The backend had a full dev story for everything except the one thing the
break loop reads: emit.ml pushes a shadow-stack frame on entry to every body
in a dev build and x86.ml pushed none, so an --x86 session answered backtrace,
locals, inspect and globals with "this program was not built with --dev",
which was false of it.
emit_fn now builds the same four words runtime/flan_dev.c describes -- the
previous head, a static descriptor, and a table of one address per slot, null
until the binding that fills it has run. The push is emitted at the head of
the body buffer rather than into the prologue's, because the prologue is where
the incoming registers still hold the arguments and every instruction of it
clobbers rax. The pop is one store at the epilogue label: this backend has one
epilogue and a return, the body's tail and the transfer exit all arrive there,
where emit.ml needs the same restore at five separate rets.
Two decisions are taken from emit.ml rather than reasoned out again, so that
the two dev builds answer a stopped frame identically. A function with no
named slot gets no table and reports no slots -- there the gate buys mem2reg
an alloca that does not escape, here it buys nothing but agreement, and
agreement is the only thing the break loop can check. And the descriptor's
bytes do not go through string_const: that counter is what decides whether an
expression thunk's module may be unloaded, and counting a descriptor in it
would pin every C-x C-e module's mapping for ever.
The descriptors are the one constant this backend emits that holds an address,
so they go in .data.rel.ro and not .rodata -- a relocation the loader applies
needs a section it may write, and a redefinition module is such an object.
Verified on an assembled module: R_X86_64_RELATIVE for the two pointers,
flan_frame_head through the GOT, no TEXTREL.
layout_ctx fills its globals table, which it never had to before. Reach's
ref_fingerprint asks whether a name is a global, and with an empty table no
name is one -- every frame would carry the hash of the empty set, the daemon
would recompute the real one, and the globals section would refuse a body
nobody had touched while its locals still read.
Release builds gain nothing: all of it is behind md.dev, and three corpus
programs emit byte-identical assembly before and after.
The new coverage stands up an --x86 daemon on dev-locals.flan and asks the
four questions, against the values the LLVM block above asserts of that same
program -- the claim is that the answers are the same and not merely
plausible. The existing --x86 two-process daemon gains one more: a backtrace
through a redefined body, which reports the installed body's own file, since
that descriptor travels in the module and a frame pointing at the host's would
name the body it replaced.
`arena-new` is a builtin, so it is in no program's symbol table and `defs`
never mentioned it — which made the editor answer "the running program
defines no arena-new" about a name that works. The fix is not a better
refusal: it is that the seventy-eight names the checker answers without
being told are now in the reply, under a kind of their own.
check.ml carries the table, beside the arms it describes, because a table
in another file drifts from them with nothing said. test_flan reads both
the arms and the table and fails on either having a name the other does
not, in both directions — there is no reflecting over a match, so it reads
the source.
Each entry is a signature in `signature_of_fn`'s shape and one line. The
arms that do not have one shape say what is true instead of pretending:
`?` for an argument that may be left out, `|` for the types an arm really
takes, and the checker's own predicate names for the type-directed ones.
The three user-allocator names carry a bare name and no bracket list,
because they are refused wherever they are written.
`defs` grows a fifth string for the prose, and builtins are appended last
so a completion table does not bury the names being worked on. A defn has
no docstring to put there and does not get one here: the Tast keeps none,
and that is a different piece of work.
The editor reads the kind, not a special case. `flan-doc--where` and the
xref backend both answer before their empty-location branch, because a
global's missing location and a builtin's absent one are different facts
and only one of them is about the daemon.
Three gaps in the first pass. The park flushes after the poll, which is the
only thing that gets a printing expression's line out of a fully buffered pipe
before the next run — now pinned by its position in the transcript. The globals
section under a parked break answers with everything skipped, which is a
different reply from the timeout a job nobody polled would give, so it is asked
for rather than assumed. And the render thunk's five-second sentence no longer
asks a parked program whether it is calling agent/poll.
The park waited on one flag and could do one thing, so C-x C-e on (+ 1 1) was
refused for want of a frame boundary — an expression that needs nothing from
the program, in a process holding every global the run left.
It waits on two now. A re-run leaves the park; a wake drains the agent's ring
and waits again, with the state still PARKED, which is what makes running the
thunk there exactly as safe as running it at a frame boundary: while parked
there is no concurrency to be unsafe against. The break loop is the precedent
and CL's spawned worker is deliberately not copied.
A thunk that stops now stops on a parked thread, so the restart ops refuse on
whether a break is engaged rather than on the state, and a paused expression
against the park is resumable.
[close] was the only way out of the accept loop that anybody ever took, and
an editor that is killed rather than quit never sends it. Four of the eight
orphans found on the author's machine were that.
The signal is the socket, not a new op: Emacs opens one connection for the
whole session and drops it only on paths that send [close] first, so an
editor left open overnight is an editor still attached and the grace cannot
accumulate under it. Armed only once a client has been there, so a daemon
still waiting for its first one is untouched. Parked sessions get the short
clock and live ones six times it, because a parked program is invisible and
a running one is a window somebody may be looking at.
Under --two-process the compiled program is a child of the daemon, and the
daemon is the only thing that ever ends it: the program is a loop with no
reason of its own to stop, and lib/dev.ml kills it from a Fun.protect finally
block on the way out. That block does not run when the daemon is SIGKILLed —
by a harness tearing down a daemon it gave up on, by a watchdog, by anybody —
and what was left behind was a program with ppid 1, sleeping, holding a socket
nothing will ever connect to. Eight of those were on the machine this was
written on, the oldest six days old, and one had been minted by this suite.
