bytes-copy.flan came in with the lane that made (bytes s) allocate, and
it went into test_acceptance.ml and nowhere else. It is the one program
in the corpus that takes a block from an allocator, writes through it
immediately, and then takes another from an arena that is freed and
destroyed under it -- which is the shape both opt-in sweeps exist for,
and neither was running it.
Both, not one, because the two tools answer different halves: a copy one
byte short is a heap overflow ASan names, and a copy whose tail was
never written is an uninitialised read only memcheck can see.
Clean under both.
The trio the author decided on 2026-09-20 is now all built: def is CL's
defparameter — its initialiser runs on every daemon re-run, unguarded, so
an edited initialiser repaints the same storage on C-c C-c plus re-run —
defonce (Clojure's name for CL's defvar, per the author) initialises once
behind the .init~once. flag, and defconst stays the image.
One parse arm reads both forms; the difference is Ast.reinit, carried to
Tast.global's grerun. Emit.startup_plan gives a def no guard flag, and
Check.check_global lifts every def initialiser — zero and literal
included — into global/<n>, so the host's startup reaches it through the
function cell and a re-evaluated def swaps it (Session's def_inits;
Emit.redefinition declares the cell for a non-sibling target). The old
defvar spelling is refused with the rename and both compiling spellings,
and every program, test, doc and editor list is swept — except sand.flan,
the author's live WIP, whose seven defvar lines are flagged in FIX.org
and keep its three dependent tests red on this branch.
The INSERTIONSORT crash, all three rulings (FIX.org 2026-09-20):
- (bytes s) allocates a writable copy through the allocator surface —
context or (bytes s a), StorageExhausted with retry, a registry note in
dev builds (flan_bytes_dup, lowered like vec-new). (bytes-view s) is the
old zero-cost reinterpret, renamed, read-only by convention; every
in-repo reader swept over to it. (string b) unchanged.
- String constants were already read-only on both backends at -O0; now
pinned — bytes-copy.flan rows on LLVM/-O0/--x86, and dies_segv rows
asserting the write-through-view trap on both backends.
- A dev build installs a SIGSEGV/SIGBUS handler by the same dev-only
constructor slot that arms the registry: one line naming the address and
the innermost frame, then the trap-hook park — stopped, not dead, the
daemon serving. No agent: message and re-raise. Release builds untouched.
Pinned by trap_park over dev-segv.flan.
CLHS 4.3.6's update protocol, minus the user hook, on the dyn side's
defclass. Redefining a class used to be silent: a class is sugar for a
constructor defn, so the edit replaced a body and the instances already in
the program kept their old keys for ever.
Three pieces. A registry in flan_dyn.c holding each class's current slot
list and a generation, made only of interned kw_entry pointers so the
collector has nothing to trace in it and no root to push for it. A uint32
generation on the instance, fitted into the padding kind and mark leave in
front of len's alignment — sizeof(flan_obj) is 48 with it and was 48
without, and flan_dyn_obj_size is there so a later field that moves it
fails a test. And a registration thunk per reload, run by the agent
through flan_reload_call after the module's bodies are published: it has
to be a thunk, because the case this exists for is a class redefined and
not constructed.
Migration is lazy, at want_map, len's map arm and dyn_equal's. Slots kept
by name, gained slots nil, dropped slots gone, identity preserved, entries
rebuilt in the class's order so a migrated instance is indistinguishable
from a fresh one. Equality migrates both operands first, so it is over the
class as it is now.
The session had to stop refusing the constructor's signature change, and
does so only for a defclass and only when no compiled caller is left
behind. The checker gets there first in practice; the walk in eval holds
the reason locally rather than inheriting it.
The registry is advisory: a class instance is an open map, so a key a raw
put wrote that the class never declared is dropped by the next migration.
FIX.org says that plainly rather than pretending enforcement.
A defclass is a named dyn map with a shape tag, and a generic function
dispatches on it two ways: CLOS's, where the dispatch value is the class
of the first argument, and Clojure's, where a body computes it. They are
one mechanism and not two — a class dispatcher is (class-of arg0) as the
dispatch function, which is what lets a method written for the class
point and one written for the value :point be the same branch.
(defclass point [x y])
(point 3 4) ; the constructor, positional
(class-of p) ; :point, or nil for anything else
(defgeneric area [self] dyn)
(defmethod area point [p] (* (get p :x) (get p :y)))
(defmulti describe [x] dyn (get x :kind))
(defmethod describe :square [s] ...)
(defmethod describe :else [s] ...)
A slot is a key in the instance's own map, so get, put and has-key? are
how one is read and written and no operation was added for any of it.
What the class adds is the tag, and the tag lives in the object's header
rather than in a reserved entry — the queue's note said a reserved key
and this departs from it, because a key would be counted by len, walked
by the renderer and compared by equality, so every instance would answer
a length one larger than its slot count and print a key nobody wrote. A
header field cannot be reached by get or put at all, so no user key can
collide with it. It costs nothing: the map arm of flan_obj's union grows
to the size the view arm already had, and sizeof(flan_obj) is unchanged.
It needs no tracing either — the tag is an interned keyword entry, which
is immortal and is not a collector object.
The tag shows up in exactly three places: class-of answers it, equality
compares it (two instances of one class compare by their slots; an
instance and a plain map with the same entries do not, which is
Clojure's answer for a record beside a map), and both renderers print it
— #point{ :x 1 :y 2}, Clojure's own spelling.
