Dep optimization #1

nojb · 2024-08-05T15:09:30Z

No description provided.

nojb · 2024-08-05T15:11:18Z

@MA0010 this PR is for comments on your branch; you should receive an email notification each time there is a new comment/review.

nojb · 2024-08-05T15:16:13Z

doc/concepts/dependency-spec.rst

+- ``(order_only <dep_spec>)`` read the s-expression in ``<dep_spec>`` and interpret it as 
+a dependency specification (i.e. any type mentioned in this list), and declares this dependency as order_only,
+meaning that if a change occurs to it, the rule will not be run again.


This is not quite right; (order_only ...) does not read any s-expression. Also the description is not as accurate as it could be. I suggest:

``(order_only <deps>)`` declares ``<deps>`` as "order only" dependencies. This means that these dependencies will be guaranteed to be up-to-date before building the target, but changes in them will not cause the target to be rebuilt.

nojb · 2024-08-05T15:21:04Z

otherlibs/stdune/src/stdune.ml

-[@@@alert "-unstable"]
+


This diff is not needed. You can test this change with 5.1 or even 4.14 to avoid the bug in OCaml.

nojb · 2024-08-05T15:23:45Z

src/dune_engine/action_exec.ml

+  let maybe_async f =
+
      (match Config.(get background_actions) with
-       | `Enabled -> Scheduler.async_exn
-       | `Disabled -> fun f -> Fiber.return (f ()))
+       | `Enabled -> Scheduler.async_exn f
+       | `Disabled -> Fiber.return (f ()))
  in
-  fun f -> (Lazy.force maybe_async) f
+  fun f ->  maybe_async f


Is this diff needed?

nojb · 2024-08-05T15:24:37Z

src/dune_engine/action_exec.ml

+  { done_or_more_deps : Action_plugin.done_or_more_deps;
+
+    needed_deps : Dep.Set.t
+  }


Don't forget to ocamlformat!

nojb · 2024-08-05T15:25:29Z

src/dune_engine/action_exec.ml

+    let complete_path = function
+    | Dep.File x -> Dep.file (Path.relative eenv.working_dir (Path.to_string x))
+    | Dep.File_selector x -> 
+                let path = Path.to_string (File_selector.dir x) in
+                let dir = Path.relative eenv.working_dir path in
+                Dep.file_selector (File_selector.edit_dir ~dir x)
+    (*| Dep.Alias x ->
+                let name = Alias.name x in
+                let dir = Path.to_string (Alias.dir x) in
+                let dir = Path.relative eenv.working_dir dir in
+                let alias = Alias.make name dir in
+                Dep.alias alias *)
+    | x -> x


As discussed, this logic should be part of the decoder.

nojb · 2024-08-05T15:26:10Z

src/dune_engine/action_exec.ml

+type action_res = 
+  { done_or_more_deps : Action_plugin.done_or_more_deps;
+
+    needed_deps : Dep.Set.t
+  }


This type should be defined in its own submodule with a union function instead of action_res_union below.

nojb · 2024-08-05T15:26:47Z

src/dune_engine/action_exec.ml

@@ -264,14 +278,15 @@ let exec_run_dynamic_client ~display ~ectx ~eenv prog args =
           that is incompatible with this version of dune."
          prog_name
      ]
-  | Ok Done -> Done
+  | Ok Done -> {done_or_more_deps = Done; needed_deps = Dep.Set.empty}


Suggest define return : ?needed_deps:Dep.Set.t -> done_or_more_deps -> action_res and use that instead of these record literals.

nojb · 2024-08-05T15:27:34Z

src/dune_engine/rule.mli

@@ -59,6 +59,7 @@ type t = private
  ; mode : Mode.t
  ; info : Info.t
  ; loc : Loc.t
+  ; requires_a_deps : bool


I thought this wasn't necessary anymore.

nojb · 2024-08-05T15:28:12Z

src/dune_lang/action.ml

+    (if List.length xs > 1
+     then Syntax.since ~what:"Passing several arguments to 'needed_deps'" Stanza.syntax (3, 4)
+     else return ())


What is this for?

It should not be here. This was used in the cat parser to alert passing several arguments, since then the execution of cat depends on the version of dune.