The child now arms PR_SET_PDEATHSIG when it starts its agent, gated on a
FLAN_DEV_PARENT the daemon sets and nothing else does, and closes the race
where the daemon died before the arming by comparing getppid against that pid
rather than against 1 — under a subreaper an orphan is not reparented to init.
The signal is SIGPWR and not SIGTERM, because the daemon's ordinary teardown
already sends the child SIGTERM and the child's stderr is the daemon's own; a
handler there would print "the daemon is gone" into the daemon's output on
every clean close. The handler writes one sentence, unlinks the agent socket
it bound, and exits 0, and it does nothing else because it can run on the game
thread inside stdio's lock.
A merged build arms none of this and must not. There the daemon is the
process, so a dead daemon is a dead program; its parent is whoever typed
flan dev, which is a shell or an emacs and not the session's owner. Merged
daemons do leak on this machine — four of the eight — but for a different
reason, which is that accept_loop has no idea whether its editor still exists.
That wants a mechanism that tracks the client rather than the parent, and it
is not this change.
The daemon's read end of the program's stdout pipe is also cloexec now. The
child used to inherit it, which is both a leak and the reason waiting for EOF
on that pipe could never have been the mechanism: the pipe always had a
reader. A printing program now takes SIGPIPE once its daemon is gone, which is
a second net under chatty programs and not a substitute — the test uses
dev-watch, which is silent, so that only the thing under test can end it.
The test spawns the two-process shape, sends one request so that what it kills
is demonstrably serving, reads the program's pid out of the daemon's
/proc children before killing it — it is a grandchild, so waitpid is ECHILD —
and SIGKILLs the daemon. SIGKILL and not SIGTERM: SIGTERM lets the finally
block run and the assertion would be green with this change reverted. It fails
that way, and passes with it, in well under a second.
Six refusals in the runtime called _exit(134) where every other error had
learned to park: no restart by that name, a restart taken with the wrong
arguments or with none, a defer that invoked one, a null allocator, and
free-all on something with no region. Under a merged flan dev the compiler is
in that process, so a program that named a restart nobody established took the
session down with it, which is the one thing the break loop exists to prevent.
They park now. Not through flan_break_hook, which is what bounds and
arithmetic use: that hook may answer by aiming a transfer channel, and these
six are called by emitted code that falls off the end with no channel anywhere
in the call, so a restart chosen against one would be accepted and dropped.
flan_trap_hook says the other thing instead — stop here, let everything be
read, and refuse the resume with a reason.
All six park, for two reasons rather than one. Four are guards that fire
before the operation they guard, so nothing is half done and the frame reads
like any other. The other two fire mid-transfer, with the frame's defers
possibly half run, and they park only to be looked at: stopping on a torn
unwind is strictly more than exiting before anyone can ask what tore it.
The break loop grew a per-snapshot resumable flag for it. Restarts are still
listed and still numbered, the terminal marks them untakeable and the socket
reports the same positions as unreachable, and the listener refuses a choice
with the trap's own sentence rather than the thunk-boundary one.
Standalone builds die exactly as they did: nothing installs the hook in a
program that did not import the agent, and the acceptance case for free-all
still wants exit 134 and the same message.
The review entry that asked for this named flan_exit_hook, which is normal
termination and not this at all; it is struck out with the correction.
The three named arms are not every way a step after the check can fail.
write_file makes the daemon's copy of the module's text before llc is called at
all, and a Sys_error from it went out through serve's guard with the session
already holding the declaration — the same stranded declaration under a
different exception. Dev.eval now restores on the way past anything that
escapes, and carries an accepted flag so that a module the agent has already
taken is never rolled back out from under the process.
That is also the seam the crash can be reproduced through. test_dev takes the
daemon's working directory away from it — its own, not the macro cache the two
failures above use, which cannot fail a build with no macro in it — evaluates a
defn, and then calls it. Without the restore the daemon lists probe-two in
describe and the call closes the socket: the null cell, the game thread, and
address 0.
Two seconds was measured on one warm machine and has no margin against the
6.8s cold build listening already records. The status and the absence of the
frame-boundary sentence are what discriminate; this bound exists only so a
timeout cannot be read as an answer, so it moves to 4.5 and says so.
The merged build's stdout is a 64K pipe back into the daemon's own process,
and the accept loop is the only thing reading it -- which it is not doing
while serve is answering a request. The two five-second waits for a frame
boundary now drain the pipe on every tick, so a program stopped inside fwrite
is one the daemon lets go rather than one it waits out and then accuses of
not calling agent/poll.
drain and not take: the text stays in the buffer until with_output puts it on
the reply, which is where the output an evaluation caused belongs. And the
drain sits beside the sleep rather than inside the select, because a readable
pipe would make the tick free and count the timeout out in a fraction of it.
dev-chatty.flan prints 4K a frame, which is the only fixture here that fills
the pipe at all; without the drain it fails in 5.1s with the old sentence.
The agent refuses a module it has no room to queue with "the program is not
calling agent/poll". For a running program that is the cause. For a parked one
it is confidently wrong: there is no game thread left to poll with, nothing
drains the ring until somebody runs the program again, and the ring is full
precisely because Dev.eval accepts evaluations while parked and promises they
install at the next run. At install sixty-five that promise breaks, and the
reply sent the reader to inspect a loop that is not running.
The agent cannot know this — parking is the merged shim's state — so the
daemon, which has just asked liveness, rewrites that one reply the way
abi_mismatch rewrites dlerror's. Every other refusal is still quoted in the
agent's own words, and a running program's is unchanged.