None of the four forms reaches the checker. lib/classes.ml turns the
whole declaration list into ordinary defns at the top of build_program,
the way Shim.expand already turns a declare-c into a declare plus a
defn: a class becomes its constructor, a generic becomes one function
whose body binds the dispatch value and compares it down a chain, and a
method becomes a branch of that chain. It is a pass and not a macro
because a macro sees one form and the generic's body is not decidable
until every method is in hand — a method may be written above its
generic, below it, or arrive at a reload an hour later.
That last case is why the method bodies are inlined rather than lifted.
A generic is exactly one top-level name, so adding a method to a running
program is the ordinary redefinition of one function, through the cell
every call site already goes through. session.ml names the generic
alongside the method's own declaration name for that reason. The cost,
recorded rather than hidden: a method is not separately callable and is
not a frame of its own.
A dispatch that finds no method signals NoMethod, a prelude struct
carrying the generic's name and the dispatch value that missed. A
condition and not a trap, because a miss is something a program can be
written to answer, and handler-case around the call is the shape. Its
value field is dyn, the first condition here with one; the per-type
descriptor an item-2 struct carries is what the collector reaches it by.
No restart is established at the miss, which is BoundsError's decision
taken for BoundsError's reason.
Both backends, identically: the two new runtime entry points are
declared in emit.ml and the x86 backend needs nothing, since a dyn call
is a dyn call there. Deferred and written down in FIX.org: inheritance,
multi-argument dispatch, :before/:after/:around, named-slot
construction, unknown-slot checking, and computed dispatch values.
Ten test binaries share a directory and had shared nothing in it but
watchdog.ml. Everything else each one needed it wrote out again: the
failure counter and its FAIL line, the three-line report tail, a poll,
a socket connect, the wait for a [flan dev] daemon to bind, a substring
search, and the Load -> Check -> Reach.link front half of a compile.
[listening] was the clearest case. Three copies, byte for byte apart
from one comment, and two of them said in that comment that they were
kept separate because "these three files have no module between them".
That was not true when it was written: watchdog.ml was already named in
the same (modules ...) stanzas. test_support.ml is the second such
module, wired the same way, and those two sentences go with the copies
they were explaining.
test_repl.ml's [quote] was Wire.quote character for character, in a file
that already links Wire and already names Wire.quote in a comment about
what the case below it is checking. It is Wire.quote now.
One real behaviour change, and it is a fix. [connect] existed twice over
with different retries: the agent's narrowed to ECONNREFUSED with a
comment saying why -- the socket file appears at bind, a moment before
listen -- while dev's and repl's retried any Unix_error, which meant an
ENOENT or an EACCES was retried to the full timeout before raising
something the reader still had to interpret. The shared one takes the
narrow version. Every caller connects to a socket [listening] has
already seen on disk, so the race it does catch is the only one left.
The rest is left where it is, on purpose. The three output-capturing
[run]s differ in what they wrap -- a pid suffix, a sanitizer environment,
a valgrind invocation -- and are not the same function. The report tails
in test_repl, test_web and the two sweep binaries print different things
for different reasons. The per-file scratch prefixes are the feature that
keeps two suites running at once from unlinking each other's sockets, so
the shared helper takes the prefix rather than choosing one. And the
[match Sys.command "command -v clang ..."] probes stay as they are:
their skip lines are output this suite pins.
bin/main.ml has the compile pipeline written out twice more. Left alone
-- this was a test/-scoped change and bin/ should not be reaching into a
test module -- and noted in FIX.org as what it actually needs, which is
the pipeline moving into lib/.
dune test: exit 0, and its output is the same line for line once the
temp-directory hash and the millisecond counts are normalised.
flan_merged_park called flan_dyn_root_reset, which emptied the collector's
root stack. The frames' roots had to go — main is left by longjmp, so they
name stack the next run overwrites — but the dyn globals' roots are on that
same stack, pushed once by the emitted main and never popped, and the park
took them with the frames.
The park is not a quiet state. It services evaluated thunks, a thunk
allocates, and an allocation collects. So a program with (defvar config dyn)
answered (get config :s) with its string before any thunk ran and with nil
after one that allocated past the heap's floor — a read of memory the sweep
had freed, answering nil by luck of what the freed words decoded as.
The emitted main now brackets its global pushes: flan_dyn_root_globals_begin
empties the stack, the pushes go on, flan_dyn_root_globals_end records how
many of them there are, and the park resets to that line instead of to zero.
Nothing between the two allocates, which is what keeps the globals from being
swept in the window where they are unrooted — and [begin] emptying the stack
rather than adding to it is what makes a re-entered main re-root the same
globals rather than push a second copy of each, which also closes the other
half: a re-run used to re-push roots over slots left dangling by the park.
Both emitters, because the dev loop's default backend is x86 and a fix in one
lowering is not a fix. A program with no dyn globals emits neither call and
its root stack still resets to empty, which is what an empty push list should
leave behind.
flan_dyn_root_pop now clamps at the globals rather than at zero. An
over-popping frame eating the globals is the one way that clamp could turn a
miscount into this same use-after-free.
Covered twice. test/dyn_ops.c's park mode is the runtime's half — a run, a
park with a collecting thunk in it, and another run, three times over,
asserting both that the global survives and that the frame's five hundred
objects do not. Under ASan the old reset reports heap-use-after-free in
flan_dyn_tag with the free in gc_sweep; under memcheck it reports 24 errors
and still prints the right answer, which is the shape of the bug. test_dev.ml
drives the whole daemon over its socket on both backends against
programs/dev-dyn-global.flan.
