Merge pull request #87 from P403n1x87/devel

Release v3.1.0

.github/workflows/release.yml

@@ -55,7 +55,7 @@ jobs:
           export PATH="/c/Program Files (x86)/`ls /c/Program\ Files\ \(x86\) | grep \"[wW]i[xX] [tT]oolset\"`/bin:$PATH"
           export VERSION=$(cat src/austin.h | sed -r -n "s/.*VERSION[ ]+\"(.+)\"/\1/p")
 
-          gcc -s -Wall -O3 -Os -o src/austin src/*.c -lpsapi -lntdll
+          gcc -static -s -Wall -O3 -Os -o src/austin src/*.c -lpsapi -lntdll
 
           git checkout "packaging/msi"
           git checkout master

ChangeLog

@@ -1,3 +1,10 @@
+2021-08-18 v3.1.0
+
+  Added garbage collection state sampling for Python 3.7 onward.
+
+  Bugfix: the MinGW libwinpthread library is now linked statically on Windows.
+
+
 2021-06-13 v3.0.0
 
   Added pipe mode.

README.md

@@ -286,6 +286,7 @@ Austin -- A frame stack sampler for Python.
   -e, --exclude-empty        Do not output samples of threads with no frame
                              stacks.
   -f, --full                 Produce the full set of metrics (time +mem -mem).
+  -g, --gc                   Sample the garbage collector state.
   -i, --interval=n_us        Sampling interval in microseconds (default is
                              100). Accepted units: s, ms, us.
   -m, --memory               Profile memory usage.
@@ -367,6 +368,17 @@ Austin can be told to profile multi-process applications with the `-C` or
 process.
 
 
+## Garbage Collector Sampling
+
+Austin can sample the Python garbage collector state for application running
+with Python 3.7 and later versions. If the `-g`/`--gc` option is passed, Austin
+will append `:GC:` at the end of each collected frame stack whenver the
+garbage collector is in the collecting state. This gives you a measure of how
+*busy* the Python GC is during a run.
+
+*Since Austin 3.1.0*.
+
+
 ## Logging
 
 Austin uses `syslog` on Linux and macOS, and `%TEMP%\austin.log` on Windows

configure.ac

@@ -2,7 +2,7 @@
 # Process this file with autoconf to produce a configure script.
 
 AC_PREREQ([2.69])
-AC_INIT([austin], [3.0.0], [https://github.com/p403n1x87/austin/issues])
+AC_INIT([austin], [3.1.0], [https://github.com/p403n1x87/austin/issues])
 AC_CONFIG_SRCDIR([config.h.in])
 AC_CONFIG_HEADERS([config.h])
 AM_INIT_AUTOMAKE

snap/snapcraft.yaml

@@ -1,5 +1,5 @@
 name: austin
-version: '3.0.0+git'
+version: '3.1.0+git'
 summary: A Python frame stack sampler for CPython
 description: |
   Austin is a Python frame stack sampler for CPython written in pure C. It

src/argparse.c

@@ -52,6 +52,7 @@ parsed_args_t pargs = {
   /* children            */ 0,
   /* exposure            */ 0,
   /* pipe                */ 0,
+  /* gc                  */ 0,
 };
 
 static int exec_arg = 0;
@@ -217,6 +218,10 @@ static struct argp_option options[] = {
     "pipe",         'P', NULL,          0,
     "Pipe mode. Use when piping Austin output."
   },
+  {
+    "gc",         'g', NULL,          0,
+    "Sample the garbage collector state."
+  },
   #ifndef PL_LINUX
   {
     "help",         '?', NULL
@@ -322,6 +327,10 @@ parse_opt (int key, char *arg, struct argp_state *state)
     pargs.pipe = 1;
     break;
 
+  case 'g':
+    pargs.gc = 1;
+    break;
+
   case ARGP_KEY_ARG:
   case ARGP_KEY_END:
     if (pargs.attach_pid != 0 && exec_arg != 0)
@@ -464,6 +473,7 @@ static const char * help_msg = \
 "  -e, --exclude-empty        Do not output samples of threads with no frame\n"
 "                             stacks.\n"
 "  -f, --full                 Produce the full set of metrics (time +mem -mem).\n"
+"  -g, --gc                   Sample the garbage collector state.\n"
 "  -i, --interval=n_us        Sampling interval in microseconds (default is\n"
 "                             100). Accepted units: s, ms, us.\n"
 "  -m, --memory               Profile memory usage.\n"
@@ -617,6 +627,10 @@ cb(const char opt, const char * arg) {
     pargs.pipe = 1;
     break;
 