Pinned on Dev.refusal rather than end to end: filling the ring is sixty-four
real clang runs, and what the daemon does with the agent's words is the whole
of the change.
flan dev's merged build is the program and the compiler in one -rdynamic
executable, so it exports every flan.* body it has, and ELF gives it precedence
over anything dlopened afterwards. The compiler expands a macro by dlopening a
module into that same process, and the module is built by Emit.program whatever
backend the session uses -- so under --x86 the caller was LLVM's and the body it
landed in was the dev backend's, which is a crossed pair. It died with SIGSEGV
inside flan.[clamp] during the first expansion, before the program had run a
line, and Dev.start refused the combination rather than do that.
Build.macro_module now asks Emit.program for hidden visibility on the module's
own Flan definitions. There is nothing left for the host to interpose, and the
flan.macro.* thunks stay exported because dlsym is how the compiler reaches
them -- nm -D on the built module lists those three and nothing else of Flan's.
The -Wl,-Bsymbolic that had been binding everything locally since 65d14f4 goes
with it: the module links its own flan_rt.c, and binding that locally aimed its
calls at a runtime flan_rt_init never ran on, with a null flan_exit_hook, so a
trap raised inside an expansion would have exited the process instead of parking
it.
Nothing about the host moved, which is what keeps redefinition modules reaching
its cells, its globals and flan_dev_cell. hidden defaults to false, and the 540
IR files this compiler emits for the test corpus are byte-identical to the ones
before it.
test_dev.ml's assertion that the merged daemon refuses --x86 becomes the session
it was standing in for: dev-macro.flan calls a prelude macro at the top level,
so the daemon coming up at all is the old crash not happening, and one build
then carries C-x C-e, a C-c C-c whose body calls a macro again, the park and the
rerun.
--two-process has no parked thread to wake. Its program is a child, and a
child that finishes is gone, so running main again is not something that shape
can do — and saying "the program is already running" would send somebody back
to try again after it had exited. It names itself instead.
--x86 arrives here too and is the reason this is worth a round trip rather
than none: --x86 refuses the merged daemon, for the -rdynamic reason the case
below already records, so every x86 dev loop is a two-process one and this is
the answer it gets.
Flan's tagged sum has been spelled defunion since it landed, which was
accurate right up until the language wanted C's untagged union as well.
Both cannot be called the same thing, and the tagged one is the one with
an alternative name that says what it is: a case, its fields, and a tag
that steers which case is live is a data type, not a union.
So the form is defdata everywhere -- the parser, the AST, the checker,
both backends, the prelude's Form, the editor's font-locking and imenu,
the docs and every .flan file in the tree. The internal vocabulary moves
with it: Tast.union is Tast.data, uname is dname, the tables the checker
and the emitter keep are datas. Leaving them would have inverted the
words permanently, with surface defunion meaning one thing and
env.unions meaning the other, which is exactly the kind of drift the
comments in those files exist to prevent. What did not move is case,
variant and vfields: a tagged sum still has cases, and it still has one
live at a time.
defunion is not kept as an alias. An alias would compile the day the
untagged form lands and mean the opposite of what it used to -- the same
silent misparse that made defn's return type mandatory, and worse,
because the reader would have no reason to look. The old spelling is a
named refusal instead, parse/defunion-renamed, which says what it is now
called and that the name is reserved for something else. It fires on the
head alone, so (defunion U [A B]) -- which would otherwise have parsed
cleanly as one field A of type B -- is refused with the rest.
You run a program under flan dev, it opens a raylib window, you close the
window, main returns — and there is no way to get another window short of
flan-dev-restart-program, which throws away the build, the session and every
global with it. In Common Lisp or Clojure the image outlives main, so you call
it again. The process here already outlived main: the exit hook flushed, closed
stdout and sat in for (;;) pause(). Nothing could wake it.
So main() is a loop. The hook records the status and longjmps back into a
setjmp in main() — there is no return available, since flan_exit is reached
from wherever the program happened to be — and the thread waits on a condition
variable until the new rerun op signals it. The main thread is the one that
runs main again: a window belongs to the thread that opened it, and on macOS to
the first thread of the process. A longjmp pops no frame, so the park first
empties the handler stack, the restart stack and the shadow frame chain, each
of which was a chain of allocas in stack the next run is about to write over.
Nothing else is reset; the second run reads whatever the first left in the
globals, which is the semantics that was asked for.
Closing stdout had to go with it. That was how the compiler learned the program
was done, but a pipe delivers EOF once, so the signal and the program's output
were the same resource and spending it left the second run with nowhere to
print. The descriptor hazard the old code reopened /dev/null for goes away with
the close that caused it. Liveness is asked for instead, through a weak symbol
in the same style as the agent's, and is now three states rather than two: Live,
Parked and Gone. Every guard branches on that before consulting the break
state, because the agent's listener answers "running" while the program is
parked and telling somebody whose program has finished that it is running is
worse than saying nothing. Only eval accepts a parked program — it queues and
waits for nothing, and the queued module installs at the first frame boundary
of the next run, so a body can be fixed while parked and the re-run executes
it. Everything else needs a frame boundary or a stopped stack, has neither, and
says which, naming the command that gets the program back.
A re-run while the program is running is refused rather than queued: the test
and the signal happen under one mutex, so two mains writing the same globals at
once never starts.
:parked rides on every reply beside :stopped, for the reason :stopped does —
finishing is as unannounced as stopping, more so when the way it happens is a
mouse click on a title bar. Emacs shows flan:parked in the modeline and binds
flan-rerun to C-c C-M-x.