None of this code needs to exist in the engine and lives just as happily as a separate library. Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

In 35b3e0d we started passing the configured parallelism to package builds by setting their `jobs` variable to the degree of parallelism configured for dune, but it seems this change was only applied for builds started with `--watch`. This commit applies the configured parallelism to package builds when dune is run without `--watch`. Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

MA0010 · 2024-08-06T10:40:20Z

@nojb I have added a new commit containing the suggested tweaks + some other changes. I still need to figure out the issue with working_dir type; it seems for now that it is only pointing to the build directory _build/default, but it is used in a lot of functions that require a Path.t type instead of the more specific Path.Build.t.

nojb · 2024-08-06T10:39:56Z

src/dune_engine/action_exec.ml

+  let return ?needed_deps done_or_more_deps =
+    match needed_deps with
+    | None -> { done_or_more_deps; needed_deps = Dep.Set.empty }
+    | Some needed_deps -> { done_or_more_deps; needed_deps }
+  ;;


Suggested change

let return ?needed_deps done_or_more_deps =

match needed_deps with

| None -> { done_or_more_deps; needed_deps = Dep.Set.empty }

| Some needed_deps -> { done_or_more_deps; needed_deps }

;;

let return ?needed_deps done_or_more_deps =

let needed_deps = Option.value ~default:Dep.Set.empty needed_deps in

{ done_or_more_deps; needed_deps }

;;

nojb · 2024-08-06T10:40:26Z

src/dune_engine/action_exec.ml

+    Action_res.return Done
+  | Needed_deps needed ->
+    let ds = Dep.decode eenv.working_dir in
+    (*the files should first be parsed*)


Suggested change

(*the files should first be parsed*)

nojb · 2024-08-06T10:42:24Z

src/dune_engine/file_selector.ml

+let edit_dir ~dir t =
+  let predicate = t.predicate in
+  let only_generated_files = t.only_generated_files in
+  of_predicate_lang ~dir ~only_generated_files predicate
+;;


This is not necessary, right?

nojb · 2024-08-06T10:43:40Z

src/dune_lang/action.ml

@@ -350,6 +351,10 @@ let cstrs_dune_file t =
    , Syntax.since Stanza.syntax (2, 7)
      >>> let+ script = sw in
          Cram script )
+  ; ( "needed_deps"
+    , Syntax.since Stanza.syntax (2, 7)


The version seems wrong.

nojb · 2024-08-06T10:44:17Z

src/dune_rules/dep_conf_eval.ml

+    (*let l = List.fold_left ~init:[] ~f:(fun acc x -> match (dep expander x) with
+      | Simple _ as a -> (to_action_builder a) :: acc
+      | Other x -> x :: acc) s*)


Suggested change

(*let l = List.fold_left ~init:[] ~f:(fun acc x -> match (dep expander x) with

| Simple _ as a -> (to_action_builder a) :: acc

| Other x -> x :: acc) s*)

nojb · 2024-08-06T10:45:35Z

src/dune_engine/build_system.ml

+                match exec_result.action_exec_result.needed_deps with
+                | deps, facts -> Some (deps, Dep.Facts.digest facts ~env:action.env)


Use let deps, fact = ....

nojb · 2024-08-06T10:45:51Z

src/dune_engine/rule_cache.ml

@@ -9,14 +9,19 @@ module Workspace_local = struct
        { rule_digest : Digest.t
        ; dynamic_deps_stages : (Dep.Set.t * Digest.t) list
        ; targets_digest : Digest.t
+        ; needed_deps : (Dep.Set.t * Digest.t) option


I think the option is not needed, right?

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

This solves a lot of the waiting leftover HTTP processes etc. The new fake curl will read the list of files to serve from `fake-curls` and map each line to a port. To match the behavior of the oneshot-webserver it will only serve each port once, subsequent requests for the same port will be answered by 404 (so not exactly the same but similar enough), the ports that were served are recorded into an `already-served` file. The same file can be served multiple times, as long as it is present multiple times in `fake-curls`. Signed-off-by: Marek Kubica <marek@tarides.com>

Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

This changes the variable that enables toolchains from DUNE_CONFIG__TOOLCHAINS_ENABLED=[true|false] to DUNE_CONFIG__TOOLCHAINS=[enabled|disabled], following the established convention for such variables in dune, and for compatibility with upcoming support for compile-time configuration. Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

do not pass entire package when we only need the depopts Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

And separate some Memo code into separate functions Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

First pass at progress indication for lockdir generation, where the status line displays a message while the package repos are being updated, and a second message during solving. This might be updated in the future to display finer-grained status updates. Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

MA0010 · 2024-08-12T12:03:20Z

Hello @nojb, I have added 2 new commits. The first contains some code tweaks and corrections, and the second contains tests written for the new features.

