i#4083 func trace: Change the -record_function interface

Changes -record_function interface to *not* specify an id.  The tracer
now assigns id's, with separate id's per library!function instance.

Each "library!function" string is written to a new "funclist" file in
the destination directory, with the line # as the id.  Uses the
passed-in file write interface to write the file.

Exposes drmemtrace_get_funclist_path() in the drmemtrace.h interface
and adds a test of it.

A separate change will add a new option "op_funclist_file" and file
reading support for the forthcoming func_view tool.

Issue: #4083

api/docs/release.dox

@@ -144,6 +144,10 @@ changes:
    This is a binary compatibility change for the OPSZ_ enum.
  - Added a new encoding hint field to #instr_t.
  - Added a requirement that a C++11-complient compiler be used with \ref page_droption.
+ - Changed the syntax of the drcachesim -record_function option.  It now takes
+   pairs instead of triples, with the identifier computed by the tracer.
+   The tracer writes out a file listing the "library!symbol" functions traced,
+   with the line ordinal indicating the identifier.
 
 The changes between version \DR_VERSION and 7.1.0 include the following minor
 compatibility changes:

clients/drcachesim/common/options.cpp

@@ -405,22 +405,24 @@ droption_t<std::string> op_record_function(
     "Record invocations trace for the specified function(s).",
     "Record invocations trace for the specified function(s) in the option"
     " value. Default value is empty. The value should fit this format:"
-    " function_name|function_id|func_args_num"
-    " (e.g., -record_function \"memset|10|3\"). The trace would contain"
+    " function_name|func_args_num"
+    " (e.g., -record_function \"memset|3\"). The trace would contain"
     " information for function return address, function argument value(s),"
     " and function return value. We only record pointer-sized arguments and"
-    " return value. The trace is labeled with the function_id via an ID entry"
-    " prior to each set of value entries."
+    " return values. The trace identifies which function is involved"
+    " via a numeric ID entry prior to each set of value entries."
+    " The mapping from numeric ID to library-qualified symbolic name is recorded"
+    " during tracing in a file \"funclist.log\" which simply lists the names in ID"
+    " order starting from 0, one name per line."
     " If the target function is in the dynamic symbol table, then the function_name"
     " should be a mangled name (e.g. \"_Znwm\" for \"operator new\", \"_ZdlPv\" for"
     " \"operator delete\"). Otherwise, the function_name should be a demangled name."
     " Recording multiple functions can be achieved by using the separator"
     " \"" OP_RECORD_FUNC_ITEM_SEP
-    "\" (e.g., -record_function \"memset|10|3" OP_RECORD_FUNC_ITEM_SEP
-    "memcpy|11|3\"), or"
+    "\" (e.g., -record_function \"memset|3" OP_RECORD_FUNC_ITEM_SEP "memcpy|3\"), or"
     " specifying multiple -record_function options (e.g., -record_function"
-    " \"memset|10|3\" -record_function \"memcpy|11|3\")."
-    " Note that the provided function id should be unique, and not collide with"
+    " \"memset|3\" -record_function \"memcpy|3\")."
+    " Note that the provided function name should be unique, and not collide with"
     " existing heap functions (see -record_heap_value) if -record_heap"
     " option is enabled.");
 droption_t<bool> op_record_heap(
@@ -433,10 +435,10 @@ droption_t<bool> op_record_heap(
 droption_t<std::string> op_record_heap_value(
     DROPTION_SCOPE_ALL, "record_heap_value", DROPTION_FLAG_ACCUMULATE,
     OP_RECORD_FUNC_ITEM_SEP,
-    "malloc|0|1" OP_RECORD_FUNC_ITEM_SEP "free|1|1" OP_RECORD_FUNC_ITEM_SEP
-    "tc_malloc|2|1" OP_RECORD_FUNC_ITEM_SEP "tc_free|3|1" OP_RECORD_FUNC_ITEM_SEP
-    "__libc_malloc|4|1" OP_RECORD_FUNC_ITEM_SEP "__libc_free|5|1" OP_RECORD_FUNC_ITEM_SEP
-    "calloc|6|2",
+    "malloc|1" OP_RECORD_FUNC_ITEM_SEP "free|1" OP_RECORD_FUNC_ITEM_SEP
+    "tc_malloc|1" OP_RECORD_FUNC_ITEM_SEP "tc_free|1" OP_RECORD_FUNC_ITEM_SEP
+    "__libc_malloc|1" OP_RECORD_FUNC_ITEM_SEP "__libc_free|1" OP_RECORD_FUNC_ITEM_SEP
+    "calloc|2",
     "Functions recorded by -record_heap",
     "Functions recorded by -record_heap. The option value should fit the same"
     " format required by -record_function. These functions will not"

clients/drcachesim/tests/burst_malloc.cpp

@@ -39,8 +39,11 @@
 /* Like burst_static we deliberately do not include configure.h here */
 