Not touched, and it wants a decision rather than a patch: a re-run re-enters
flan_program_main, which re-runs the lifted startup function, so every global
with a computed initialiser is reset by a re-run. That contradicts dev.ml's
own note and FIX.org item 1. It is independent of this — the roots are right
whether or not the values are re-initialised.
Nor is this the reload path. A defvar added by an evaluation gets its storage
from flan_dev_global (emit.ml's new_globals, x86.ml's counterpart) and there
is no flan_dyn_root_push anywhere on that path in either backend, so a dyn
global added to a live session is unrooted. That is a separate defect with a
separate fix, and nothing here makes it better or worse.
A (Vec T), a slice or a fixed array crossing into dyn no longer refuses; it
is a view, one word in the box, over the container's own storage. Reads box
the element on the way out; writes tag-check the dyn value's tag against the
element type on the way in and trap, by name, on a mismatch, never coercing
or silently storing.
The open question the decision left — whether the descriptor points at the
container or snapshots pointer and length beside it — is settled by kind. A
Vec view holds the address of the Vec's own header (flan_rt.c's flan_vec,
restated in flan_dyn.c under the file's standing "if either table changes,
change both" rule) and reads ptr and len live on every operation, so a push
that reallocates cannot leave it stale: flan_vec_grow overwrites that same
header in place, and there is nothing captured at the crossing for the
growth to invalidate. A slice and a fixed array cannot grow, so a flat view
snapshots data and length once; pointing it at the value's own slot instead
would be worse, since a slot's lifetime is not the slice's.
The element set is i64, f64 and bool, not everything box already handles
typed-to-dyn. A string element's dyn form is a pointer into the collector's
heap, and a typed container's storage is arena or stack memory the collector
never scans — a wider set would let a write plant a live reference nothing
ever traces, which no care at the write site closes. (Vec string) and a
typed (Map K V) keep the "does not cross into dyn yet" refusal, now for that
reason.
flan_dyn.c gains a fourth object kind, OBJ_VIEW, and flan_dyn_len/at/set_at/
push and the printer each grow one branch for it beside the existing vec
one. A view's own stale-container check is the runtime's own spelling
(flan_trap, park-and-inspect) rather than flan_rt.c's rt_die, per the
duplicity doctrine; growing a Vec through a view calls flan_rt.c's own
flan_vec_push rather than re-implementing doubling and allocator adoption a
second time. (set (at target i) x) against a dyn target — a plain dyn vec or
a view alike — was a hole in the base dyn milestone rather than something
item 3 introduced; it is wired to flan_dyn_set_at here because a view's
writes needed it to exist at all.
Both backends: emit.ml and x86.ml both already passed a Vec or a Map to a
runtime call by address rather than by value; a fixed array crossing into a
view needed the same arm added in both, for the same reason — a copy would
view the copy and never see a write to the caller's own array.
test/dyn_ops.c drives the runtime directly with a hand-built Vec header and
a plain C array, ahead of any compiler involvement: reads, writes on both
element kinds, the tag-check refusal on every element kind, the range
refusal, and the push that grows and moves a hand-built header out from
under the view watching it. test_flan.ml turns the old "does not cross into
dyn yet" refusal into acceptances for Vec/slice/array, keeps it for a string
element and for Map, and adds the element-restriction refusal by name.
test/programs/dyn-view.flan is the compiler-level survey: a Vec view mutated
through both sides including the grow-and-move case, a fixed array's and a
slice's views, a bool Vec's view, and its own two trapping modes for the
acceptance rows to run against. test_sanitize.ml carries the survey's happy
path; test_dyn.ml's new refusals are the runtime's own.
Both directions of the boundary go through expect, the way every other
dyn crossing does. A dyn's tag decides which case an (Option T) becomes
on the way in; an Option's own tag decides nil or a boxed payload on the
way out. box_option/unbox_option build the same If-over-a-tag shape get
and map-remove already build for the same reason, reading an Option's
tag and payload with the raw Field access Render's structural printer
already uses — nothing new for either backend to lower. A literal
Some/None skips the runtime check entirely, since the checker already
knows which case it is.
A bare T has no None to become. The literal nil the checker can see is
refused right there, at compile time, in expect itself — the author's
decision to do both halves rather than settle for the runtime trap
alone. Everything one step removed from the syntax — a dyn that only
turns out to be nil once the program runs — reaches flan_dyn_need_i64's
existing DynType trap, unchanged; there is no dataflow in this checker
for it to be otherwise (see "Ownership tracking repealed").
(Some nil) is refused the same way: the literal at compile time, with a
message saying why nil and None would collide; a dyn that turns out to
be nil only at run time through the new flan_dyn_need_not_nil, which
traps by the same route flan_dyn_need_i64 does.
(Option (Option T)) does not cross either direction — boxing Some of an
inner None would box it as nil, indistinguishable from the outer None,
the same ambiguity (Some nil) is refused for. The type itself stays
legal on the typed side; only the crossing does not exist for it.
