flan/test/test_session.ml
Joseph Ferano 3ace7c262f A hang is a failure the suite never reported
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.
2026-09-12 10:49:07 +07:00

266 lines
14 KiB
OCaml

(* The session: the declarations a running process was built from, plus every
change accepted since (NEXT.md, the dev loop).
Two halves. First, what a session refuses — every case here is a change that
would compile, load, install, and then be wrong, because a cell is a bare
pointer and storage that already exists already has a shape. Second, that
refusing leaves the session usable, which is the failure people actually hit:
one typo must not poison every later evaluation. *)
open Flan
(* The watchdog first: a hang is the one failure mode that reports
nothing at all. See watchdog.ml. *)
let () = Watchdog.arm ~seconds:600 "test_session"
let failures = ref 0
let fail fmt = Printf.ksprintf (fun s -> incr failures; print_endline ("FAIL " ^ s)) fmt
let has hay needle =
let n = String.length needle and h = String.length hay in
let rec go i = i + n <= h && (String.sub hay i n = needle || go (i + 1)) in
go 0
(* Every rejection is asserted on its reason, not just on the failure: the
reason is the part that has to survive a refactor. *)
let checked_program file =
Check.program
(Load.program ~file (Parse.program (Reader.read_file file))).Load.decls
let refuses ?(file = "programs/reload.flan") name src reason =
let t, _ = Session.create ~file () in
match Session.eval t src with
| _ -> fail "%s was accepted" name
| exception Loc.Error (_, msg) ->
if not (has msg reason) then
fail "%s\n said: %S\n wanted it to mention: %S" name msg reason
let () =
(* A cell carries no signature, so every call site compiled before the change
still passes the old arguments through it. *)
(* [outer] is called only from C, and [spare] is read by nothing, so the
checker has no complaint about either change and the session is the only
thing that can refuse them. A change something else in the program uses is
an ordinary type error first, which is a different and louder failure. *)
refuses "a changed parameter type"
"(defn outer [x i64] i64 (bump))"
"changes signature";
refuses "a changed return type"
"(defn outer [] i32 (i32 (bump)))"
"changes signature";
refuses "a changed arity"
"(defn outer [a i64 b i64] i64 (bump))"
"changes signature";
(* The storage exists and has a shape: reusing it reads at the wrong offsets,
and replacing it discards the state the reload exists to preserve. *)
refuses "a retyped global"
"(defvar spare i32)"
"changes type";
(* Values of the type are already in the running program's memory. *)
(* A defconst the *checker* consumed is in the shape of the program — an
array length is decided before any type resolves — so no store can reach
it. One that is only read at run time is a different matter; see below. *)
refuses "a changed defconst used at compile time"
"(defconst folded i64 8)"
"used at compile time";
(* An enum member is erased to an i32 literal in the caller, so the same
applies. It is compared over declarations because Tast.program carries no
enums at all, for exactly that reason. *)
refuses "a changed enum member"
"(defenum Colour [red 0 green 2])"
"changes its members";
refuses ~file:"programs/values.flan" "a restructured struct"
"(defstruct P [x i32 y i32])"
"changes layout";
(* An ordinary redefinition, and what the session works out about it. *)
let t, _ = Session.create ~file:"programs/reload.flan" () in
let c = Session.eval t "(defn bump [] i64 (set counter (+ counter 5)) counter)" in
if not c.Session.installs then fail "a redefined function had nothing to install";
if c.Session.fns <> [ "bump" ] then
fail "redefining bump reported %s" (String.concat " " c.Session.fns);
(* The prelude is in the checked program and in no accumulated AST, so a
session that derived [known] from declarations would call rand-seed
through a registry cell nobody ever publishes. *)
if not (has c.Session.ir "@\"flan.cell.rand-seed\" = external global ptr") then
fail "the prelude was treated as new";
if has c.Session.ir "flan_dev_cell" then
fail "a name the host has went through the registry";
(* DWARF in a redefinition module, which is a property of the session and
not of the call. [Emit.redefinition] has taken a ~debug argument all
along and was tested with it; what was missing was anyone passing it, so
every body installed by C-c C-c lost its debug info in a running process.
The defect was one unpassed argument, so the test is that the argument
arrives — asserted on the emitted text, which is the only place it shows.