Item 3, and the reason the backend was written. Until now --x86 was read only
by flan build's argument list; the daemon built both halves through LLVM, so
none of this reached the dev loop at all.
The choice is a session setting, not a per-command flag, and it is spelled
exactly as [debug] already is -- one field on Session.t, set once in Dev.start,
carried on every change the session emits. session.ml's comment on [debug]
already gives the reason and it is the same one: the modules have to match the
process they are loaded into. Session.redefinition is the single place that
picks a backend, so the six call sites cannot disagree and the refusal has one
home. Session.change carries the answer beside the text, so the builder and the
text can never come from two different decisions.
There is no fallback and there must not be one. X86.redefinition refusing a form
is reported to the editor; quietly building an LLVM module instead is precisely
the crossed pair flan.abi.x86 exists to refuse at dlopen. A refusal reaches the
editor as a diagnostic like any other -- X86.Unsupported is re-raised as a
Loc.Error at the form it is about, because every caller already handles that and
none handled the other, and a session that died on the first unsupported form
would be worse than one that says so and stays up.
flan reload got the same flag at the same time. A command that could build a
module for a host the other backend compiled is how the crossed pair was
reachable from the CLI at all; the aggregate handoff's two-line reproduction no
longer has a second half.
And the finding: flan dev --x86 refuses the merged daemon. A merged build is the
program and the compiler in one process, and the compiler expands macros by
dlopening a module Build.macro_module made through Emit.program, cached on disk
by the macro source rather than by the backend. The merged host is linked
-rdynamic so a redefinition module can reach its cells, which also exports every
flan.* body it has -- so the macro module's own copy of a prelude function is
interposed by the host's. With an LLVM host nobody notices. With an --x86 host
the caller is LLVM and the body it lands in is this backend's, and the process
dies inside flan.[clamp] during the first macro expansion, before the program
has started. flan.abi.x86 does not catch it and was never meant to: a macro
module deliberately neither defines nor requires a marker. The honest fix is
hidden visibility on a macro module's Flan bodies, which changes the cached
object for both backends and wants a lane of its own. Until then the refusal
names the mechanism and the remedy, and --two-process has no such meeting.
start_merged keeps its --x86 plumbing, unreachable for now, because it is the
half that is right and will be wanted the day the macro module is fixed.
test_dev.ml drives an --x86 daemon through C-c C-c, C-x C-e, a literal, a new
defvar with a value of its own and a new defn, and asserts (twice fresh) is 82 --
which only holds if both registry lookups resolved. The merged refusal is
asserted there too. bin/main.ml learned to print a bare Failure as a sentence
rather than an uncaught exception and its backtrace.
Eight hit-and-run connections against the robustness daemon, then a check
that the session is still there and still knows what it installed. Fails
with 'killed by SIGPIPE' against the parent of this commit.
The daemon caught Loc.Error at each op and nothing else. That was survivable
while the frontend was the only thing that could refuse a form; it is not now
that expansion is part of evaluating. Both C-c C-c and C-x C-e run a clang
driver through Build.macro_module, which answers with an exit status and a
Failure, and a dlopen that finds no symbol answers with another one. Neither
is a Loc.Error, so neither was answered, and an exception past serve is not a
refused evaluation — it is a dead daemon with the program still on screen and
a closed socket waiting for the editor's next request.
The boundary is now one place, around the whole of a request, rather than a
new arm at each of the dozens of calls. Out_of_memory, Stack_overflow and
Sys.Break go through it: those say the process cannot continue, and answering
"error" to them would claim a session survived something it did not.
Everything else is about the form that was sent, and the message it carries
is the one the user can act on, so a clang exit status reaches :message
instead of being flattened to "internal error".
The session's own state goes with it. Session.eval wrote the imported macro
set above the checker, so a form that did not check left the session holding
a package's macros and none of its declarations; it is held and committed at
the bottom with decls, program and env. Session.eval_expr committed the
generic copies it had instantiated before emitting the module that carries
them, which is the session believing it holds a body nothing was written for;
that assignment moved below Emit.
Both are pinned. test_session drives the two rollbacks in process, and
test_dev drives a real daemon whose macro module cannot be built — the
expression path and the redefinition path, each followed by the same
evaluation succeeding and by the session still knowing the program.
Two loose ends.
The arena was invisible to memcheck. free-all is retain-capacity, so from
malloc's point of view nothing died and round two of a reset arena could read
a byte it never wrote, print round one's value, and draw no report.
flan_arena_proc now issues memcheck's MAKE_MEM_UNDEFINED over the whole
capacity beside its registry call. Measured on the same machine: the control
produced ERROR SUMMARY 0 before and 6 errors from 4 contexts after, with
--track-origins naming the client request. It is a control in
test_valgrind.ml now rather than a printed note.
The macro is vendored, not included, and the argument is measurement: the
machine that runs the sweep has valgrind and not valgrind-devel, so a guarded
#include would compile to nothing exactly where it matters and the control
would go quiet with no diagnostic. There is also nowhere to put an -I --
flan_rt.c is cat'd into an OCaml string literal and handed to clang in a
scratch directory. The __x86_64__ guard is load-bearing: the same runtime is
built for wasm32-wasi and emscripten.
Cost outside valgrind: 23 instructions on the free-all path only, about 1ns
per reset over fifty million of them, against a run-to-run spread wider than
the effect. Nothing on alloc, resize or free. valgrind.supp still holds no
suppressions; the corpus stayed clean across the change, which is its own
finding.
merged_serve's warning path deserved a test and has one. The discriminating
fact is not the log line but the policy: two_process kills its child and
fails where merged_serve warns and serves anyway, and nothing held that
second answer in place. dev-noagent.flan plus the last block of test_dev.ml
assert the session still answers describe after the wait runs out. Verified
by reverting the policy: the block reports rather than passing. It costs the
full ten seconds and there is no way to spend less. HANDOFF-f1.md is deleted.