(Option dyn) needs no case of its own in the boundary code — the
payload is already dyn, so box_option/unbox_option treat it as the
identity — but it is not yet a value a program can hold anywhere. The
per-type-descriptor pass (M2 item 2) refuses it at every storage site
today, the same way it refuses (Vec dyn), because a struct's dyn fields
are marked by byte offsets and (Option dyn)'s payload has none. Item 4
does not lift that gate; it only makes the boundary already correct for
the day items 2/3 do.
expect grew a ctx parameter to build the fresh slot the two new
crossings need — every call site threaded through, one context
mismatch caught and fixed in check_fn's tail-expression case along the
way. var's None case grew a direct Dyn arm: None at a dyn want is nil
outright, with nothing to build.
nil-option.flan carries the crossings that succeed and ends on the
bare-T trap; some-nil.flan is (Some nil)'s run-time half, kept in its
own file the way dyn-boundary.flan is one trap per program. Both are
in no_fallback_slots and test_sanitize.ml: the new dyn temporary
unbox_option's tag test mints is rooted, and reads its Option's tag and
payload through ASan clean, --sanitize matching the unsanitized run
byte for byte.
The crux was never where to put a descriptor; it was how an instance finds
one. A bare struct on the stack has no header to hang a pointer off, and
giving it one would change the layout C interop agrees on, change the stride
of an array and change what embedding a struct in another costs. So it has
none. The instance never carries a pointer to its type and the collector
never derives one from the bytes: the pairing of an address with a descriptor
is made at the *push*, by the code that put the value there and therefore
knows its static type. That is the same trick the shadow stack has always
used, and it makes the stack case the easy one rather than the impossible one.
A descriptor is the size of an instance, a count, and a table of byte offsets,
emitted once per type as private static data. Flattened, not a graph — a
struct held by value contributes its offsets shifted by where it sits, and a
fixed array contributes its element's once per element — so nesting costs
nothing at run time and there is no recursion in the marker. The offsets of a
big array would be a big table, and that is capped with a sentence rather than
half of the repeat form item 3 will bring.
Four places a value of such a type can live, and all four are rooted: a frame
slot, a global, the temporary a call's by-value return is spilled into, and
the slot a condition that is not a place is evaluated into. The last two are
new and are the ones that were not obvious. A callee roots its dyn words and
pops them in its epilogue, so between the return and the caller's store the
only copy is a register, which a collector that finds its roots by address
cannot see; the same hole was open for a Flan call answering a bare dyn and is
closed here too. And a condition crosses as a pointer into the signalling
frame while a handler allocates, which is exactly what the original refusal
said could not be made safe.
dyn_roots grows into root_plan and both backends read it, which is what the
older note about one counter deciding both ends was always for. The aggregate
temporaries are pooled by type rather than handed out in mint order: a
positional supply that drifted would pair an address with another type's
descriptor, and marking arbitrary offsets off a base is corruption where a
missed root is only a bug. Pooled, the worst a drift can do is run out.
What is still refused is a dyn no static offset can reach — inside a typed
container, in a data type's payload or a union's members where the cases
overlay, or under an Option where the payload exists only beneath the tag.
A (Ptr S) and a [S] are deliberately not on that list: neither owns storage,
and the only storage this compiler hands out for such a type is a frame slot,
a global or a fixed array in one, all of them already rooted. That is what
lets a handler clause take its (Ptr Cond) and read a dyn payload.
test/programs/dyn-struct.flan is the evidence. It runs forty thousand rows
past flan_dyn.c's one-megabyte floor, so marks and sweeps really happen, and
it holds live values through them in all four places at once. It has teeth:
with the descriptor walk stubbed out of the marker, the kept vector's length
comes back 24 instead of 628 and its first element is a stale word. Clean
under ASan and UBSan, same output at -O2, -O0 and --x86. dyn_ops.c grows an
aggregate-root mode so the runtime half can be wrong on its own, with a
header word holding a bit pattern that looks boxed and is not a dyn slot.
--no-gc still refuses, and had to be told how: a struct with a dyn field is a
collected value even when no expression in the program ever has the type dyn,
because a zeroed one still has a word the collector is asked to mark.
dune test --force: green, 0 failures across every suite.
The report was that handler-case segfaults in a top-level global
initialiser. It does not, and never did. What crashes is one frame
further down, in a position that has nothing to do with startup:
(defn read-file [path string] dyn
(let [src (slurp path (heap-allocator))]
(defer (free src))
(read (as-slice src))))
slurp signals FileError, a handler further out unwinds, and the
transfer leaves this frame through its defers -- all of them. But src
was never written: the form that would have written it is the form
that transferred. free then reads whatever the stack held under that
slot, which at -O0 in a small program is zero and at -O2 is a live
pointer, which is why the same program looked like an optimiser bug
from one direction and a startup bug from the other.
return never had this. The checker splices the defers registered above
it and no others, and says so where it does it. The transfer exit took
the whole list, because it is one landing block per function and
nothing in the IR said where each defer had come into being.
So the count is kept. The first defer in a function mints an i64 slot
zeroed at the top of the body; each defer leaves a store of its own
number where it was written; and fdefers -- the transfer path's copy,
and only that copy -- tests the count before running each one. The
normal paths are untouched and still need no test. It is all in the
checker: what reaches a backend is a slot, a store and an if, so
neither emitter learned anything and the x86 one needed no frame of
its own.
test/programs/init-conditions.flan is the survey. The top half is the
part of the report that was never true: handler-case with its
condition firing and with its body completing, handler-bind, and a
restart-case, all four in a global initialiser, all four answering
what they answer anywhere. The bottom half is the part that was: a
defer below the signalling form, which must not run, beside one above
it, which must -- a fix that took the unregistered one off by taking
them all off would have traded the crash for a leak, and 302/2 is the
line that would catch it. Three acceptance rows, LLVM, -O0 and --x86.