Both directions matter. A session that always emitted debug info would
force -O0 on every reloaded body ([Build.shared] does that, and must),
which would change the frame time of the one function being iterated on.
Off unless asked for is the behaviour, so off is asserted too. *)
let dt, _ = Session.create ~debug:true ~file:"programs/reload.flan" () in
let dc =
Session.eval dt
"(defn bump [] i64 (let [step (i64 5)] (set counter (+ counter step)) counter))"
in
if not (has dc.Session.ir "!DILocalVariable(name: \"step\"") then
fail "a debug session's redefinition carries no name for its local";
if not (has dc.Session.ir "!DISubprogram(name: \"bump\"") then
fail "a debug session's redefinition carries no subprogram";
let pt, _ = Session.create ~file:"programs/reload.flan" () in
let pc =
Session.eval pt
"(defn bump [] i64 (let [step (i64 5)] (set counter (+ counter step)) counter))"
in
if has pc.Session.ir "!DILocalVariable" then
fail "a plain session's redefinition carries debug info it was not asked for";
(* The same for an expression evaluation, which takes the other path out of
the session and so can lose the flag on its own. *)
let ec = Session.eval_expr dt "(+ counter 1)" in
if not (has ec.Session.ir "!DISubprogram") then
fail "a debug session's eval thunk carries no debug info";
(* A form that does not check must leave the session exactly as it was. This
is the one that decides whether a REPL survives a typo. *)
(match Session.eval t "(defn bump [] i64 nonsense)" with
| _ -> fail "an unresolvable name was accepted"
| exception Loc.Error _ -> ());
(match Session.eval t "(defn bump [] i64 (set counter (+ counter 6)) counter)" with
| c -> if c.Session.fns <> [ "bump" ] then fail "the session did not recover"
| exception Loc.Error (_, m) -> fail "the session was poisoned by a typo: %s" m);
(* A declaration the program already has, with no body and no new storage,
is accepted and has nothing to send. Building a module for it would report
success for a change that cannot have taken effect, and would cost the
program a reload it did not need. *)
(match Session.eval t "(defvar counter i64)" with
| c -> if c.Session.installs then fail "an empty change claimed to install"
| exception Loc.Error (_, m) -> fail "redeclaring a var unchanged: %s" m);
(* A constant that is only ever read at run time is just bytes in the
program's memory. A dev build emits it as a mutable global and the module
stores the new value at the frame boundary, which is how a colour table
gets tuned live. *)
(match Session.eval t "(defconst palette [2 u32] [9 9])" with
| c ->
if not c.Session.installs then
fail "a changed run-time constant had nothing to install";
if not (has c.Session.ir "store [2 x i32]") then
fail "a changed run-time constant published no new value"
| exception Loc.Error (_, m) -> fail "changing a run-time constant: %s" m);
(* And in a dev build its storage is writable, where a release build keeps
it immutable and gets all the folding back. *)
let host = checked_program "programs/reload.flan" in
if not (has (Emit.program ~dev:true host) "@\"flan.palette\" = global") then
fail "a dev build left a constant immutable";
if not (has (Emit.program host) "@\"flan.palette\" = constant") then
fail "a release build made a constant mutable";
(* A new global carries its declared initial value, copied once when the
storage is allocated and never again — calloc alone would make it zero. *)
let c = Session.eval t "(defvar started i64 42) (defn read-started [] i64 started)" in
if not (has c.Session.ir "@\".init.") then
fail "a new global's initialiser was dropped";
if not c.Session.installs then fail "adding a global had nothing to install";
(* Names the process was never built with go through the registry instead of
binding to a symbol, and adding one is allowed where retyping one is not. *)
let c = Session.eval t "(defvar fresh i64) (defn use-fresh [] i64 (set fresh 3) fresh)" in
if not (List.mem "fresh" c.Session.names && List.mem "use-fresh" c.Session.fns) then
fail "adding a var and a function reported %s" (String.concat " " c.Session.names);
if not (has c.Session.ir "call ptr @flan_dev_global") then
fail "a new global did not go through the registry";
(* And once added, it is part of the session: a later form can use it. *)
(match Session.eval t "(defn use-fresh [] i64 (set fresh 4) fresh)" with
| _ -> ()
| exception Loc.Error (_, m) -> fail "a name added earlier was forgotten: %s" m);