There are two tests for now, but I think that the needed-deps.t is enough (there is no need for the separate order-only.t, because the feature is tested inside needed-deps.t).

Please when you are free, try to look at the test files and give your opinion on the structure and any suggestions that you come up with. I will now gaurd the new features under the new dune version and do a primary commit.

Thanks!
Have a nice holiday!

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

unreachable packages are those that aren't reachable via the dependencies of local packages typically, such packages are included when solving via post deps Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Signed-off-by: Marek Kubica <marek@tarides.com> Co-authored-by: Etienne Millon <me@emillon.org>

This reverts commit 7ef8c00.

nojb · 2024-08-13T15:03:56Z

doc/concepts/dependency-spec.rst

@@ -46,6 +46,9 @@ Dependencies in ``dune`` files can be specified using one of the following:
 - ``(include <file>)`` read the s-expression in ``<file>`` and interpret it as
  additional dependencies. The s-expression is expected to be a list of the
  same constructs enumerated here.
+- ``(order_only <deps>)`` declares ``<deps>`` as "order only" dependencies. This means that these
+dependencies will be guaranteed to be up-to-date before building the target, but changes in them will
+not cause the target to be rebuilt.


Please resepect existing style (indentation, line width, etc).

nojb · 2024-08-13T15:04:55Z

doc/reference/actions/needed_deps.rst

+
+   Example::
+
+   (needed_deps deps)


This is not a full example; a full example and an explanation of the expected behaviour should be added.

nojb · 2024-08-13T15:05:23Z

doc/reference/actions/needed_deps.rst

+   ``file <filename>
+    alias <alias_name>
+    file_selector <glob>
+    universe``


Here we need a fuller description of each constructor. See for example the section on "Dependency Specification" in the Dune manual. https://dune.readthedocs.io/en/stable/concepts/dependency-spec.html

nojb · 2024-08-13T15:06:32Z

src/dune_engine/action.ml

@@ -300,6 +303,7 @@ let is_useful_to memoize =
    | System _ -> true
    | Bash _ -> true
    | Extension (module A) -> A.Spec.is_useful_to ~memoize
+    | Needed_deps _ -> memoize


What does memoize mean here?

It means that it makes sense to store the rule in the cache, but it does not make sense to look up the target of the action in the distributed (or shared) cache. However, giving a second look at our final definition of needed_deps, I think that both of these operations are not necessary, so this should be set to false.

nojb · 2024-08-13T15:07:20Z

src/dune_engine/dep.ml

+          (*          let dir = Path.Build.of_string (Path.to_string w_dir) in*)
+          (*let dir = Path.Build.of_string "_build/.actions/default" still under test*)


Please remove stale comments.

nojb · 2024-08-13T15:07:43Z

src/dune_engine/alias.mli

@@ -3,6 +3,7 @@ open Import
 module Name : sig
  include module type of Dune_util.Alias_name with type t = Dune_util.Alias_name.t

+  val of_string : string -> t


Is this change needed?

nojb · 2024-08-13T15:08:00Z

src/dune_engine/dep.ml

@@ -28,6 +29,30 @@ module T = struct
  let universe = Universe
  let file_selector g = File_selector g

+  let decode w_dir =


Suggested change

let decode w_dir =

let decode dir =

nojb · 2024-08-13T15:09:56Z

src/dune_lang/dep_conf.ml

@@ -103,8 +105,13 @@ let decode files =
        , let+ () = Syntax.since Stanza.syntax (3, 1)
          and+ filename = filename in
          Include filename )
+      ; ( "order_only"
+        , let+ () = Syntax.since Stanza.syntax (3, 18)


The last released version of Dune is 3.16; isn't 3.17 enough?

Suggested change

, let+ () = Syntax.since Stanza.syntax (3, 18)

, let+ () = Syntax.since Stanza.syntax (3, 17)

If that is the case, then yes. I thought that 3.17 was released because the version was bumped in the dune codebase.

MA0010 · 2024-08-14T09:15:21Z

Hello @nojb, just to explain my last commits. I have removed the version bumping of dune. Also, I applied the suggested tweaks to the code, and finally I expanded the docs.
Please when you have time, check the new written docs and give me any suggestions you might have.

Thanks!

