Add documentation and better naming around checker callbacks

Summary:
1. Some invariants are tricky enough to be documented. This is especially
important for cases related with error reporting. Lets document it.
2. Cluster callback -> File callback rename.

Reviewed By: ngorogiannis

Differential Revision: D20093932

fbshipit-source-id: e716f1f5b

infer/src/IR/IssueLog.mli

@@ -7,7 +7,9 @@
 
 open! IStd
 
-(** Module for maps from procnames to error logs. *)
+(** Module for storing issues detected outside of per-procedure analysis (and hence not serialized
+    as a part of procedure summary). Normally those are issues detected at the file-level analysis
+    step. Individual checkers are responsible for storing their own list of such issues. *)
 type t
 
 val empty : t

infer/src/backend/InferAnalyze.ml

@@ -60,8 +60,8 @@ let analyze_target : (SchedulerTypes.target, Procname.t) Tasks.doer =
         analyze_source_file exe_env source_file
 
 
-let output_json_makefile_stats clusters =
+let output_json_makefile_stats files =
-  let num_files = List.length clusters in
+  let num_files = List.length files in
   let num_procs = 0 in
   (* can't compute it at this stage *)
   let num_lines = 0 in
@@ -144,7 +144,7 @@ let analyze source_files_to_analyze =
   else (
     L.environment_info "Parallel jobs: %d@." Config.jobs ;
     let build_tasks_generator () = tasks_generator_builder_for source_files_to_analyze in
-    (* Prepare tasks one cluster at a time while executing in parallel *)
+    (* Prepare tasks one file at a time while executing in parallel *)
     RestartScheduler.setup () ;
     let runner =
       Tasks.Runner.create ~jobs:Config.jobs ~f:analyze_target ~child_epilogue:BackendStats.get

infer/src/backend/Summary.mli

@@ -47,6 +47,10 @@ type t =
   ; status: Status.t
   ; proc_desc: Procdesc.t
   ; err_log: Errlog.t
+        (** Those are issues that are detected for this procedure after per-procedure analysis. In
+            addition to that there can be errors detected after file-level analysis (next stage
+            after per-procedure analysis). This latter category of errors should NOT be written
+            here, use [IssueLog] and its serialization capabilities instead. *)
   ; mutable callee_pnames: Procname.Set.t }
 
 val poly_fields : t PolyFields.t

infer/src/backend/callbacks.ml

@@ -7,37 +7,35 @@
 
 open! IStd
 
-(** Module to register and invoke callbacks *)
-
 type proc_callback_args =
   {get_procs_in_file: Procname.t -> Procname.t list; summary: Summary.t; exe_env: Exe_env.t}
 
 type proc_callback_t = proc_callback_args -> Summary.t
 
-type cluster_callback_args =
+type file_callback_args =
   {procedures: Procname.t list; source_file: SourceFile.t; exe_env: Exe_env.t}
 
-type cluster_callback_t = cluster_callback_args -> unit
+type file_callback_t = file_callback_args -> unit
 
 type procedure_callback =
-  {name: string; dynamic_dispatch: bool; language: Language.t; callback: proc_callback_t}
+  {checker_name: string; dynamic_dispatch: bool; language: Language.t; callback: proc_callback_t}
 
-type cluster_callback = {name: string; language: Language.t; callback: cluster_callback_t}
+type file_callback = {checker_name: string; language: Language.t; callback: file_callback_t}
 
 let procedure_callbacks = ref []
 
-let cluster_callbacks = ref []
+let file_callbacks = ref []
 
-let register_procedure_callback ~name ?(dynamic_dispatch = false) language
+let register_procedure_callback ~checker_name ?(dynamic_dispatch = false) language
     (callback : proc_callback_t) =
-  procedure_callbacks := {name; dynamic_dispatch; language; callback} :: !procedure_callbacks
+  procedure_callbacks :=
+    {checker_name; dynamic_dispatch; language; callback} :: !procedure_callbacks
 
 
-let register_cluster_callback ~name language (callback : cluster_callback_t) =
+let register_file_callback ~checker_name language (callback : file_callback_t) =
-  cluster_callbacks := {name; language; callback} :: !cluster_callbacks
+  file_callbacks := {checker_name; language; callback} :: !file_callbacks
 