The detector was valgrind, not ASan: reading a stack slot nobody wrote
is not ASan's bug class and it reported nothing on the broken binary,
while memcheck named the conditional jump in flan_vec_free with the
unwinding frame directly above it. The program is in both lists --
test_valgrind.ml because that is what saw it, test_sanitize.ml because
that is where the transfer exit's frames are already watched.
spec-conditions.md section 5 now says which defers a transfer runs.
The author's (defvar game-data dyn (handler-case (edn/read-file ...)
[(FileError [c] nil)])) works.
Every mark it left was an addition, and addition commutes, so a backend that
ran the defers outermost-first produced byte-identical output and the row that
was supposed to be watching the order could not have told. The claim was in the
comments and not in the numbers. cleanup.flan already had the device for this —
a shift rather than a sum — so the log here is a digit trace now, and the two
frames under a catch read 12 where a wrong order reads 21.
Rewriting the trace made room for the three behaviours that worked and nothing
pinned. A return inside a clause is an ordinary return from the function that
wrote the form, because that is where a clause runs: it leaves through the
function's own exit, runs the defer registered there after the two the unwind
already ran, and leaves the handler stack empty behind it, which the bare
signal that follows in main is the check on. A defer inside a clause is refused
for the reason every nested form is refused one. And a handler-case inside a
defer works, because a defer may not start a transfer that leaves it and this
one begins and ends its own.
The program is registered with the sanitizers, where the interesting failure is
not the heap but a handler or restart frame left on a stack pointing into an
alloca that has gone — an output comparison cannot see that until something
much later calls through it. It is clean; it was also leaking sixteen bytes out
of the vector main allocates to prove the allocator context came back, which is
the test's own litter and is freed now.
docs/PORTING.md ranked handler-case as one site handler-bind covers. It still
is one site, and handler-bind still covers it, but it is no longer the closer
translation: a catch block is assumed everywhere it is written to see the
locals around it, and only the clause that runs at the form does.
The handler clauses are lifted left to right rather than by List.map, whose
order is unspecified. Each lift names itself after the count already on the
list, so an order nobody chose would number the clauses of one handler-bind
differently between builds, and those names go into a redefinition module.
The dyn runtime gets a map object and an interned keyword, alongside the
vec it already had. {:a 1 :b s} is a map literal wherever a struct
literal isn't — the parser tells the two apart by whether the first form
in the braces is a .field symbol — and a bracket literal builds the
runtime's own vec rather than a typed array wherever a dyn is wanted, which
is what lets a map literal's values nest arrays and maps freely. get, put,
len and has-key? all learn a dyn-map arm alongside the typed-map one they
already had, and (keyword s) builds the same interned value a :foo literal
does, for a name that only exists at run time. nil is now a literal, the
dyn absence value that get answers for a key a map does not hold.
On the runtime side, flan_dyn.c gets an OBJ_MAP that shares the vec's
storage arm and doubles its accounting, a linear-scan intern table for
keywords that makes equality an identity compare, and structural map
equality by lookup rather than position. The marker traces a map's
interleaved keys and values the same way it already traced a vec.
edn/read and its callers move off the old (Option Value) union entirely:
a document is plain dyn now, sets are dyn maps to true, and arena-edn.flan
is retired along with the union it demonstrated. The acceptance suite's
edn-read and json rows were recaptured against the new shape, and a new
dyn-map.flan program exercises the map and keyword operations end to end,
including a 200k-iteration churn loop against a rooted map that runs
GC for real, across the LLVM, -O0 and x86 rows, and under the sanitizer.
Keywords are dyn everywhere an enum isn't expected, which changed what a
couple of existing checker tests actually see refused; both were updated
to the sentence the checker gives now rather than the one it used to.
The sweep's note said there was no Flan program that reached flan_dyn.c, so
the only sanitized run over the collector was dyn_ops.c's -- which pushes
its roots by hand. That note stopped being true when the dyn programs
landed, and it stayed in the file. The gap it left is the one that matters
for the backend lane just committed: a root the *emitter* forgot is a live
object swept, and no amount of C testing can see a mistake the compiler
made. dyn-vec and dyn-defer are in the corpus now -- the second for the
roots that come off on a transfer's path out rather than a return's -- and
so is p13-dyn-collect, which is the only program anywhere that allocates
past flan_dyn.c's one-megabyte floor and therefore the only one under which
a mark and a sweep actually run. Everything else in that list agrees with
ASan by never collecting at all.
p13 lives in spike/x86 because that is the lane that wrote it, so the
alias's deps grew a glob for that directory; it is in this sweep for what it
does and not for where it sits.
What this cannot cover, and the compiler says so itself when asked: there is
no sanitizer pass over hand-written assembly, so --x86 --sanitize is refused
by name. ASan sees the x86 lane's roots only from the collector's side of
the call, never as frame slots. p13 through --x86 under the @x86 sweep is
what stands in for it, and it is a weaker check honestly labelled rather
than a stronger one assumed.
--force @sanitize: clean, and non-empty, which the previous run was not --
an alias satisfied from cache prints nothing and reads exactly like a pass.
dune test --force still green.
Milestone 1 of dynamic-by-default, the runtime half: NaN-boxed values in one
machine word, a mark-sweep heap, and the operations over them.
