flan/lib/dev.ml
Joseph Ferano 52d3898116 Four ways C-c C-c could lie, found by trying a defconst
Asked whether a defconst could be redefined, probed it, and got ":status ok"
for a change that did nothing at all - the module was built, delivered,
installed, and the program went on using the old value. That is the
silent-wrongness class the house rule exists to prevent, so it is now four
refusals and a fix.

A defconst's value is folded into its call sites - into an array length at
worst, which is decided before any type resolves - so it lives in the running
program's code and not only in its storage. Refused. A defenum member is the
same thing: :space is erased to an i32 literal in the caller. Refused, and
compared over declarations rather than over Tast.program, which carries no
enums at all for exactly that reason.

A defvar's initial value is deliberately not refused. Its storage holds live
state the program moved past long ago, and refusing to change the initialiser
would be refusing "edit the code, keep the sand". Same Tast.global record as a
defconst, opposite answers, told apart by gconst.

The value comparison is structural and conservative - anything it does not
recognise counts as changed. Comparing emitted text would be wrong, since
Emit.const on a string allocates a name off a per-module counter and two
different strings in two throwaway modules both come out as @".str.0".

Third: a new global's declared initial value was being dropped. flan_dev_global
callocs, so (defvar n i64 42) added at run time was silently zero. It now takes
the initial value as a blob, copies it on the allocation and ignores it
afterwards - the second half being where "a reload must not reset the program's
state" lives. In the allocation path rather than a branch at the call site, so
it cannot be got wrong at one of them.

Fourth: a change with no body to publish and no storage to allocate now answers
"nothing to install" instead of shipping an empty module. That is what the
defconst probe actually did, and it cost the program a frame's worth of reload
it did not need.
2026-09-10 22:36:55 +07:00

