The lock records the session, the owner process and the emulator, each by process ID and start
time, so a process that later reuses an ID does not match. The emulator cannot outlive the owner
(step 4 above), so once every process a lock records has exited, no emulator from that session
runs, including one the owner launched but had not recorded yet (an unrecorded emulator may
still be finishing its exit for a moment after the owner reads as exited). A session that finds the lock
refuses to start (LockHeld) and its report says which recorded processes still run. Nothing
removes a lock automatically. When every recorded process has exited, remove it with:
dosbox-session stale-lock [--lock PATH] [--json]
It checks the processes again first, and refuses (exit code 1, nothing removed) when one still runs or cannot be queried, when the file is not a record this package wrote, or when deleting it fails. Exit code 0 means it removed the lock or found none; 2 is a usage error.
The session record
session.json in the run directory is rewritten at each step and holds:
| Field | Meaning |
|---|---|
checkout |
The checkout's path and revision. |
emulator |
The emulator's path and SHA-256. |
build_link |
States that the two facts above are separate: nothing shows the emulator was built from that checkout. |
run_lock |
The lock path this session took. |
endpoint |
The session's named pipe, unique to the session. |
owner_process, emulator_process |
Process ID and start time of each. |
host_sound |
muted or kept. |
readiness |
observed once the guest wrote its marker. |
request_id_prefixes |
One per client. |
capabilities |
What the server reported. |
debugger_session |
The debugger session's ID. |
writes |
Each guarded write in order: the contract's name, the field, its address (as the address object's repr) and length, the expected hash, the hash of the data the write carried (null when it was not bytes), verified or failed, and the failure, which says how far a failed write got. |
run_failure |
Why the run failed, or null. |
event_log |
The event log's path, or null when the session keeps none. |
Calls, capabilities and request IDs
session.client and each session.open_diagnostic_client() wrap a client from your factory.
Every call they make carries a request ID from that client's own namespace
(<session>.c<client>.<n>), so a diagnostic client on the same session cannot collide with the
first. A call that needs a capability the server did not report as true raises
CapabilityRefused and is not sent: every debugger call needs debugger, a memory_change
breakpoint needs breakpoints.memory_change, and CPU tracing needs trace.cpu. The wrappers offer
no writes to guest state; session.write makes them, as the next section describes.
Guarded writes
session.write(contract, field, data, expected_sha256) writes to the stopped guest's memory. The
contract is yours: a FieldContract with a name you choose and the WritableFields you support,
each with a name, an address built with the client's MemoryAddress and a length. The package
has no default contract and supports no field on its own, so a write without a contract is
refused. Field layouts and the rules for when a field may be written stay in your restoration.
import hashlib
from dosbox_agent import MemoryAddress
from dinorefurb_dosbox_session import FieldContract, WritableField
contract = FieldContract("startup-state/1", (WritableField("counter", MemoryAddress.segmented(cs, 0x0200), 2),))
session.write(contract, "counter", b"\x21\x43", expected_sha256=hashlib.sha256(b"\x34\x12").hexdigest())
Each write, in this order:
- Refuses a field the contract lacks,
datathat is not bytes, ordataof another length than the field (WriteOutsideContract). Nothing is sent. - Refuses an
expected_sha256that is not 64 hexadecimal digits, and a guest whose status is notstopped(WriteFailed). - Reads the field and refuses unless its bytes hash to
expected_sha256(WriteHashMismatch). Nothing is written. - Sends
memory.writewith the sameexpected_sha256, so the server checks it again. The hashes the server reports for the bytes it replaced and the bytes it left must match the expected hash anddata; when the server wrote but reports replacing other bytes, the field may hold the new bytes (WriteFailed). - Reads the field back and compares it with
data(WriteReadbackMismatch).
It returns a VerifiedWrite and appends it to writes in session.json.
Any refusal or failure, including a transport error during the write, fails the run. The write is
not retried, the failure goes into session.json as run_failure, and from then on further
writes, continue_ and step raise RunFailed without sending anything. pause is still sent,
so a guest that was running when the write failed can be stopped, and reads still work, so the
failed state can be inspected. Closing the session cleans up as usual. Start a new
run to try again.