The allocation registry had a recording side and half a reader. This is the
rest of the reader: point at any heap address, a breakdown by type, what is
still held, and the test that stops dev-ptr.flan's header from being read by
hand.
The recorded name, back to a type. The table records a string and has to —
the note is built where the concrete type exists and what crosses into the
runtime is bytes. What closes it is that the string is Types.to_string, which
is the source spelling, so the round trip is the language's own reader,
Parse.texpr and Check.resolve. No table of spellings is written down, so
nothing can fall behind Types.to_string, and a name that is not a type —
"pool slots" — is refused with the name quoted rather than defaulted.
The address root renders a (Ptr T) and not the pointee, which puts it through
render.ml's pointer arm: permission is asked in one place in the compiler, and
an address root and a slot root reach the same two answers by the same code.
Flan has no integer-to-pointer cast, so flan_dev_reg_addr is an extern beside
flan_agent_frame_slot, for the same reason.
One walk and two questions: a leak report is a breakdown with the dead left
out, so flan_dev_reg_by_type is one function and the agent formats it.
"At exit" is not a hook. A program killed by a signal runs no handler, which
is how a game under the editor ends, so (:op "leaks") is the authoritative
reader and can be asked at any moment including the one before the kill. The
atexit hook is for the program that returns from main, is registered from
inside flan_dev_reg_enable rather than by a file-scope destructor so that a
release build does not grow a third not-free place, and is off unless
FLAN_DEV_LEAKS is set because the acceptance table reads stderr.
The memcheck half of item 6 is deliberately not here.
FLAN_CLANG=/bin/false against test_repl: 'the daemon exited with status
2 before binding /tmp/flan-repl-dev.sock' in 45ms, where the old wording
would have waited out the whole timer and then blamed the socket.
Build.cachedir sat under TMPDIR, which dune makes private per run, so no
test run ever reused an object and every build in the suite was cold. It
moves to $XDG_CACHE_HOME/flan/objcache (FLAN_CACHE_DIR overrides), which
is safe because the keys are total: compile_c digests the source text,
the compiler's stamp and every flag; wasm_resource_dir digests the
builtins archive; compiler_object digests flan.cmxa and flan.a. Writes
were already .tmp-then-rename, so concurrent dune jobs are fine.
Macro.key was the one key that was not total -- prelude text plus the
call's forms, and nothing about the compiler whose codegen produced the
.so it names, which is dlopened straight back into this binary. Under a
per-run TMPDIR that never showed; under a durable cache it is a stale
expander that crashes rather than a compile error. It carries the
compiler's stamp now, handed across start_merged's exec in
FLAN_COMPILER_STAMP because a merged dev binary lives at a per-session
path and keying on that rebuilt a macro module every dev start.
Measured on dev-repl.flan, launch to bound socket: 2.0s cold against
0.48s warm. Whole-program flan build: 1.44s against 0.06s. Full dune
test 25.7s/30.1s before, 24.0s after, user CPU ~50s down to ~34s.
And the await: one timer covered two waits, a build then a bind, so
'the daemon never listened' was a wrong diagnosis of a build that had
not finished. listening now polls the process alongside the socket and
says which -- exited with a status, or still running and therefore still
building. A daemon that dies fails in milliseconds instead of costing
the whole timeout. Thirty seconds, down from a minute, because the build
it waits on is warm now.
Two follow-ups to the marking commit.
`Dev.eval_expr`'s new wait matched `Stopped _`, which fires on the first
iteration when the program is already parked on something else — the
break loop allows evaluating, so that is reachable — and answers for a
thunk that has not run yet, on a reply whose own `:condition` names the
other condition. It now waits for `Stopped "Pause"`, which the agent
reports under a nested break because `condition_name` is overwritten on
the way in and restored on the way out. `dev-pause.flan` grows a
`Missing` and a `boom` so the test can park the program on something
else first and tell the two apart.
And the flake NEXT.md had as "seen once and unexplained": `the daemon
never listened` is not a race, it is an llc-and-link of the whole
program before `flan dev` binds — ~600ms idle, measured at 6.6s and 6.8s
with the rest of the suite beside it, against a 5s and 8s await. All
three test binaries now wait a minute; the watchdog is what bounds the
run. Two consecutive full runs green.
Finishes DISCUSS.md §9's `pause` marking: the daemon half was already
built, this is the editor half plus the one daemon path it was missing.
`C-u C-c C-c` marks the form point is inside, `C-u C-u C-c C-c` the
top-level form (stop on entry), `C-u C-x C-e` the expression before
point. The buffer is never edited — the position rides beside the code
and the `(pause)` call goes into the tree after parsing, so no source
location moves.
`C-x C-e`'s path needed the daemon: its 5s `wait` answered "the program
did not reach a frame boundary", which is exactly what a thunk parked at
a breakpoint looks like from out here. `wait` is now three-way and asks
`state t = Stopped` only when a pause was requested, so the no-pause
shape `test_dev.ml` pins is unchanged.
The overlay is an annotation and not feedback, so unlike an error marker
it survives `pre-command-hook`; what takes it down is an accepted
evaluation with no `:pause` on it, which is the same thing that takes
the mark itself down.