Dune_sexp belongs to dune's frontend and the engine has no business relying on it. This is a change we've done internally, so this is just syncing. Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Signed-off-by: Etienne Millon <me@emillon.org>

* Better patch format parsing Signed-off-by: Ambre Austen Suhamy <ambre@tarides.com> * Clarify that we don't care about the timestamp itself Signed-off-by: Ambre Austen Suhamy <ambre@tarides.com> * Undo style changes Signed-off-by: Ambre Austen Suhamy <ambre@tarides.com> --------- Signed-off-by: Ambre Austen Suhamy <ambre@tarides.com>

Signed-off-by: Marek Kubica <marek@tarides.com>

* Use backslashes when passing staged pps driver on Windows Signed-off-by: Nicolás Ojeda Bär <n.oje.bar@gmail.com>

* perf: better names for trace events Signed-off-by: Javier Chávarri <javier.chavarri@gmail.com>

* feat: create an alias from pkg to package Signed-off-by: Etienne Marais <dev@maiste.fr>

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

support for ocamlformat is hardcoded anyway, so just setup the rules with the extra expansion step for actions defined using the dune lang. Co-authored-by: Stephen Sherratt <stephen@sherra.tt> Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Make it possible to attempt to read a lockdir from a directory without raising a [User_error] if the lockdir is not present. Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

…void useless recompilations when Dune's stderr is redirected (ocaml#10883) Signed-off-by: Nicolás Ojeda Bär <n.oje.bar@gmail.com>

Signed-off-by: Javier Chávarri <javier.chavarri@gmail.com>

…ocaml#10885) Signed-off-by: Marek Kubica <marek@tarides.com>

) Signed-off-by: Marek Kubica <marek@tarides.com>

* fix: enabled_if with `dune describe` Fixes ocaml#10779. Signed-off-by: Alpha DIALLO <moyodiallo@gmail.com>

…aml#10899) Signed-off-by: Marek Kubica <marek@tarides.com>

This makes it possible to check whether the user is currently inside a dune project. This is needed for dev-tools to have fallback behaviour when run outside of a dune project which is necessary to allow users to configure their editors to invoke dev-tools via dune. For example, if a user configures their editor to start ocamllsp with the command `dune ocamllsp`, dune needs to be able to handle the case where the user edits an ocaml file outside of a dune project, for example by attempting to invoke ocamllsp from PATH. Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

MA0010 · 2024-09-10T14:42:19Z

Hello @nojb, I have added 3 new commits that (hopefully) contain a functional code for the shared cache and build_deps edit, ready to be tested.

However, for now, I removed passing the Loc.t to the function that parses the csexp of the file stored in the shared cache. Also, it remains to pass the correct location of the rule when raising the error while building needed_deps. I will work on that, but I wanted to commit the code so that it can be tested, since there is some behaviour that needs to be verified.

nojb · 2024-09-10T14:47:57Z

Thanks, I'll take a look.

nojb · 2024-09-03T11:07:38Z

src/dune_cache_storage/dune_cache_storage.ml

+      "Cannot parse file digest %s in cache metadata entry"
+      deps_digest)))
+      | _ -> Error (Failure "Cannot parse cache metadata entry")
+      ;;*)


Let's remove debug code.

nojb · 2024-09-11T05:48:30Z

bin/print_rules.ml

@@ -108,6 +108,7 @@ let rec encode : Action.For_shell.t -> Dune_lang.t =
  | Pipe (outputs, l) ->
    List (atom (sprintf "pipe-%s" (Outputs.to_string outputs)) :: List.map l ~f:encode)
  | Extension ext -> List [ atom "ext"; ext ]
+  | Needed_deps xs -> List (atom "needed_deps" :: List.map xs ~f:path)


"needed deps" does not read very well; I suggest to change it to "actual_deps" everywhere.

nojb · 2024-09-11T06:39:34Z

doc/concepts/dependency-spec.rst

@@ -46,6 +46,9 @@ Dependencies in ``dune`` files can be specified using one of the following:
 - ``(include <file>)`` read the s-expression in ``<file>`` and interpret it as
  additional dependencies. The s-expression is expected to be a list of the
  same constructs enumerated here.
+- ``(order_only <deps>)`` declares ``<deps>`` as "order only" dependencies.


similarly, maybe we could replace "order_only" by "maybe_deps" ?

nojb · 2024-09-11T06:41:05Z

src/dune_engine/action_exec.ml

@@ -138,6 +139,27 @@ module Exec_result = struct
  ;;
 end

+module Action_res = struct


What does Action_res stand for? Maybe Action_deps would be better?