The event log
With SessionSettings.event_log set, the session writes events.jsonl in the run directory: one
header line, one line per event, and one outcome line, as JSON. Each line is written, flushed and
synced to disk before the call that writes it returns, so anything logged before a continuation
survives the owning process being killed. Everything the log checks comes from you:
from dinorefurb_dosbox_session import EventLogSettings, EventSchema, OutcomeContract
event_log = EventLogSettings(
contract=OutcomeContract("startup-probe", 2, {"reached_menu": "boolean", "frames": "integer"}),
schemas=(
EventSchema("stop", required={"frame": "integer", "ip": "string"}),
EventSchema("input", required={"key": "string"}, optional={"note": ["string", "null"]}),
),
modules=("probe.adapter", "probe.startup"), # imported before the session starts
)
settings = SessionSettings(..., event_log=event_log)
with DosboxSession(settings) as session:
session.log_event("stop", {"frame": 0, "ip": "0x0100"})
...
session.finish_log({"reached_menu": True, "frames": 412})
- An
EventSchemanames an event kind and itsrequiredandoptionalfields, each with a JSON type (string,integer,number,boolean,null,array,object) or a list of them. Anintegeris also anumber. An event carries every required field, may carry the optional ones, and carries nothing else. Only the top level is checked. - The
OutcomeContracthas a name, a version and the fields the final outcome must carry, all required and no others. Give a changed contract a higher version. - The header records the schemas, the contract and, for each module in
modules, the file it was imported from and that file's SHA-256, taken after the modules are imported and before the emulator starts. A named module that is not imported then, or that has no file (a built-in or a namespace package), is refused withModuleRefusedbefore the lock is taken. The header is written once the session holds the lock. session.finish_log(values)ends the log with a completed outcome, andsession.fail_log(failure, values=None)with one that records a failure. The outcome records how many events came before it, a SHA-256 over them in order and a SHA-256 of the header.- Once the log has its outcome, by either call or by a refusal,
continue_,stepand writes raiseRunEnded(orRunFailedafter a failure) without sending anything, so the log covers everything the run did to the guest.pauseand reads still work. An event logged after the outcome raisesLogEntryRefusedand does not fail the run. - Entries go through
log_event,finish_logandfail_log.session.event_log_pathgives the log's path. - An event the log refuses (a kind with no schema, data that does not fit it or that JSON cannot
hold) or outcome values that do not fit the contract end the log with a failure outcome that
names the refusal, raise
LogEntryRefusedand fail the run, as a failed guarded write does. A failed guarded write ends the log with the run's failure too, and so does a session that fails to start after writing the header. - A failed write to the file raises the
OSErrorand fails the run. The log takes nothing more and has no outcome, so it reads as truncated or incomplete. - A session closed without an outcome leaves the log without one, and it reads as incomplete.
The package does not decide what an event or an outcome means. Boundary names, what counts as a pending operation, replay and comparison with your own journal stay in your restoration.
Reading a log
from dinorefurb_dosbox_session import read_event_log
log = read_event_log(run_directory / "events.jsonl", {"reached_menu": True, "frames": 412})
read_event_log(path, expected_outcome, max_bytes=DEFAULT_MAX_LOG_BYTES) returns an EventLog
with the recorded contract, schemas, module hashes, events and outcome, or raises a subclass of
LogRejected that says which check failed. The checks run in this order:
| Error | The log |
|---|---|
LogOversized |
is larger than max_bytes (64 MiB by default). It is not parsed. |
LogMalformed |
has a line that is not a JSON record this package writes, a header that is not first or has another format, or a record after the outcome. |
LogTruncated |
ends partway through a line, or is empty. |
EventSchemaViolation |
has an event whose kind or data does not fit the schemas its header records. |
LogIncomplete |
has no outcome. |
EventsMismatch |
has events or a header that differ from the count and hashes its outcome recorded: an event is missing, extra, changed or reordered. |
OutcomeFailed |
ends with an outcome that records a failure. |
OutcomeContractViolation |
has an outcome with a contract field absent or of the wrong type, or a field the contract lacks. Nothing is filled in. |
OutcomeMismatch |
has outcome values that differ from expected_outcome, which is compared whole, so a field you do not expect fails too. |
Each error's line is the line the check failed on, or None for LogOversized and
LogIncomplete. A log is read against the contract its own header records, never against a
newer version you have defined since, so a version 1 log still reads as version 1 and a field
version 2 added is not filled in. log.contract.version says which one it was. The header is
covered by the outcome's hash, so a header changed to name another contract is rejected.