(* A file with imports, re-evaluated whole — the C-c C-k case. The session
keeps the *expanded* declarations, so the package's names are replaced in
place rather than appended a second time and rejected as duplicates. *)
let t, _ = Session.create ~file:"../sand.flan" () in
let src = In_channel.with_open_bin "../sand.flan" In_channel.input_all in
(match Session.eval t src with
| c ->
if not (List.mem "game-draw" c.Session.fns) then
fail "reloading sand.flan did not include its own functions"
| exception Loc.Error (_, m) ->
fail "reloading a file with imports failed: %s" m);
(* A form typed into a file that is *imported as a package* has to be
qualified the way the import qualified it, or it splices as a brand-new
unrelated name: the evaluation reports success and the running program
goes on calling the one it already had. The alias is chosen by the
importer and written nowhere in the file, so the path is the only thing
that can decide it — which is why it is derived here and not sent by the
editor. *)
let t, _ = Session.create ~file:"../sand.flan" () in
(match
Session.eval ~origin:"../vendor/agent/agent.flan" t
"(defn poll [] i32 (poll-raw))"
with
| c ->
if c.Session.fns <> [ "agent/poll" ] then
fail "a form from a package file reported %s, wanted agent/poll"
(String.concat " " c.Session.fns)
| exception Loc.Error (_, m) -> fail "redefining agent/poll: %s" m);
(* A package that is a single file, which is what sand.flan is to the
headless driver. The file being edited *is* the package rather than a
member of a directory, so matching on the directory alone would answer
"not a package" — and the failure is the silent one above: the form
splices as a bare [step] and the running program keeps the one it had. *)
let t2, _ = Session.create ~file:"programs/sand-headless.flan" () in
(match Session.eval ~origin:"../sand.flan" t2 "(defn step [] Unit (do))" with
| c ->
if c.Session.fns <> [ "sand/step" ] then
fail "a form from a single-file package reported %s, wanted sand/step"
(String.concat " " c.Session.fns)
| exception Loc.Error (_, m) -> fail "redefining sand/step: %s" m);
(* And a file that is not a package keeps its names as written. *)
(match Session.eval ~origin:"../sand.flan" t "(defn game-draw [] Unit (do))" with
| c ->
if c.Session.fns <> [ "game-draw" ] then
fail "a form from the program's own file reported %s"
(String.concat " " c.Session.fns)
| exception Loc.Error (_, m) -> fail "redefining game-draw: %s" m);
(* An expression's thunk leaves nothing behind, and the module says so, which
is what lets the agent unload it: nothing may point into its text
afterwards. So it is called directly rather than through a cell, and it
must not take a registry slot either — there are 4096 of those and an
expression evaluated in a loop would exhaust them. A module that publishes
a body can never say this; its whole purpose is to leave a pointer. *)
let t, _ = Session.create ~file:"programs/reload.flan" () in
let e = Session.eval_expr t "(+ 1 2)" in
if not (has e.Session.ir "@flan_reload_transient") then
fail "an expression's module did not declare itself unloadable";
if not (has e.Session.ir "define void @flan_reload_call") then
fail "an expression's module carried no thunk to run";
if has e.Session.ir "flan.cellp.eval" then
fail "an expression's thunk took a registry slot";
if has e.Session.ir "call ptr @flan_dev_cell" then
fail "an expression's thunk was looked up by name";
let c = Session.eval t "(defn bump [] i64 (set counter (+ counter 1)) counter)" in
if has c.Session.ir "@flan_reload_transient" then
fail "a module that publishes a body claimed to be unloadable";
(* And a third condition, about data rather than text. A string literal lives
in the evaluating module's own image, and an expression may store one
anywhere: [(set msg "x")] on a string global would leave that global
pointing into a mapping the agent then drops — and since the next thunk can
be mapped at the same address, the result is silent garbage rather than a
fault. A module carrying any string constant keeps its mapping. *)
let str = Session.eval_expr t "(println \"tuned\")" in
if not (has str.Session.ir ".str.0") then
fail "the fixture stopped carrying a string constant, so it proves nothing";
if has str.Session.ir "@flan_reload_transient" then
fail "an expression holding a string claimed to be unloadable";
if !failures = 0 then print_endline "session: all tests passed"
else begin
Printf.printf "\n%d failure(s)\n" !failures;
exit 1
end