 
-(** Invoke all registered procedure callbacks on the given procedure. *)
 let iterate_procedure_callbacks exe_env summary =
   let proc_desc = Summary.get_proc_desc summary in
   let proc_name = Procdesc.get_proc_name proc_desc in
@@ -55,11 +53,11 @@ let iterate_procedure_callbacks exe_env summary =
   in
   let is_specialized = Procdesc.is_specialized proc_desc in
   List.fold ~init:summary
-    ~f:(fun summary {name; dynamic_dispatch; language; callback} ->
+    ~f:(fun summary {checker_name; dynamic_dispatch; language; callback} ->
       if Language.equal language procedure_language && (dynamic_dispatch || not is_specialized) then (
         PerfEvent.(
           log (fun logger ->
-              log_begin_event logger ~name ~categories:["backend"]
+              log_begin_event logger ~name:checker_name ~categories:["backend"]
                 ~arguments:[("proc", `String (Procname.to_string proc_name))]
                 () )) ;
         let summary = callback {get_procs_in_file; summary; exe_env} in
@@ -69,9 +67,8 @@ let iterate_procedure_callbacks exe_env summary =
     !procedure_callbacks
 
 
-(** Invoke all registered cluster callbacks on a cluster of procedures. *)
+let iterate_file_callbacks procedures exe_env source_file =
-let iterate_cluster_callbacks procedures exe_env source_file =
+  if not (List.is_empty !file_callbacks) then
-  if not (List.is_empty !cluster_callbacks) then
     let environment = {procedures; source_file; exe_env} in
     let language_matches language =
       match procedures with
@@ -85,4 +82,4 @@ let iterate_cluster_callbacks procedures exe_env source_file =
         if language_matches language then (
           Language.curr_language := language ;
           callback environment ) )
-      !cluster_callbacks
+      !file_callbacks

infer/src/backend/callbacks.mli

@@ -7,7 +7,18 @@
 
 open! IStd
 
-(** Module to register and invoke callbacks *)
+(** Module to register and invoke checkers' callbacks. *)
+
+(* There are two types of callbacks:
+   1) Procedure level callback.
+      The checker is responsible for updating checker-specific summary artifacts for that procedure, including
+      writing errors that were detected during per-procedure analysis.
+   2) File level callback. Guaranteed to be invoked after all procedure-level callbacks for this files are invoked,
+      and generated summaries are serialized.
+      The checker is responsible for doing any additional work, but SHALL NOT modify summaries at this point, including updating
+      summaries' error log.
+      Additional errors can be issued at this stage using capabilities of [IssueLog].
+ *)
 
 type proc_callback_args =
   {get_procs_in_file: Procname.t -> Procname.t list; summary: Summary.t; exe_env: Exe_env.t}
@@ -20,20 +31,21 @@ type proc_callback_args =
     - Procedure for the callback to act on. *)
 type proc_callback_t = proc_callback_args -> Summary.t
 
-type cluster_callback_args =
+type file_callback_args =
   {procedures: Procname.t list; source_file: SourceFile.t; exe_env: Exe_env.t}
 
-type cluster_callback_t = cluster_callback_args -> unit
+type file_callback_t = file_callback_args -> unit
 
 val register_procedure_callback :
-  name:string -> ?dynamic_dispatch:bool -> Language.t -> proc_callback_t -> unit
+  checker_name:string -> ?dynamic_dispatch:bool -> Language.t -> proc_callback_t -> unit
-(** register a procedure callback *)
+(** Register a procedure callback (see details above) *)
 
-val register_cluster_callback : name:string -> Language.t -> cluster_callback_t -> unit
+val register_file_callback : checker_name:string -> Language.t -> file_callback_t -> unit
-(** register a cluster callback *)
+(** Register a file callback (see details above) *)
 
 val iterate_procedure_callbacks : Exe_env.t -> Summary.t -> Summary.t
 (** Invoke all registered procedure callbacks on the given procedure. *)
 
-val iterate_cluster_callbacks : Procname.t list -> Exe_env.t -> SourceFile.t -> unit
+val iterate_file_callbacks : Procname.t list -> Exe_env.t -> SourceFile.t -> unit
-(** Invoke all registered cluster callbacks on a cluster of procedures. *)
+(** Invoke all registered file callbacks on a file. Guaranteed to be called after all
+    procedure-level callbacks are invoked *)

infer/src/backend/ondemand.ml

@@ -381,13 +381,13 @@ let analyze_procedures exe_env procs_to_analyze source_file_opt =
   Option.iter source_file_opt ~f:(fun source_file ->
       if Config.dump_duplicate_symbols then dump_duplicate_procs source_file procs_to_analyze ) ;
   Option.iter source_file_opt ~f:(fun source_file ->
-      Callbacks.iterate_cluster_callbacks procs_to_analyze exe_env source_file ;
+      Callbacks.iterate_file_callbacks procs_to_analyze exe_env source_file ;
       create_perf_stats_report source_file ) ;
   unset_exe_env () ;
   Language.curr_language := saved_language
 