Observation
session.observe(operation, timeout) waits on one operation. Each poll first checks that the owned
emulator still runs (EmulatorExited if not). When the time runs out the result is pending,
which is neither a failure nor a result: observe the same operation again to keep waiting.
Transport errors from the client propagate. Observation never restarts the guest or sends another
continuation, and while an operation is pending a further continue_ or step raises
OperationPending. A pause ends the continuation it interrupts, so observing either one clears
both. When a continue_ or pause request itself raises, the server may still have received it:
continue_ and step raise OperationPending until client.status() shows the guest
stopped, exited or failed.
Gated breakpoints
A guest that polls in a loop until a timer reaches some value passes the loop's top many times
per timer tick. A breakpoint there stops every pass, although nothing a timer condition reads can
change between two ticks. A GatedBreakpoint stops at the boundary only at the first pass after
the guest runs one of the wake addresses you name, such as the timer interrupt handler. The
design is ADR 0032.
from dinorefurb_dosbox_session import CodeAddress, GatedBreakpoint
# Stopped at the loop's top: check once here, then remove your own breakpoint at it.
if not ready(session):
session.client.delete_breakpoint(session.session_id, poll_breakpoint.id)
with GatedBreakpoint(session, CodeAddress(cs, 0x0120), (CodeAddress(0xF000, 0xFEA5),)) as gate:
while True:
stop = gate.run(timeout=10)
if stop.pending:
continue # still running; the same continuation is observed again
if stop.kind == "ended":
break
if stop.kind == "other":
record_other_stop(stop.session) # your own breakpoints still stop the guest
continue
if ready(session): # stop.kind == "boundary"
break
# The guest is stopped at the loop's top, the first pass at which ready() held.
- Opening needs a stopped guest in real or virtual-8086 mode, or in protected mode with a resolver
(below). It sets an execution breakpoint at
each wake.
run(timeout, poll_ms=100)continues the guest through the session. At a wake stop it sets a breakpoint at the boundary that the server removes when it is hit, unless one is set, and continues. It returns aGateStop:boundary,other(any stop that is not the gate's, with an armed boundary left armed),ended(the debugger session exited or failed) orpending.stop.wakescounts the wake stops it continued from during the call, andgate.wakes_seensince it opened. Closing removes the breakpoints the gate set and leaves the guest where it is. Once a run returnedended, the gate refuses to run again and closing sends nothing to the session. - The gate never reads or evaluates your condition. The boundary stop is the first pass at which your condition can hold only if everything it reads changes in code that runs a wake address before the guest reaches the boundary again. Showing that, choosing the addresses, and checking the condition at the stop where you open the gate are yours.
- Two breakpoints at one address stop the guest once and the server reports only one of them, and
at the pinned revision a breakpoint removed when hit is removed without a stop of its own when
another breakpoint at its address is reported. So the gate refuses a boundary at a wake's
address, two wakes at one address, and any breakpoint it did not create at the boundary or a wake
address (
GateRefused). In real and virtual-8086 mode it compares addresses assegment * 16 + offset, with and without the A20 wrap, and refuses an offset of its own above 0xFFFF. It checks when it opens and before eachrunthat continues the guest. A gate works in the mode it opened in: it refuses to continue in another, and stops if a wake stop that arms the boundary is in another. - Continuations go through the session, so a pending operation, a failed run or an ended log
refuses them as usual. A pending continuation is observed again by the next
run, which never sends a second one. Transport errors propagate. Eachrunthat continues the guest first reads the session's status. After a continue request raised, the nextruncompares the status with thestate_revisionthe guest had when that request was sent: a higher one is handled as an observed stop, so a wake the lost reply reached still arms the boundary. Breakpoint requests raise the revision too, so after arming the boundary the gate reads the registers and keeps their revision. A write, step or breakpoint request the session sends after a failed continue request would look like a stop, so the nextrunraisesGateRefusedand the gate stops (session.changescounts them): run the gate again before changing anything. After a breakpoint request raised, the gate raisesGateRefusedinstead of continuing, and closing removes any breakpoint left at its addresses. Closing after a continuation the gate has not seen end reads the session's status: while it shows the guest running, closing raisesOperationPendingand keeps the breakpoints; once you have paused the guest and observed the pause, closing removes them. The session's own teardown is unaffected.