A double is itself, which is what a language with a physics loop and a float
calculator in its corpus wants; everything else hides in the quiet-NaN space,
three tag bits and a 48-bit payload that is exactly an x86-64 user pointer.
The negative-NaN collision is answered by canonicalising every NaN on the way
in, which flan_rt.c had already decided was the right thing to print. An i64
past the payload goes on the heap rather than becoming a 48-bit integer with a
64-bit name.
The collector is mark-sweep and nothing else -- no generation, no barrier, no
free list -- because the answer to wanting it faster is to type the program.
Roots are pushed, not scanned: NaN-boxing makes a conservative guess wrong in
both directions, and flan_dev.c's frame chain is the precedent. A fixed ring
of the last sixty-four allocations is marked unconditionally, which closes the
window where an expression with two constructors in it can collect its own
first result before the compiler has rooted either.
A type mismatch traps rather than aborting, through a flan_trap exported from
flan_rt.c so it takes the same path the six existing traps take: parked for
inspection in a dev session, dead where it stands otherwise. The sentence
names the operation, both tags as words, and both values.
flan_dyn.c is its own translation unit and nothing in the release runtime
names a symbol in it, so a program with no dyn operation links no collector
and --no-gc can be file-level selection rather than an argument with the
linker.
docs/SPIKE-DYNAMIC.md carries the argument. test/dyn_ops.c drives every
operation and all twenty-four refusals from C, the way dev_limits.c does,
including a million allocations against a hundred live and the control that
says an unrooted object really is reclaimed.
vendor/json is vendor/edn's shape with one decision reversed. edn never
allocates, so its tokens are views into the source buffer and escaped
strings are refused for want of anywhere to put the unescaped copy. This
one has an allocator, so it unescapes, and to unescape it copies —
string-of is the only function in the package that allocates, and it
copies even when there was no escape to resolve, because a Value whose
lifetime depended on which bytes happened to be in it is not a contract
anyone can hold. Odin answered the same question the same way:
tokenizer.odin allocates nothing, parser.odin's unquote_string does the
copy, and it clones in the no-escape branch too.
What that buys is at the bottom of test/programs/json.flan, which is
programs/edn.flan and programs/arena-edn.flan in one file because for
JSON they are one claim. The source buffer is overwritten with `?` bytes
while the document is live and the strings read back afterwards are
still the strings. arena-edn's header has a section admitting it cannot
do that.
Strict JSON and not Odin's JSON5 default, and the difference is where
most of the refusals come from: comments, single quotes, +1, .5, 1.,
0x1f, 01, NaN, Infinity and unquoted keys each get a sentence naming the
dialect they belong to, rather than one shared unexpected-byte. A lone
surrogate is refused too, and that one is forced rather than chosen —
rune-size answers None for the whole D800-DFFF block, so encode-rune!
would write nothing and the character would vanish.
The tokenizer refused #{} because "it needs a hash set to even
represent" — which is a claim about a reader, and a tokenizer represents
nothing. #{ now pushes } on the same balance stack { does, there is one
new token kind and no new closer, and err-set is gone rather than kept
with a message it no longer earns. skip-value needed nothing: it is
written against the depth and not against the kinds.
The dynamic reader moves out of test/programs/arena-edn.flan and into
vendor/edn/read.flan as (edn/read bytes), answering an (Option Value)
against whichever allocator the caller bound. Two decisions are written
down where they are made:
* a set is a Value.Set holding a deduplicated (Vec Value), because
(Map Value bool) does not typecheck — keyable refuses a key holding
a Vec or a Map — and restricting elements to keyable Values would
refuse #{[0 0] [1 0]}, which is the file this was built for. Insert
is O(n) against a structural value=?, so building the tileset's 54
pairs is 1458 comparisons, once.
* a Value copies every string into the allocator where a Token stays
a view. A view handed back out of the function that owns the buffer
is a dangling pointer, and free-all would not even take it. Odin's
json parser clones for the same reason.
An imported defdata was a refusal in load.ml — "not implemented yet
(milestone 4)" — and it had to go first. It is the type's name plus the
Type. half of a constructor symbol, which arrives as a Var node when the
case has no fields and a Struct node when it has; a match pattern needed
nothing, because a case resolves against the scrutinee's type and was
never a top-level name. programs/pkg-data.flan is that on its own.
programs/edn-read.flan reads assets/edn/tileset.edn, which is the
editor's real output: :texture-path and a :selected-cells of 54 integer
pairs, with no type declared for any of it. It also overwrites the
source buffer in place after reading and prints the document back, which
is the copy contract asserted rather than described.
The lo <= hi test in check_slice and slice-from-ptr's n >= 0 sat behind
--no-bounds-checks in both backends, while the comment beside each said
they could not be dropped. They are not bounds checks: hi <= len asks
whether a range fits inside a length, and lo <= hi asks whether the word
about to be written into a %slice's length field is a count at all. The
first stays behind the flag, the second is now emitted everywhere, the
way flan_vec_as_slice has always validated its own l > h in plain C.
emit.ml emits two signal blocks rather than one and i1, so an unchecked
build carries one compare. x86.ml keeps all three frame temporaries
stored outside the flag and gates only the second compare, because the
third is the length the message prints.
The IR assertion in test_acceptance now says the two slice calls are
present under --no-bounds-checks rather than absent, and the same build
is run: case 2 and case -2 of bounds.flan must still die.