Tests: a `test_dev.ml` block over the new `dev-pause.flan` that marks,
stops, re-evaluates plainly and then polls half a second confirming it
does not stop again — one sample after `continue` proves nothing, the
resumed frame is still in the old body — and an `emacs/test-flan-dev.el`
block for which form a prefix picks, the byte column, the overlay's
lifetime, and one live round trip.
PORTING.md Tier 1 item 5. The spy half of the watch was already built — the
pushed table, the buffer, the inline ghost text. What was missing is spy-num,
which is the part that item calls least obvious and most valuable, and it is
what this is.
A slot keeps count, min, max, last and mean. Each answers a question you can
ask without building a query: n is the first thing wrong when a loop is wrong,
the range is what one sample can never show you, last is what the scalar watch
would have given you, and the mean is a running sum divided at read time
because a mean accumulated as a mean drifts. A small ring of the last N
samples was the other candidate and loses — N out of 91,200 is a sample of the
tail of the loop rather than of the loop, and past five numbers every richer
answer is a UI for building a query.
The write path does no formatting, which is the feature rather than an
optimisation: a snprintf per sample at thousands a frame is a HUD that costs
more than the game. A sample is a load, five compares and the slot's seqlock;
the listener thread renders once per editor tick.
The window is since the editor's last tick, and that is a deliberate
divergence from watch.clj, where the stats are cumulative until reset-spies!.
Cumulative min and max reach the session's extremes within seconds of play and
then never move again, so the two most useful of the five go dead exactly when
you start interacting with the thing you are debugging — and this tool exists
to show you a number while you drag the mouse. Reset is its own message and
never a side effect of reading, because a destructive read makes looking
change what is there and anything that polls would shorten the window under
the editor that owns it. It bumps one epoch counter and clears no slot; a slot
clears itself on its next sample, so the reader never writes the table.
Ghost text needed one character. The call regexp allowed one hyphenated
segment, so watch-num-i64 backtracked to failure and a numeric watch got no
inline value while appearing normally in the buffer.
dune test is green, run twice. HANDOFF-f3.md carries the reasoning, the two
small gaps left behind it, and what did not work on the way.
test_dev's first block checked for agent.sock without having completed a
round-trip, so nothing put the check after the bind. merged_setup binds the
dev socket and returns; the main thread calls agent/start only after that, and
a connect lands in the backlog because merged_serve does not accept until it
has seen agent.sock itself. 2 runs in 8 failed.
Moving the describe up in front of the check supplies the ordering: a reply
arriving means merged_serve is past its own await, so it has already observed
the path. That closes the second race too — unlinking before merged_serve
looked made it wait out its full ten seconds and warn, which is the ten
seconds separating the passing runs from the failing ones.
0 in 10 after, and every run in the fast band.
Ported from the author's Clojure watch.el. Three of its decisions survive
unchanged — the program decides what is shown, the request is async, and the
paint is replace-buffer-contents so point survives every tick. The transport
is the part that had to turn round.
An eval here compiles a module and dlopens it, so the planned fix of
compiling the render thunk once and re-invoking it per tick was still a poll,
and a poll has a defect caching does not touch: a thunk runs at a frame
boundary and a stopped program has no more of those. So the program writes
into a table in flan_dev.c from its own loop and Emacs reads the table, which
is memory. The values are then as fresh as the last frame whatever the
repaint interval is, and they are still there while the program is stopped.
The frame thread's constraints decide the storage: no allocation, so names
are fixed arrays rather than strdup'd; no lock, because the reader is the
listener thread; and not the result buffer, which is written once per C-x C-e
and would be overwritten sixty times a second by watch traffic. One seqlock
per slot rather than one for the table, so a reader retries one slot instead
of having to catch the gap between two frames' writes; a snapshot from
adjacent frames is what a HUD looks like anyway.
Sixty-four slots, and past that a name is dropped rather than fatal — killing
the program because somebody watched a 65th value is the diagnostic shooting
the patient. Reported as a flag and not a count: the only number the write
path could keep is of write attempts, which at frame rate says "3847 names"
about one name.
Nothing writes the table until a watch buffer is open, so a watch call in a
program nobody is debugging is a load and a branch that is not taken — the
same number in a release build, since flan_dev.c is linked into both.
Scalars work today through declare-c against four runtime entry points, which
is why this needed no compiler change. A struct or a slice needs a walk over
its type, which is one arm in check.ml beside print; BUILT.md writes it out
rather than reaching into a file another lane holds. Ghost text turns out to
be gated on the same arm, for a different reason: nothing in the table
carries a source location, and a hand-written declare-c call cannot supply
one that does not drift when the line moves.
flan dev is one process, so the unix socket between the compiler and the
program was a connect, a write and a read that never left the machine's own
memory. A delivery is now a call into the agent's verb table.
flan_agent.c's handlers write into a sink that is either an fd or a growing
buffer, so serve() and the new flan_agent_request() share one verb table rather
than two that would drift -- every answer in there is assembled from several
pieces, so the seam had to be the writing. Dev.deliver, Dev.result and Dev.ask
were three copies of the same socket dance and are now three lines over one
Dev.request.
Which path is taken is the linker's answer, not a flag: flan_agent_request is
weak in dynload_stubs.c, null in every binary that links no flan_agent.o, and
Agent.request is None there. So --two-process, flan reload and every test go on
using the socket with nothing to configure.
Three things the direct path must not quietly change. The ring's room check is
separate from its store, which was safe only while the accept loop was the sole
producer, so handle_line runs under a mutex. A delivery is still only published
-- the install is one store per function on the game thread at a frame
boundary, and doing it on the spot would be a frame running half in the old
code and half in the new. And the OCaml runtime system is released across the
call, because a delivery is a dlopen.