Protected mode
The debugger places a protected-mode breakpoint at the linear address its selector's descriptor
gives when the breakpoint is created, and the Agent reports neither descriptor bases nor linear
addresses. So in protected mode the gate places addresses only through resolve, which you
supply (ADR 0037):
def resolve(address: CodeAddress) -> int | None:
# The linear address the debugger places selector:offset at, from descriptors you verified:
# base + offset, with the offset wrapped to 16 bits for a 16-bit segment. None if unknown.
descriptor = verified_descriptors.get(address.segment)
if descriptor is None:
return None
offset = address.offset if descriptor.big else address.offset & 0xFFFF
return descriptor.base + offset
gate = GatedBreakpoint(session, CodeAddress(0x0028, 0x00012345), (CodeAddress(0x0030, 0x0100),), resolve=resolve)
- The gate asks
resolvefor its own addresses and for every other execution breakpoint when it opens and before eachrunthat continues the guest. It refuses a breakpointresolvecannot place, and two places that are equal. - When
resolveplaces one of the gate's own addresses somewhere else than where its breakpoint was set, the gate stops: the breakpoint stays where the debugger put it. At a wake stop it asksresolvefor the boundary again before arming it, and stops without arming it when the place differs from the one it compared when it opened. An error fromresolvewhile the gate judges a stop also stops the gate. resolveanswering as the debugger places addresses is yours to show, including the hidden base of the currentCSand a descriptor that changes while the gate is open. A breakpoint you set in another mode is placed where it was created;resolvehas to answer with that place.- All of a gate's addresses are placed in the mode it opened in, so a wake that runs in real mode (such as a BIOS handler the extender reflects an interrupt to) cannot arm a protected-mode boundary. Pick wakes in the boundary's mode.
The pinned DOSBox-X has no breakpoint conditions, run-until, hit counts or batch reads, so the package offers none of them. To read several fields at one stop, read one range that covers them.
rust-dos sessions
RustDosSession runs rust-dos instead of DOSBox-X.
rust-dos serves its debugger over HTTP and JSON on a local port, in its normal release binaries,
so there is no source build, no client to import and no console to hide. The package speaks that
API with the standard library and imports nothing of rust-dos.
The session needs a rust-dos with debug server API version 1 and the features journal, ops,
if_state_seq, request_ids, write_expect, run, program_stops and quit
(RUST_DOS_FEATURES). No rust-dos release has them yet; build kibertoad/rust-dos main (on
Windows, the linker needs the SDL2 VC development package's lib/x64 on LIB, and SDL2.dll
goes beside rust-dos.exe).
from pathlib import Path
from dinorefurb_dosbox_session import Media, RustDosConfig, RustDosSession, RustDosSettings, Target
settings = RustDosSettings(
emulator=Path("tools/rust-dos/rust-dos.exe"),
emulator_sha256=None, # or the executable's SHA-256, to refuse any other
run_directory=Path("artifacts/runs/2026-10-11-startup"), # new or empty for each run
target=Target("GAME.EXE"),
prepare_drive=lambda drive_c: ..., # put the target's files on the new, empty C:
emulator_config=RustDosConfig(media=(Media("D", Path("game.cue"), "iso"),)),
)
with RustDosSession(settings) as session:
cs = session.entry["registers"]["cs"] # stopped at the target's first instruction
session.client.set_breakpoint(f"{cs}:0120")
operation = session.continue_()
observation = session.observe(operation, timeout=10)
while observation.pending:
observation = session.observe(operation, timeout=10)
print(observation.stop["reason"], session.client.registers()["ax"])
What a rust-dos session does
Entering RustDosSession (or calling start()):
- Refuses a platform other than Windows, a missing emulator, an emulator whose SHA-256 differs
from
emulator_sha256when that is given (EmulatorRefused), options the session sets itself inemulator_arguments(ConfigurationRefused), a deterministic run without a numeric speed, media on C: or two media on one letter, a target command line with a line break, a deterministic target command line that is not ASCII or holds%(it goes intoAUTOEXEC.BAT, which would expand it), and a run directory that is not empty. Withevent_logset, hashes the named modules. - Takes the run lock, the same one
DosboxSessiontakes, or refuses withLockHeld. - Creates