Every number-to-text conversion wrote into one file-static in the runtime and
answered a slice over it, and nothing copied. Two of them in one expression
printed the second number twice — no crash, no diagnostic, and nothing a
sanitizer could find, because every byte read was inside an object that was
alive. The wrong object.
The buffer is now the caller's, one frame slot per call site. The slot is
allocated in the checker rather than in either backend: a slot is a
function-lifetime location in both of them, where an x86 backend temporary is
bump-allocated and reclaimed at the end of the expression that made it — which
is the one lifetime a returned slice must outlive. Each backend gains one
pointer argument and no reasoning of its own, which is what keeps them
symmetric.
The static is gone rather than left unused, since a buffer with nothing but a
comment beside it is a loaded gun. What remains is the ordinary lifetime a
pointer into a frame has: storing one of these slices in a container that
outlives the frame, or returning it, is still a copy the caller has to make.
NEXT.md's sharp edge now says that instead of what it used to say.
Both corpora are explicit lists and not globs, so a program added to
test/programs is covered by dune test and by @x86 and by nothing else until
somebody types its name here. files.flan, math3.flan and time.flan are typed.
time.flan is the one with something to say. getenv hands back a slice viewing
the process environment and never a copy, which is the exact shape a
use-after-free or an off-by-one length would be, and neither ASan nor memcheck
had ever seen it. files.flan brings three more path buffers through
flan_path_cstr. math3.flan is the cheap one and is here for completeness.
files.flan makes and removes its own tree, so the sweeps' two runs of it see
the same directory both times.
@sanitize is clean with all three in. @valgrind is not run here -- it is tens
of minutes and opt-in -- so those three entries are checked by the next person
who runs the alias.
A (Vec Value) where a Value may itself hold a (Vec Value) — the recursive
dynamic value an EDN reader has to answer with when nobody hands it a target
struct type — was refused five different ways, and every one of the five gave
the same reason: the container runtime is type-erased, so it copies and
releases slots bytewise and cannot reach inside a slot. A free would release
the slots and leave every block they point at stranded.
That reason is about teardown, and it does not hold for a region. free-all
never releases an individual slot; it takes the whole arena, and every block
the elements own is in it, because they came out of it. The refusals were
over-broad, and what they were guarding was never ownership — ownership
tracking is untouched here, moves are still moves, and Types.is_move_only is
the same function it was.
So the question moved rather than disappeared. It could not stay at the type,
because can-free is a capability on an allocator value and with-allocator
rebinds a dynamic variable: which tier a (vec-new) will meet is not a property
of the place its type is written. What is decided at compile time is only
whether to ask, which is a property of the element type; the answer is a
run-time branch on the allocator, one per container and never per element,
because the alternative is a walk at release and a walk at release is the
registry of destructors the frame tier's reset exists to not have. It is
emitted at every growth and not only at the construction, because ZII means a
container can exist without ever passing through (vec-new) — a case field left
out of a literal, a global that starts zeroed — and those adopt the context on
their first push.
free on such a container is refused rather than made quietly shallow. It cannot
recurse, which is the whole premise, and releasing the outer block alone would
be "I freed it" written over a program that stranded everything inside; this
runtime refuses that collapse everywhere else. The message names free-all,
which is reachable by construction. clone stays refused for a reason the region
does not dissolve, and the old message had bundled the two failures under one
sentence: what disqualifies clone is not that it copies a header — so do at and
get, and they are fine, because they promise nothing — it is that clone
allocates a new block and promises independence, and a bytewise copy hands back
elements still pointing into the original's region.
A struct or union field is admitted only where the field's container holds
owning elements, because that container can only have been built against a
region. A field holding a plain (Vec u8) stays refused: nothing would force
that one into a region, and two copies of the aggregate would be two headers
over one heap block. vec-in-struct.flan still pins that.
The epoch already covered use after free-all, including the case this makes
reachable — an inner header copied out of an arena-held element into a local
still traps, because an Allocator is a pointer and a copied-by-value one would
carry its own epoch.
arena-value.flan builds the value by hand; arena-edn.flan reads a real document
through the tokenizer, and its reader takes no allocator and names none,
because spec-memory.md already puts the allocator in the calling convention.
arena-region.flan is the branch itself: run 0 is the (Vec (Vec i32)) control
that must not trap, and runs 1 and 2 are the two ways this dies.
Load.program takes forms: it reads the import forms, resolves them with the
one resolver it always had, and parses the file with the packages' macros in
front of it. The refusal said this needed a second import resolver at the Form
level. It did not notice that the file being compiled is parsed before Load
runs too, so no shape of the feature could have left import resolution where
it was.
Names arrive qualified, as a defn's do. (mac/twice 4) is a call and (twice 4)
is an unknown name.
Stopped mid-task: dune test was never run and the acceptance wiring is
unfinished. HANDOFF-macros.md has what is left.
handles.flan joins the ASan and memcheck corpora; pool-stale-region.flan
joins memcheck as a seventh program that aborts by design, for the reason
the other six are kept — a trap that stopped firing would be silent. Clean
both ways.
Loading a package kept one table, keyed by real path, and used it for two
different questions. Already loaded meant "skip", which is right for the second
route of a diamond and wrong for a ring: a package that imported itself round a
chain met its own entry, contributed nothing, and appeared to work. The comment
said so and called it a feature.