nojb · 2024-09-11T06:43:08Z

src/dune_engine/action_exec.ml

+      Produce.return
+        (Action_res.return need ~needed_deps)


This is Produce.return action_res, right?

nojb · 2024-09-11T06:45:13Z

src/dune_engine/action_exec.ml

+     let* action_res = exec t ~display ~ectx ~eenv in
+     (match action_res with
+      | { done_or_more_deps = d; needed_deps } ->
+        Produce.return (Action_res.return d ~needed_deps))


This is Produce.return action_res?

nojb · 2024-09-11T06:47:08Z

src/dune_engine/action_exec.ml

+      }
+  ;;
+
+  let return ?needed_deps done_or_more_deps =


Since you never pass needed_deps or Need_more_deps _ at the same time, I suggest to have instead:

val no_deps: t val need_more_deps: ... -> t val actual_deps: Dep.Set.t -> t

nojb · 2024-09-11T06:49:44Z

src/dune_engine/rule_cache.ml

           | false -> Fiber.return (Miss Miss_reason.Dynamic_deps_changed))
      in
-      loop prev_trace.dynamic_deps_stages
+      loop prev_trace.dynamic_deps_stages prev_trace.needed_deps


This is overly complicated; the check should have the form check1 prev_trace.dynamic_deps_stages || check2 prev_trace.neede_deps, there is no need to mix the check for both types of dependencies.

nojb · 2024-09-11T06:57:41Z

Let's hold off on renaming "order_only" to "maybe_deps" and "needed_deps" to "actual_deps", we may have other ideas later.

Signed-off-by: HasanA <mhmd_alameen1023@outlook.com>

MA0010 · 2024-09-17T15:43:09Z

Hello @nojb, the code should be ready for review.

nojb commented Aug 5, 2024

View reviewed changes

rgrinberg and others added 3 commits August 6, 2024 10:25

refactor: move diff action + promotion to library (ocaml#10796)

dbce52e

None of this code needs to exist in the engine and lives just as happily as a separate library. Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

test: fix formatting in promotions test (ocaml#10800)

beb8724

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

nojb commented Aug 6, 2024

View reviewed changes

rgrinberg and others added 13 commits August 6, 2024 14:44

refactor: switch to regular sexps for custom actions (ocaml#10807)

4b25456

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

test: formatting cleanup (ocaml#10809)

1dd8015

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Add test of the toolchains mechanism (ocaml#10798)

546661c

Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

refactor(pkg): trivial binding simplifications (ocaml#10812)

9af2225

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

refactor(pkg): simplify depopts passing (ocaml#10813)

857ee98

do not pass entire package when we only need the depopts Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

refactor(pkg): move [variables] to [Dune_pkg] (ocaml#10816)

ebd58f2

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

refactor: move [Dune_project_name] to [dune_lang] (ocaml#10817)

0862778

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

refactor: remove [Dune_rules.Package] (ocaml#10818)

1f9dd2e

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

refactor: move [Dialect] to [Dune_lang] (ocaml#10819)

8735d9d

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

refactor(pkg): move all IO in [Dune_project] (ocaml#10820)

7788f53

And separate some Memo code into separate functions Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

rgrinberg added 2 commits August 13, 2024 00:49

refactor(pkg): remove some dead code (ocaml#10815)

230e480

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

fix(pkg): do not lock unreachable packages (ocaml#10814)

3e56902

unreachable packages are those that aren't reachable via the dependencies of local packages typically, such packages are included when solving via post deps Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

MA0010 force-pushed the dep_optimization branch from d6157a2 to 1f83184 Compare August 13, 2024 09:25

Leonidas-from-XIV and others added 2 commits August 13, 2024 11:57

Add configure-time toggles (ocaml#10811)

7ef8c00

Signed-off-by: Marek Kubica <marek@tarides.com> Co-authored-by: Etienne Millon <me@emillon.org>

Revert "Add configure-time toggles (ocaml#10811)" (ocaml#10823)

96ed5fb

This reverts commit 7ef8c00.

nojb commented Aug 13, 2024

View reviewed changes

MA0010 force-pushed the dep_optimization branch from bf03c78 to 2e906ea Compare August 14, 2024 09:12

refactor(engine): remove [Dune_sexp] from the engine (ocaml#10826)

ca8ab93

Dune_sexp belongs to dune's frontend and the engine has no business relying on it. This is a change we've done internally, so this is just syncing. Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

emillon and others added 16 commits September 4, 2024 10:54

chore: update JS deps to 0.17 (ocaml#10878)