drive-cempty in the run directory and callsprepare_driveon it. A prepared drive that holdsAUTOEXEC.BATor the readiness marker is refused. The session writesAUTOEXEC.BAT, which writes the readiness markerC:\DRREADY.TXT. - Launches rust-dos in a job object that ends it with its owner, with
--no-config,-dondrive-c,--debug-server 127.0.0.1:0,--logonrust-dos.login the run directory, and--cycles,--deterministicand--start-timefromRustDosConfig. SDL's dummy video driver hides the window and its dummy audio driver mutes the host, unlessshow_windoworkeep_host_soundis set; the emulated sound devices stay. The emulator's output goes torust-dos.out, where the session reads the port the server took. - Checks
/api/version: the namerust-dos, API version 1 and every feature above, or refuses withEmulatorRefused. A server that refuses/api/versionwith a status below 500 is refused the same way. - Mounts each medium read-only through the server (an
isoas a CD-ROM drive, any CD image rust-dos reads; adirectoryas a hard disk) and fails unless the server reports it read-only. - Waits for the readiness marker, then runs the target at the DOS prompt with
stop_at_entry. A deterministic machine starts paused beforeAUTOEXEC.BATruns and takes no run command until the batch file ends, so in that modeAUTOEXEC.BATstarts the target after the marker and aprogram_startstop catches it. Either way the start fails unless the machine stops at the entry of a program with the target's file name.session.entryholds that stop.
Any failure cleans up as leaving does, then raises.
Leaving (or close()) asks the emulator to quit, terminates it if it has not exited, and releases
the run lock once it has. A process that outlives termination keeps the lock and raises
CleanupFailed, as for DOSBox-X.
The rust-dos session record
session.json has the format dinorefurb-dosbox-session.rust-dos-record/1 and holds backend
(rust-dos), emulator (path, SHA-256 and the expected SHA-256 or null), server (the
/api/version answer, with the instance ID that is new at every start), endpoint
(127.0.0.1:<port>), arguments (the launch command line), emulator_log, run_lock,
owner_process, emulator_process, host_sound, window (hidden or shown),
deterministic (the speed and start time, or null), media (each drive as the server reported
it), readiness, request_id_prefixes, entry (the entry stop's CS:IP, program and journal
number), writes, run_failure and event_log.
Calls, operations and observation
session.client and each session.open_diagnostic_client() are a RustDosClient. Every request
carries an X-Request-Id from that client's own namespace (<session>.c<client>.<n>), and the
server copies it into each journal entry the request causes. A reply with a status of 400 or above
raises ServerRefused with the status, the body and, for a refused precondition, its name. A
reply cut short or unreadable raises ConnectionError, like a dropped connection. The
client offers reads (status, registers, read_memory, disassemble, screen_text,
screenshot, drives, log), breakpoints and watchpoints (set_breakpoint with once,
break_on for program_start and program_exit), the journal and operations. request sends any
other call.
Requests that resume or change the guest go through the session:
continue_(until=None, until_ms=None),step(count=1)andstep_over()start a continuation and return aRustDosOperation. One continuation runs at a time: another raisesOperationPendinguntil an observation has seen the first end.send_input(kind, event)queues keys, typed text, mouse events or a batch; the operation ends once the input is delivered. Input may be sent while a continuation runs.pause()stops the guest. The continuation it interrupts endsstoppedwith reasonrequest; observe it to clear it. A pause is still sent after the run failed.observe(operation, timeout, poll_ms=500)waits on any operation, checking on each poll that the emulator still runs (EmulatorExited);poll_msmust be a positive integer. It returns aRustDosObservation:completedwith the state the operation ended in (stopped,done,failed,supersededorcancelled) and the stop, orpendingwhen the time ran out.- When a continuation request raises a transport error, the server may have taken it. Before the
next continuation the session looks for the request's ID in the journal: if it is there, its
operation is the pending one; if it is not and the machine is paused and quiet, the server did
not take it; otherwise the session raises
OperationPendingand sends nothing.
Continuations, steps, input and writes raise RunFailed after the run failed and RunEnded once
the event log has its outcome, as for DOSBox-X. The event log calls (log_event, finish_log,
fail_log, read_event_log) are the ones described above.