Measured, and the number is the point: the transport was ~50us of a 23ms
redefinition, so end to end did not move. Code generation is 19 of the 22
milliseconds. The merge's prize was never latency.
The test is a deletion, because a reply cannot say which way it came:
test_dev.ml unlinks the agent's socket file once the merged program has bound
it, and every evaluation after that still installs.
A reply over the socket looks the same either way, which is the point of the
merge and also why nothing in the suite noticed it. So the daemon's pid is
checked directly: /proc/<pid>/exe is the merged program, at a path named for
the pid that built it, and there is no child at all.
The same check runs the other way round over --two-process, which gets one
round trip of its own — describe, an eval, and the output coming back — because
it is the escape hatch for a machine that cannot build the compiler object and
an escape hatch nobody exercises is not one.
Linux only, by /proc, and skipped rather than faked elsewhere: what is being
asked about is the process table.
The `inspect' verb had no coverage. The discriminating case is not a path
step, it is the frame: dev-inspect.flan gives `mark' to a global holding 99
and to a local of the OUTER frame holding a Point, so evaluating the name and
rooting at the frame answer differently and not even with the same type. One
`eval-expr' and one `inspect' of that name is the bug and the fix in a pair.
The slot index comes off the locals listing's fourth element rather than being
written as a literal, which exercises the field the editor depends on and
keeps the test from passing for the wrong reason if slot allocation shifts.
The rest is what a path can and cannot do: a struct field, an array element,
an option's payload and a union case's field — the last two having offsets but
no accessor form in the language — and four refusals, each checked for naming
the step and saying why. A `:path' of `nil' is read as the slot itself,
because Emacs has no other spelling for an empty list.
Two claims about the frame, since `stopped_frame' being shared is an assertion
about code rather than about behaviour until something proves it: the frame
whose body was redefined under it is refused, and so is the whole stack once
the program resumes.
The globals section attributed a frame by its slot fingerprint, which is the
wrong cut for it: a redefined body can name entirely different globals while
binding identical locals, so the check saw no change and the new body's
reference set went into the union under the old body's frame, with the frame
numbers beside an entry saying so.
So a second fingerprint. Reach.ref_fingerprint hashes the set of globals a body
names — sorted and deduplicated, because a reference set is not ordered, where
slot indices make the slot fingerprint order-sensitive on purpose — and it
travels the path the first one already cut: %fninfo, flan_dev_frame_refsig, the
agent's snapshot, the backtrace line, Dev.globals_op. Different means the frame
is skipped by name with its reason, and the rest of the stack still contributes.
Two numbers rather than one, because they are two facts. A frame whose slots
match and whose globals do not has locals that are perfectly readable and
attribution that is not, and a combined hash would make locals refuse a frame
with nothing wrong with it. locals still checks the slot fingerprint alone.
It lives in reach.ml because expr_refs is already the walk that answers what a
body refers to, and is the walk the union itself is built from. One consequence:
emit now reaches reach, which closes a cycle through Load if cimport calls
Build.cachedir, so the header cache spells the object cache directory itself.
test_dev.ml drives the exact case — a body that binds identical locals and names
untouched where the stopped frame names pressure. With the check disabled it
fails twice: the missing refusal, and untouched appearing under frame 0.
A global is program state a frame happened to touch, not part of it, so
nesting it under one implies an ownership that is not there and repeats the
name once per frame that reads it. One section instead, holding the union of
the globals every frame on the stack references — the compiler does the
choosing, since Reach.expr_refs already answers a body's reference set, and
listing every global a program has would bury the one that matters under the
prelude's PRNG state.
Each entry says which frames touch it, by the index the stack section already
numbers them with, which recovers what per-frame nesting would have told you
at no cost in duplication. Ordered by the innermost frame that touches it:
a deep stack makes the union large and proximity to the error is what puts
the likely culprit on top.
Simpler than locals, because a global is reached by name rather than by
address. Emit.redefinition writes a global the host has as external, so the
thunk binds to the program's own storage and nothing is asked of the stopped
thread — no dev-slot round trip and no not-yet-bound case to refuse.
A frame that cannot be attributed contributes nothing and is named in
:skipped; the union being incomplete and the union being complete are
different answers. The hole in that is stated rather than papered over:
slot_fingerprint hashes a body's slots, which is the right cut for locals and
not for this, so a body that names different globals while binding the same
locals is not caught. The test drives the case that is.
MANUAL.md also loses a stale paragraph claiming the fingerprint check never
fires with a failing test pinned to it. It fires, and test_dev covers it.
render.ml's output and emacs/flan-inspect.el's parser are the two ends of one
wire format, which is why the printer was left on the colon when the rest of
the corpus moved: shifting it alone would have broken inspection in the dev
loop without breaking a test that said so. They move together here.
The field list in the inspector is labelled with the dot too, which is the
spelling flan-inspect-step-expr already used to build `(.x b)' — the label and
the expression it stands for now read the same.
One case needed a guard the colon never did: `...' also begins with a dot and
is the renderer saying it stopped, not a field called `..'. A field name never
starts with a second dot, so one character of lookahead separates them.
The colon is not gone from the rendered grammar. An enum member is `:green' and
is a *value*, so the two are now told apart by the character alone, which is
the only thing that distinguishes them.