 #include "dr_api.h"
+#include "drmemtrace/drmemtrace.h"
 #include <assert.h>
 #include <iostream>
+#include <fstream>
+#include <string>
 #include <math.h>
 #include <stdlib.h>
 
@@ -86,6 +89,28 @@ do_some_work(int arg)
     return (temp > 0);
 }
 
+static void
+exit_cb(void *)
+{
+    const char *funclist_path;
+    drmemtrace_status_t res = drmemtrace_get_funclist_path(&funclist_path);
+    assert(res == DRMEMTRACE_SUCCESS);
+    std::ifstream stream(funclist_path);
+    assert(stream.good());
+    std::string line;
+    bool found_malloc = false;
+    bool found_return_big_value = false;
+    while (std::getline(stream, line)) {
+        assert(line.find('!') != std::string::npos);
+        if (line.find("!return_big_value") != std::string::npos)
+            found_return_big_value = true;
+        if (line.find("!malloc") != std::string::npos)
+            found_malloc = true;
+    }
+    assert(found_malloc);
+    assert(found_return_big_value);
+}
+
 int
 main(int argc, const char *argv[])
 {
@@ -94,9 +119,12 @@ main(int argc, const char *argv[])
                    "-stderr_mask 0xc -rstats_to_stderr"
                    " -client_lib ';;-offline -record_heap"
                    // Test large values that require two entries.
-                   " -record_function \"malloc|0|1&return_big_value|42|1\"'"))
+                   " -record_function \"malloc|1&return_big_value|1\"'"))
         std::cerr << "failed to set env var!\n";
 
+    drmemtrace_status_t res = drmemtrace_buffer_handoff(nullptr, exit_cb, nullptr);
+    assert(res == DRMEMTRACE_SUCCESS);
+
     for (int i = 0; i < 3; i++) {
         std::cerr << "pre-DR init\n";
         dr_app_setup();

clients/drcachesim/tests/offline-burst_malloc.templatex

@@ -1,5 +1,5 @@
 pre-DR init
-Warning: duplicated function id .*
+Warning: duplicated function name .*
 .*
 pre-DR start
 pre-DR detach

clients/drcachesim/tracer/drmemtrace.h

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2016-2018 Google, Inc.  All rights reserved.
+ * Copyright (c) 2016-2020 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -213,9 +213,16 @@ drmemtrace_buffer_handoff(drmemtrace_handoff_func_t handoff_func,
  * Its creation can be customized using drmemtrace_custom_module_data()
  * and then modified before passing to raw2trace via
  * drmodtrack_add_custom_data() and drmodtrack_offline_write().
+ * Use drmemtrace_get_modlist_path() to obtain the full path.
  */
 #define DRMEMTRACE_MODULE_LIST_FILENAME "modules.log"
 
+/**
+ * The name of the file in -offline mode where function tracing names
+ * are written.  Use drmemtrace_get_funclist_path() to obtain the full path.
+ */
+#define DRMEMTRACE_FUNCTION_MAP_FILENAME "funclist.log"
+
 DR_EXPORT
 /**
  * Retrieves the full path to the output directory in -offline mode
@@ -233,6 +240,16 @@ DR_EXPORT
 drmemtrace_status_t
 drmemtrace_get_modlist_path(OUT const char **path);
 
+DR_EXPORT
+/**
+ * Retrieves the full path to the file in -offline mode where function tracing
+ * information is written.  Each "library!symbol" function that was traced occupies
+ * one line of the file, with the line number (starting from 0) equal to the
+ * identifier recorded in the function trace entries.
+ */
+drmemtrace_status_t
+drmemtrace_get_funclist_path(OUT const char **path);
+
 DR_EXPORT
 /**
  * Adds custom data stored with each module in the module list produced for