Guarded writes in a rust-dos session
session.write(contract, field, data, expected_sha256) takes the same FieldContract and
WritableField as a DOSBox-X session, with the field's address as a rust-dos address string, such
as "1234:0200" or "phys:12540". A segment-register form such as "DS:0200" resolves against
the registers at each request, so a contract should name fixed addresses. Each write:
- Refuses a field outside the contract or
dataof another length (WriteOutsideContract). - Reads the field. It refuses a guest that is running (
WriteFailed) and bytes that do not hash toexpected_sha256(WriteHashMismatch); nothing is written. - Writes with
expect(the bytes read) andIf-State-Seq(the journal state they were read in). The server refuses the write, and writes nothing, if anything changed the machine since the read, if it is not paused, or if an operation or input is under way (WriteFailed), or if the bytes differ (WriteHashMismatch). - Requires the bytes the server reports replacing to be the bytes read (
WriteFailedotherwise: the field may hold the new bytes) and the bytes it reports leaving to bedata(WriteReadbackMismatch). - Reads the field back, which must hold
datain the state the write left (WriteReadbackMismatch, orWriteFailedwhen the machine changed in between).
Any refusal or failure fails the run and is recorded in session.json, as for DOSBox-X.
Deterministic runs
With RustDosConfig(deterministic=True, cycles="10000") rust-dos runs on emulated time alone: the
clock starts at start_time (rust-dos's default is 1995-04-11 12:34:56), the speed is fixed, and
input is delivered at emulated times. continue_(until_ms=N) stops when emulated time reaches N
milliseconds since power-on. Two runs with the same input and the same stops reach the same state
at the same emulated time. rust-dos takes input and changes to the machine only while it is
paused in this mode. cycles must be a number, because rust-dos replaces auto or max with a
speed of its own choosing.
Gated breakpoints are offered for DOSBox-X sessions only, and neither backend offers save states yet.
Errors
| Error | Raised when |
|---|---|
PlatformRefused |
The platform is not Windows. |
CheckoutRefused |
The checkout is at another revision, has local changes, or cannot be read with git. |
RunDirectoryRefused |
The run directory is not empty, or prepare_drive wrote the readiness marker. |
ConfigurationRefused |
A setting is one the session owns or one known to break the debugger. |
LockHeld |
The run lock is held, or cannot be read. report describes it. |
EmulatorLaunchFailed |
Windows refused to start the emulator in its job. Nothing runs and the lock is released. |
EmulatorRefused |
A rust-dos executable has another hash than emulator_sha256, or its debug server is not rust-dos, speaks another API version or lacks a feature the session relies on. |
ServerRefused |
The rust-dos debug server answered a request with a status of 400 or above. status, body and precondition say why. |
EmulatorExited |
The owned emulator exited while the session needed it. |
ReadinessNotObserved |
The guest did not write its readiness marker in time. |
CapabilityRefused |
An operation needs a capability the server did not report. Not sent. |
OperationPending |
A continuation was asked for while another operation is pending. |
GateRefused |
A gated breakpoint refused its addresses, a protected-mode guest without a resolver, a guest in another mode than the gate opened in, an address it cannot place, or a breakpoint it did not create at its addresses; or it stopped because a breakpoint request of its own failed, a resolver moved one of its addresses, or the session changed the guest after a failed continue request. Raised before a continuation, nothing was sent. |
WriteOutsideContract |
A write names no field in the contract, has no contract, or differs from the field's length. The run fails. |
WriteHashMismatch |
The field's bytes do not hash to the expected value. Nothing was written; the run fails. |
WriteReadbackMismatch |
The field does not hold the written bytes afterwards. The run fails. |
WriteFailed |
A write was refused for another reason, such as a guest that is not stopped. The run fails. The three errors above derive from it. |
RunFailed |
A write, continuation or step was asked for after the run failed. Not sent. |
RunEnded |
A write, continuation or step was asked for after the event log got its outcome. Not sent. |
ModuleRefused |
A module the event log names is not imported, or has no file to hash. Nothing was started. |
LogEntryRefused |
The event log refused an event or outcome, or has already ended. A refusal ends the log as failed and fails the run. |
LogRejected |
Reading an event log refused it; its subclasses are in the table above. |
CleanupFailed |
The emulator still ran after teardown; the lock was kept. |
All of them derive from SessionError.