Also font lock, handed over with the same change: `:name' was the rule that
drew field labels, and with the colon belonging to keywords every label in the
corpus was left unfontified. `.name' is drawn as a constant, in both the places
it appears — the label in `{.x 1.0}' and the accessor in `(.x v)', which are
the same name.
render.ml's output is parsed by emacs/flan-inspect.el, which hard-codes
the colon when it reads a field out of a rendered struct. Moving the
printer on its own would break inspection in the dev loop without
breaking a test that says so, so the printer waits and moves with its
reader, in the Emacs lane.
The sweep could not tell a rendered *expectation* from a Flan *source*
snippet -- both are strings in a test -- so it converted both. The suite
named every one it got wrong, and those are back.
emacs/test-flan-dev.el:415 is the one edit inside emacs/: Flan source sent
to the daemon for eval, which the parser now refuses in the old spelling.
One label, in a fixture.
The script is in tools/ rather than thrown away, because two lanes are
writing Flan in the old spelling right now and their files need the same
pass at merge.
It works on forms, not on text: a keyword becomes a dot only where it sits
in a field-label position inside a brace, so an enum member in value
position, a map key inside an EDN string and a type-position {K V} are all
left alone. :keys keeps its colon -- it names no field.
The refusal for a frame whose body has been redefined underneath it did not
fire because four of its five hand-offs were never written. `Emit.fninfo` has
been storing `slot_fingerprint` in the last `i32` of every `%fninfo` all along;
`flan_dev.c` called that field `spare`, there was no accessor for it, the agent
never snapshotted it, the backtrace line never carried it, and `Dev.locals`
compared slot counts and nothing else. The handoff note's "every piece is
written and the refusal does not happen" was a guess, and the first step it
suggested — printing both sides of the comparison — could not have found it,
because there was no comparison.
So: `spare` becomes `slotsig` and gets `flan_dev_frame_slotsig`; the agent
snapshots it beside the slot count and puts it on the backtrace line *before*
the location, since the name is the one field that can contain a space and has
to stay last; `Dev.backtrace` parses it; `Dev.locals` compares it against
`Emit.slot_fingerprint` of the body this session holds and refuses by name when
they differ. No change to `emit.ml` — the value was already there.
The mechanism itself is right and stays. `slot_fingerprint` hashes every slot's
name together with the spelling of its type, so a rename that keeps the count
and the types — exactly the case this exists for — changes it. The count check
stays in front of it because its message is the more specific one.
The fingerprint stays off the wire. A hash is not something an editor can act
on, and the refusal says the fact in words: this frame's body was redefined
since it was entered, so its names no longer describe its values.
`test_dev.ml` gains the inverse and the control. A body that drops a `let` is
refused on the count, and `main` — untouched by the redefinition of `look` —
must still answer, which is the assertion that would catch a fingerprint that
never matched anything and made the verb useless while turning the suite green.
locals compares the frame on the stack against the body the session holds:
installing while stopped is allowed, so the two can be different bodies of
one function, and a rename that keeps the slot count pairs every name with
the wrong value. Emit.slot_fingerprint hashes each slot's name and type,
emit_fn puts it in the frame's static description, the agent reports it on
the backtrace line and Dev.locals compares it.
It does not fire. The test that drives it -- a redefinition that renames
every local of a function that is on the stack -- fails, and is committed
failing rather than deleted, because it is the only record of what is
wrong. Everything else in the suite is green; this one check is red.
It builds. See NEXT.md's handoff for where to look first.
The half the shadow stack was built for. A slot's entry in the frame is its
address, null until the binding that fills it has run, so "not bound yet at
this point" is a null and needs no liveness analysis. The daemon compiles a
thunk that renders the types it already knows -- Tast.fn.slots, with snames
beside them -- at the addresses the stopped program supplies, and reads the
text back the way C-x C-e does. Nothing is copied out, because a value with
no header is bytes with no meaning anywhere but in the program that holds
it.
That is render.ml's walk with its root changed, which is the pointer-rooted
thunk NEXT.md said this needed, and one new arm in the backend: a cast from
one pointer type to another, which emits nothing.
Only named slots are recorded. A recorded slot escapes and stops being
promotable, and the slots that would cost most are the ones with nothing to
show -- dotimes' bound, the temporaries min and max use, the walk's own
scratch. They are refused by name rather than shown under an invented one.
Recording every slot was built and timed and is inside the noise, so the
rule stands on what it shows.
Four refusals, each by name and with its reason: a slot nobody named, a
slot the program has not reached, a type the printer has no arm for, and
two whole frames -- an evaluation's thunk, and a frame running a body that
has been redefined since, where every slot index would be a guess.
Measured, minimum of nine runs: +61% on call-heavy code over globals
against +33% for the frames alone, 0.06% of a frame at 60fps.
plan.org has specified a shadow stack in the dev column since the beginning
and nothing had ever built it. A frame is four words on the calling
function's own stack: the one it displaced, a pointer to a static
description of the function, and two words reserved for its locals. The
name and the location travel on the frame, so a backtrace needs no debug
information, no symbol table, and nothing from the platform unwinder that
plan.org deliberately does not use.
The pop is at every ret, the landing block a transfer leaves through
included. That is the half that is easy to get wrong: a pop written only on
the normal path leaves a dead frame behind every handled error, and the
test takes five breaks and resumes all of them by transfer before asking
for two frames.
(:op "backtrace") answers from a snapshot the stopped thread takes, beside
the restarts and for the same reason, and marks which frames belong to the
program and which to the evaluation the break is inside. It is refused
while the program runs.
Measured, interleaved, three pairs of binaries: 29% on 600 frames of sand,
7.6% on a benchmark that is nothing but calls -- 32us per frame of sand, a
fifth of a percent of a frame at 60fps. An array with a stack pointer was
built and timed as the alternative and is worse on both.