7d50c4b

Signed-off-by: Etienne Millon <me@emillon.org>

test: Repro case for spaces in patch file names (ocaml#10879)

d9b0e2e

Signed-off-by: Marek Kubica <marek@tarides.com>

Fix staged pps on Windows (ocaml#10869)

45976a0

* Use backslashes when passing staged pps driver on Windows Signed-off-by: Nicolás Ojeda Bär <n.oje.bar@gmail.com>

perf: add names to source tree events (ocaml#10884)

9c0d65b

* perf: better names for trace events Signed-off-by: Javier Chávarri <javier.chavarri@gmail.com>

feat: create an alias from pkg to package (ocaml#10858)

77592f5

* feat: create an alias from pkg to package Signed-off-by: Etienne Marais <dev@maiste.fr>

test: prefix XDG_STATE_HOME (ocaml#10887)

783a913

Signed-off-by: Rudi Grinberg <me@rgrinberg.com>

Return [result] from Lock_dir.read_disk (ocaml#10889)

d84b9d7

Make it possible to attempt to read a lockdir from a directory without raising a [User_error] if the lockdir is not present. Signed-off-by: Stephen Sherratt <stephen@sherra.tt>

Always pass -fdiagnostics-color=always to cc-style compilers to a…

a7924e3

…void useless recompilations when Dune's stderr is redirected (ocaml#10883) Signed-off-by: Nicolás Ojeda Bär <n.oje.bar@gmail.com>

ci bench: use 5.1 (ocaml#10893)

7ec38ff

Signed-off-by: Javier Chávarri <javier.chavarri@gmail.com>

fix: Support patches without a prefix by determining the prefix depth (…

b3bd89f

…ocaml#10885) Signed-off-by: Marek Kubica <marek@tarides.com>

test: Repro case for when the patch prefix is not a//b/ (ocaml#10895

83f6e1c

) Signed-off-by: Marek Kubica <marek@tarides.com>

fix: enabled_if with dune describe (ocaml#10881)

2498355

* fix: enabled_if with `dune describe` Fixes ocaml#10779. Signed-off-by: Alpha DIALLO <moyodiallo@gmail.com>

fix: Support prefixes other than a/ & b/ when modifying files (oc…

9b805a0

…aml#10899) Signed-off-by: Marek Kubica <marek@tarides.com>

MA0010 force-pushed the dep_optimization branch from 64db594 to f7d4dcc Compare September 10, 2024 15:32

nojb commented Sep 11, 2024

View reviewed changes

Add order_only dependency specification

39a15c6

Signed-off-by: HasanA <mhmd_alameen1023@outlook.com>

MA0010 force-pushed the dep_optimization branch 4 times, most recently from 8f6a480 to eae7315 Compare September 16, 2024 13:59

MA0010 added 2 commits September 17, 2024 16:26

Add needed_deps action

7894668

Signed-off-by: HasanA <mhmd_alameen1023@outlook.com>

Add lazy mode to build_deps

a364644

Signed-off-by: HasanA <mhmd_alameen1023@outlook.com>

MA0010 force-pushed the dep_optimization branch from eae7315 to a364644 Compare September 17, 2024 14:30

	(*let l = List.fold_left ~init:[] ~f:(fun acc x -> match (dep expander x) with
	\| Simple _ as a -> (to_action_builder a) :: acc
	\| Other x -> x :: acc) s*)

		match exec_result.action_exec_result.needed_deps with
		\| deps, facts -> Some (deps, Dep.Facts.digest facts ~env:action.env)

		(* let dir = Path.Build.of_string (Path.to_string w_dir) in*)
		(let dir = Path.Build.of_string "_build/.actions/default" still under test)

	, let+ () = Syntax.since Stanza.syntax (3, 18)
	, let+ () = Syntax.since Stanza.syntax (3, 17)

		[@@@alert "-unstable"]

Dep optimization #1

Are you sure you want to change the base?

Dep optimization #1

Conversation

nojb commented Aug 5, 2024

nojb commented Aug 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MA0010 commented Aug 6, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MA0010 commented Aug 12, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MA0010 commented Aug 14, 2024

MA0010 commented Sep 10, 2024

nojb commented Sep 10, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nojb commented Sep 11, 2024

MA0010 commented Sep 17, 2024