Native verification
CI runs the tests against a stand-in emulator and a stand-in client, and against a stand-in rust-dos debug server, with no DOSBox-X, no rust-dos and no game.
DOSBox-X
Before each release that changes process, transport or drive handling, the owner runs this procedure on the pinned revision and puts its output in the pull request:
-
Clone
https://github.com/joncampbell123/dosbox-xat tagdosbox-x-v2026.10.01and check thatgit rev-parse HEADprintsb6abbd5980a885f5f310a4088c59a8688d1b116c. -
Build the
Agent Debug SDL2configuration for x64. With Visual Studio 2019 Build Tools (toolset v142) and Windows SDK 10.0.19041.0:& 'C:/Program Files (x86)/Microsoft Visual Studio/2019/BuildTools/MSBuild/Current/Bin/MSBuild.exe' ` <checkout>/vs/dosbox-x.sln '/p:Configuration=Agent Debug SDL2' /p:Platform=x64 ` /p:PlatformToolset=v142 /p:WindowsTargetPlatformVersion=10.0.19041.0 /m:2 /v:minimal
-
From this directory, with the package installed:
python native/verify_native.py --checkout <checkout> ` --emulator '<checkout>/bin/x64/Agent Debug SDL2/dosbox-x.exe' --run-directory <new directory>
The script generates a synthetic .COM program, starts it in an owned session under the
machine's run lock, sets an execution breakpoint, continues to it, and reads the registers there.
At the breakpoint it makes a guarded write to a word the program loads next, under a contract with
that one field, reads the word back and steps over the load. It passes when the breakpoint stop is
observed, AX and BX hold the values the program set, the write verifies, the readback holds
the new bytes and AX holds the new word after the step. It prints the checks, the checkout
revision, the emulator hash, the reported capabilities and the recorded writes.
rust-dos
Before each release that changes the rust-dos backend's process, transport or drive handling, the owner builds the rust-dos the package targets and runs, from this directory:
python native/verify_rustdos.py --emulator <rust-dos.exe> --run-directory <new directory>
python native/verify_rustdos.py --emulator <rust-dos.exe> --run-directory <another new directory> --deterministic
The script starts the same synthetic .COM program in an owned rust-dos session, continues to a
breakpoint, reads the registers, checks that the server refuses a write whose If-State-Seq is
stale, makes a guarded write, steps over the load and runs the program to its exit. It passes when
the breakpoint stop is observed, AX and BX hold the values the program set, the stale write is
refused with 409 and precondition: state, the guarded write verifies, AX holds the new word
after the step and the program exits with code 0. It prints the checks, the emulator's hash, the
server's version answer and the recorded writes.
Tests
python -m pip install -e packages/dosbox-session
cd packages/dosbox-session && python -B -m unittest discover -s tests -p "test*.py"
The session tests need Windows and git; on another platform they are skipped.
Metadata
Release files for dinorefurb-dosbox-session 0.7.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| dinorefurb_dosbox_session-0.7.0.tar.gz | 121.5 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| dinorefurb_dosbox_session-0.7.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 206.9 kB
Release files / dinorefurb_dosbox_session-0.7.0.tar.gz
| Download URL | dinorefurb_dosbox_session-0.7.0.tar.gz |
|---|---|
| Size | 121.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
b2f0377651d1bbe86d64fa09de848bbb602782374e431c794cf91bf8df2667b1
|
|
BLAKE2b-256 checksum How to use checksums |
ceddccdb88b449977996eb0bd1ba64759438c9250827a40923847d83e7b7a3f3
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 11, 2026.
Transparency logRelease files / dinorefurb_dosbox_session-0.7.0-py3-none-any.whl
| Download URL | dinorefurb_dosbox_session-0.7.0-py3-none-any.whl |
|---|---|
| Size | 85.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
db0d9cfcb4a78d05afd754bc048cd8638e98e055b2fcbca34b52e6d7c4514a83
|
|
BLAKE2b-256 checksum How to use checksums |
5511cbaa0cac5a5bd021c31ec6b0c899655660bcfdc68fa2c06cb1b6e9d5fdff
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 11, 2026.
Transparency log