It is not one. A ring has no package order, and a definite package order is what
the macro expander needs — every defmacro has to be compiled before anything
that calls it. So the chain currently being read is now carried separately from
the set already finished. A directory found in the first is a cycle and is
refused; a directory found only in the second is still the diamond's second
route and still a no-op.
The refusal names the ring — a -> b -> c -> a — and only the ring, not the route
that led to it. "There is a cycle" leaves the reader to find which three imports
it was.
pkgs now comes back dependencies-first, which is the topological order the
acyclic rule buys. The declaration list is left alone: check.ml collects every
top-level name before it checks any body, so declarations are order-independent
by construction and sorting them would be churn in the field every test reads.
The tests are a real tree rather than a second copy of pkg-shared. pkg-diamond
builds a shape/Box inside area/ and hands it to a function declared inside
draw/, which only type-checks if the bottom package was read once — two copies
of one struct are two types. What proves it is the numbers, not the compile.
The mutation pass turned up one defect that did not make the suite go
red: a reader branch that forgets to advance reads the same character
for ever, and dune test waits as long as it is left to. In CI that is a
job killed by the runner with nothing named and no output to read.
watchdog.ml puts an alarm on every test binary — generous, because an
alarm that fires on a slow machine is a flake — and a five-second one
around each read in test_flan, where the budget really is small. The
first read that does not return wedges the rest, so a looping reader
costs five seconds and names the row instead of costing eight minutes
or never finishing. Both were watched: the string-escape loop now fails
in five seconds with the case named, and the per-binary backstop was
armed short and observed to fire.
A negative index into a global is silent in bounds.flan, which is
measured. "Because a global has no left redzone" was the explanation
put on it, and it does not survive the obvious test: declare another
defvar in front of arr and arr[-1] is caught, landing in that global's
right redzone. Underflow detection is a question about what the linker
put in front of the object, not about the access. Corrected in
test_sanitize, BUILT.md and NEXT.md.
NEXT.md's entry also goes back to its stated size. It had grown to 78
lines saying what BUILT.md says in the same commit range -- the
attribute, the -O0 decision, the bounds.flan table -- which is the
half-build-log the file's own header warns about. What stays here is
what is next: the UBSan gap as an undecided compiler question, the four
daemon-path buffers the corpus never reaches, and Valgrind.
NEXT.md's queued section becomes a landed one. The headline is not the
flag: ASan reaches Flan code only because Emit now attributes every
define, and UBSan reaches none of it and has no lever that would, so the
shift-UB and float-cast items that section listed are still open and are
a compiler feature rather than a flag.
The clean result is written with its reach. println.flan pushes a
1100-character string through escaped[1024] on purpose, so that buffer
is genuinely covered; scratch[64] never sees more than 20 characters;
and the 4K result cap, the dev registry guard, SNAP_MAX/SNAP_NAMES and
condition_name[128] are on the daemon path and not in the corpus at all
-- read, not tested. Two defects fixed, both found by reading. Three of
bounds.flan's six out-of-bounds cases caught with the checks off, with
the other three tabulated and explained, and the caveat that ASan sees
out-of-object and not out-of-subobject access, so three of six is a
ceiling and not a measurement.
BUILT.md gets the durable half: the attribute, the absent UBSan lever,
why --sanitize does not force -O0 when --debug does, and the -O0/-O2
divergence that earned it.
(slice s 2 1) has length 2 - 1 - 2 = -1. flan_bytes_to_i64 and
flan_bytes_to_f64 both wrote their clamp as (size_t)n < sizeof buf - 1,
and (size_t)(-1) is 18446744073709551615, which is not less than 511 --
so k took the cap and the memcpy copied 63 or 511 bytes out of a
five-byte string constant. ASan calls it a global-buffer-overflow in
flan_bytes_to_i64; the regression case is in test_sanitize.
Every other (ptr, len) entry point in the runtime already guarded the
negative case -- flan_write_stdout tests n > 0, flan_escape_bytes and
flan_dev_emit both fold a negative length to zero -- so this was two
exceptions rather than a missing convention. A checked build traps on
the reversed slice before reaching either, which is why it took an
--no-bounds-checks run to show.
Also clamps the three snprintf shims that publish scratch as a slice.
snprintf returns what it would have written, not what it did, so a
format that overran the 64-byte buffer would hand out a length past its
end. No format here can: %g is 13 characters and %lld is 20. Found by
reading, and the sweep could not have found it -- nothing in forty
programs prints a number that long.
Twenty-eight programs built twice -- once plain, once sanitized -- and
compared on output and exit status, plus two positive controls that are
the only reason a clean result means anything: an out-of-bounds read
that must report, and a shift by the width of the type that must not,
because UBSan cannot see hand-written IR and this file would otherwise
be claiming coverage it does not have.
Its own alias rather than dune test. A sanitized program is a statically
linked 1.8MB binary and takes tens of seconds to link; the sweep is nine
minutes against the existing suite's seconds, and a test nobody will
wait for is a test nobody runs. dune build --root . @sanitize.
The checked sweep is clean. The unchecked variant -- ASan alone, with
Flan's own bounds checks off -- catches three of bounds.flan's six
deliberate out-of-bounds cases and is listed with why for the other
three: a global has a right redzone and nothing to its left, so arr[-1]
is invisible; a read past a string constant folds away entirely at -O2
and is caught only at -O0; and a reversed slice reads nothing at all.
ASan is not a substitute for the bounds checks, and now there is a table
saying which half it covers.