220 lines
9.0 KiB
OCaml

(** [flan dev]: one long-lived session, the program it belongs to running
beside it, and a socket an editor talks to.
This is the piece between an editor and everything else. What it adds over
[flan reload] is that the session *persists*: a [defvar] added by one
evaluation is part of the program the next one is checked against, and the
set of names the running process was built with is the one from the build
this daemon actually made. A CLI that rebuilds its session from source each
time cannot have either.
It owns the build, which is what makes its layout rules mean anything: a
session's struct layouts and global types describe the memory of a process
only if it is the session that compiled it. So the daemon launches the
program rather than attaching to one. *)
type t = {
session : Session.t;
child : int; (* the running program *)
agent : string; (* where it listens for modules *)
dir : string; (* modules are built here, one per eval *)
mutable n : int; (* dlopen caches by path: never reuse one *)
}
let await ?(ms = 5000) f =
let rec go ms =
if f () then true
else if ms <= 0 then false
else begin ignore (Unix.select [] [] [] 0.005); go (ms - 5) end
in
go ms
(* ── Delivery ──────────────────────────────────────────────────────── *)
(* The agent answers "ok" when it has queued a module, and anything else is a
refusal with a reason. Reporting that back rather than swallowing it is what
keeps a failed delivery from looking like a successful evaluation — the
whole class of bug this socket makes possible. *)
let deliver t path =
let s = Unix.socket Unix.PF_UNIX Unix.SOCK_STREAM 0 in
Fun.protect
~finally:(fun () -> try Unix.close s with Unix.Unix_error _ -> ())
(fun () ->
Unix.connect s (Unix.ADDR_UNIX t.agent);
let msg = path ^ "\n" in
ignore (Unix.write_substring s msg 0 (String.length msg));
let b = Bytes.create 1024 in
let buf = Buffer.create 64 in
let rec drain () =
match Unix.read s b 0 1024 with
| 0 -> ()
| n -> Buffer.add_subbytes buf b 0 n; drain ()
| exception Unix.Unix_error _ -> ()
in
drain ();
String.trim (Buffer.contents buf))
let alive t =
match Unix.waitpid [ Unix.WNOHANG ] t.child with
| 0, _ -> true
| _ -> false
| exception Unix.Unix_error _ -> false
(* ── Ops ───────────────────────────────────────────────────────────── *)
(* Every reply is a plist with a :status, so an editor can dispatch on one key
and never has to guess whether a missing field means failure. *)
let ok fields =
"(:status \"ok\"" ^ String.concat "" (List.map (fun f -> " " ^ f) fields) ^ ")"
let error ?loc msg =
"(:status \"error\" :message " ^ Wire.quote msg
^ (match loc with None -> "" | Some l -> " :loc " ^ Wire.quote l)
^ ")"
let eval t ~code ~origin =
if not (alive t) then error "the program exited; restart flan dev"
else
match Session.eval ~origin t.session code with
| c when not c.Session.installs ->
(* Accepted into the session and nothing to send: a declaration the
program already has, with no body and no new storage. Saying "ok" and
shipping an empty module would report success for a change that cannot
have taken effect. *)
ok
[ ":names " ^ Wire.strings c.Session.names; ":fns ()";
":note " ^ Wire.quote "nothing to install" ]
| c ->
t.n <- t.n + 1;
let out = Filename.concat t.dir (Printf.sprintf "m%d.so" t.n) in
(match Build.shared ~opts:{ Build.default with Build.dev = true }
~ir:c.Session.ir ~out () with
| timing ->
(match deliver t out with
| "ok" ->
ok
[ ":names " ^ Wire.strings c.Session.names;
":fns " ^ Wire.strings c.Session.fns;
Printf.sprintf ":ms %.1f"
(timing.Build.llc_ms +. timing.Build.link_ms) ]
| reply -> error ("the program refused the module: " ^ reply)
| exception Unix.Unix_error (e, _, _) ->
error
("cannot reach the program on " ^ t.agent ^ ": "
^ Unix.error_message e))
| exception Failure m -> error m)
| exception Loc.Error (l, msg) -> error ~loc:(Loc.to_string l) msg
let describe t =
ok
[ ":fns "
^ Wire.strings
(List.map (fun (f : Tast.fn) -> f.Tast.name)
t.session.Session.program.Tast.fns);
":globals "
^ Wire.strings
(List.map (fun (g : Tast.global) -> g.Tast.gname)
t.session.Session.program.Tast.globals);
":alive " ^ (if alive t then "t" else "nil") ]
let handle t req =
match Wire.string_field req "op" with
| Some "eval" ->
(match Wire.string_field req "code" with
| Some code ->
let origin =
match Wire.string_field req "file" with Some f -> f | None -> "<editor>"
in
eval t ~code ~origin
| None -> error "eval needs :code")
| Some "describe" -> describe t
| Some "close" -> ok []
| Some op -> error ("unknown op: " ^ op)
| None -> error "no :op"
(* ── The loop ──────────────────────────────────────────────────────── *)
(* One connection at a time. An editor is one client, evaluations are
sequential by nature — each one is checked against the program the last one
left behind — and a second concurrent evaluation would be racing for the
same session anyway. *)
(* Returns whether the client asked to end the session. One editor per daemon,
so [close] shuts the whole thing down rather than waiting for another
connection nobody is going to make. *)
let serve t fd =
let rec go () =
match Wire.recv fd with
| src ->
let op, reply =
match Wire.parse src with
| req -> (Wire.string_field req "op", handle t req)
| exception Loc.Error (_, m) -> (None, error ("bad request: " ^ m))
in
Wire.send fd reply;
if op = Some "close" then true else go ()
| exception Wire.Closed -> false
| exception Unix.Unix_error _ -> false
in
go ()
let start ~file ~sock =
let t0 = Unix.gettimeofday () in
let session, l = Session.create ~file in
let dir =
Filename.concat (Filename.get_temp_dir_name ())
(Printf.sprintf "flan-dev-%d" (Unix.getpid ()))
in
(try Unix.mkdir dir 0o700 with Unix.Unix_error (Unix.EEXIST, _, _) -> ());
let exe = Filename.concat dir "program" in
ignore
(Build.executable ~opts:{ Build.default with Build.dev = true }
~csrcs:l.Load.csrcs ~lflags:l.Load.lflags session.Session.host ~out:exe);
let agent = Filename.concat dir "agent.sock" in
(* The program's source names some socket path; the daemon is the one that
knows where it wants to talk to it, so it overrides through the
environment. Guessing instead would fail silently — everything compiles,
the module is built, and nothing ever receives it. *)
Unix.putenv "FLAN_AGENT_SOCKET" agent;
let child = Unix.create_process exe [| exe |] Unix.stdin Unix.stdout Unix.stderr in
(* Wait for it to bind before accepting an evaluation. One that arrives first
would fail for a reason that reads like a compiler bug. *)
if not (await (fun () -> Sys.file_exists agent)) then begin
(try Unix.kill child Sys.sigterm with Unix.Unix_error _ -> ());
failwith
("the program never listened on " ^ agent
^ " — does it call (agent/start ...)?")
end;
let t = { session; child; agent; dir; n = 0 } in
(try Unix.unlink sock with Unix.Unix_error _ -> ());
let ls = Unix.socket Unix.PF_UNIX Unix.SOCK_STREAM 0 in
Unix.bind ls (Unix.ADDR_UNIX sock);
Unix.listen ls 4;
Printf.eprintf "flan dev: %s ready on %s (%.0fms)\n%!" file sock
((Unix.gettimeofday () -. t0) *. 1000.);
(* [accept] would block past the program's own exit, so it is waited on with
a timeout and the child checked each time round: a daemon whose program
has finished has nothing left to do, and an editor waiting on it would
wait forever. *)
let rec accept_loop () =
if alive t then
match Unix.select [ ls ] [] [] 0.2 with
| [], _, _ -> accept_loop ()
| _ ->
(match Unix.accept ls with
| fd, _ ->
let closed = serve t fd in
(try Unix.close fd with Unix.Unix_error _ -> ());
if not closed then accept_loop ()
| exception Unix.Unix_error (Unix.EINTR, _, _) -> accept_loop ())
| exception Unix.Unix_error (Unix.EINTR, _, _) -> accept_loop ()
in
Fun.protect
~finally:(fun () ->
(try Unix.kill child Sys.sigterm with Unix.Unix_error _ -> ());
(try Unix.close ls with Unix.Unix_error _ -> ());
(try Unix.unlink sock with Unix.Unix_error _ -> ()))
accept_loop