 
-(** Invoke all procedure and cluster callbacks on a given environment. *)
+(** Invoke all procedure-level and file-level callbacks on a given environment. *)
 let analyze_file (exe_env : Exe_env.t) source_file =
   let procs_to_analyze = SourceFiles.proc_names_of_source source_file in
   analyze_procedures exe_env procs_to_analyze (Some source_file)

infer/src/backend/printer.ml

@@ -236,7 +236,6 @@ end = struct
         Errlog.update global_err_log (Summary.get_err_log summary)
 
 
-  (** Create filename.ext.html. *)
   let write_html_file filename procs =
     let fname_encoding = DB.source_file_encoding filename in
     let fd, fmt = Io_infer.Html.create filename [".."; fname_encoding] in

infer/src/backend/printer.mli

@@ -38,4 +38,5 @@ val write_proc_html : Procdesc.t -> unit
 (** Write html file for the procedure. *)
 
 val write_all_html_files : SourceFile.t -> unit
-(** Create filename.ext.html for each file in the cluster. *)
+(** Group procedures used in the file by their corresponding source files, and create
+    filename.ext.html for each such a file. *)

infer/src/checkers/registerCheckers.ml

@@ -21,7 +21,7 @@ let () =
 type callback_fun =
   | Procedure of Callbacks.proc_callback_t
   | DynamicDispatch of Callbacks.proc_callback_t
-  | Cluster of Callbacks.cluster_callback_t
+  | File of Callbacks.file_callback_t
 
 type callback = callback_fun * Language.t
 
@@ -101,8 +101,8 @@ let all_checkers =
     ; callbacks=
         [ (Procedure RacerD.analyze_procedure, Language.Clang)
         ; (Procedure RacerD.analyze_procedure, Language.Java)
-        ; (Cluster RacerD.file_analysis, Language.Clang)
+        ; (File RacerD.file_analysis, Language.Clang)
-        ; (Cluster RacerD.file_analysis, Language.Java) ] }
+        ; (File RacerD.file_analysis, Language.Java) ] }
     (* toy resource analysis to use in the infer lab, see the lab/ directory *)
   ; { name= "resource leak"
     ; active= Config.resource_leak
@@ -131,9 +131,9 @@ let all_checkers =
     ; active= Config.starvation
     ; callbacks=
         [ (Procedure Starvation.analyze_procedure, Language.Java)
-        ; (Cluster Starvation.reporting, Language.Java)
+        ; (File Starvation.reporting, Language.Java)
         ; (Procedure Starvation.analyze_procedure, Language.Clang)
-        ; (Cluster Starvation.reporting, Language.Clang) ] }
+        ; (File Starvation.reporting, Language.Clang) ] }
   ; { name= "purity"
     ; active= Config.purity || Config.loop_hoisting
     ; callbacks=
@@ -156,11 +156,12 @@ let register checkers =
     let register_callback (callback, language) =
       match callback with
       | Procedure procedure_cb ->
-          Callbacks.register_procedure_callback ~name language procedure_cb
+          Callbacks.register_procedure_callback ~checker_name:name language procedure_cb
       | DynamicDispatch procedure_cb ->
-          Callbacks.register_procedure_callback ~name ~dynamic_dispatch:true language procedure_cb
+          Callbacks.register_procedure_callback ~checker_name:name ~dynamic_dispatch:true language
-      | Cluster cluster_cb ->
+            procedure_cb
-          Callbacks.register_cluster_callback ~name language cluster_cb
+      | File file_cb ->
+          Callbacks.register_file_callback ~checker_name:name language file_cb
     in
     List.iter ~f:register_callback callbacks
   in

infer/src/concurrency/RacerD.ml

@@ -1151,7 +1151,7 @@ let aggregate_by_class exe_env procedures =
 (* Gathers results by analyzing all the methods in a file, then
    post-processes the results to check an (approximation of) thread
    safety *)
-let file_analysis ({procedures; source_file; exe_env} : Callbacks.cluster_callback_args) =
+let file_analysis ({procedures; source_file; exe_env} : Callbacks.file_callback_args) =
   let class_map = aggregate_by_class exe_env procedures in
   Typ.Name.Map.fold
     (fun classname methods issue_log ->

infer/src/concurrency/RacerD.mli

@@ -7,6 +7,6 @@
 
 open! IStd
 
-val file_analysis : Callbacks.cluster_callback_t
+val file_analysis : Callbacks.file_callback_t
 
 val analyze_procedure : Callbacks.proc_callback_t

infer/src/concurrency/starvation.mli

@@ -9,6 +9,6 @@ open! IStd
 
 val analyze_procedure : Callbacks.proc_callback_t
 
-val reporting : Callbacks.cluster_callback_t
+val reporting : Callbacks.file_callback_t
 
 val whole_program_analysis : unit -> unit