+  case 'g':
+    pargs.gc = 1;
+    break;
+
   case '?':
     puts(help_msg);
     exit(0);

src/argparse.h

@@ -43,6 +43,7 @@ typedef struct {
   int       children;
   ctime_t   exposure;
   int       pipe;
+  int       gc;
 } parsed_args_t;
 
 

src/austin.1

@@ -1,7 +1,7 @@
 .\" DO NOT MODIFY THIS FILE!  It was generated by help2man 1.47.6.
-.TH AUSTIN "1" "June 2021" "austin 3.0.0" "User Commands"
+.TH AUSTIN "1" "August 2021" "austin 3.1.0" "User Commands"
 .SH NAME
-austin \- manual page for austin 3.0.0
+austin \- manual page for austin 3.1.0
 .SH SYNOPSIS
 .B austin
 [\fI\,OPTION\/\fR...] \fI\,command \/\fR[\fI\,ARG\/\fR...]
@@ -21,6 +21,9 @@ stacks.
 \fB\-f\fR, \fB\-\-full\fR
 Produce the full set of metrics (time +mem \fB\-mem\fR).
 .TP
+\fB\-g\fR, \fB\-\-gc\fR
+Sample the garbage collector state.
+.TP
 \fB\-i\fR, \fB\-\-interval\fR=\fI\,n_us\/\fR
 Sampling interval in microseconds (default is
 100). Accepted units: s, ms, us.

src/austin.c

@@ -289,7 +289,11 @@ int main(int argc, char ** argv) {
     austin_errno = EOK;
 
   // Log sampling metrics
-  NL;meta("duration: %lu", stats_duration());
+  NL;
+  meta("duration: %lu", stats_duration());
+  if (pargs.gc) {
+    meta("gc: %lu", _gc_time);
+  }
 
   stats_log_metrics();NL;
 

src/austin.h

@@ -24,6 +24,6 @@
 #define AUSTIN_H
 
 #define PROGRAM_NAME                    "austin"
-#define VERSION                         "3.0.0"
+#define VERSION                         "3.1.0"
 
 #endif

src/linux/py_thread.h

@@ -21,6 +21,7 @@
 // along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
 #include <fcntl.h>
+#include <pthread.h>
 #include <stdio.h>
 #include <string.h>
 #include <unistd.h>

src/py_proc.c

@@ -323,7 +323,7 @@ _py_proc__check_interp_state(py_proc_t * self, void * raddr) {
   if (py_proc__get_type(self, raddr, is))
     return OUT_OF_BOUND;
 
-  if (py_proc__get_type(self, is.tstate_head, tstate_head)) {
+  if (py_proc__get_type(self, V_FIELD(void *, is, py_is, o_tstate_head), tstate_head)) {
     log_t(
       "Cannot copy PyThreadState head at %p from PyInterpreterState instance",
       is.tstate_head
@@ -347,7 +347,7 @@ _py_proc__check_interp_state(py_proc_t * self, void * raddr) {
   );
 
   // As an extra sanity check, verify that the thread state is valid
-  raddr_t thread_raddr = { .pid = PROC_REF, .addr = is.tstate_head };
+  raddr_t thread_raddr = { .pid = PROC_REF, .addr = V_FIELD(void *, is, py_is, o_tstate_head) };
   py_thread_t thread;
   if (fail(py_thread__fill_from_raddr(&thread, &thread_raddr, self))) {
     log_d("Failed to fill thread structure");
@@ -361,6 +361,11 @@ _py_proc__check_interp_state(py_proc_t * self, void * raddr) {
 
   log_d("Stack trace constructed from possible interpreter state");
 
+  if (py_v->major == 3 && py_v->minor >= 9) {
+    self->gc_state_raddr = (void *) (((char *) raddr) + py_v->py_is.o_gc);
+    log_d("GC runtime state @ %p", self->gc_state_raddr);
+  }
+
   SUCCESS;
 }
 
@@ -475,6 +480,10 @@ _py_proc__deref_interp_head(py_proc_t * self) {
       FAIL;
     }
     interp_head_raddr = V_FIELD(void *, py_runtime, py_runtime, o_interp_head);
+    if (py_v->major == 3 && py_v->minor < 9) {
+      self->gc_state_raddr = self->py_runtime_raddr + py_v->py_runtime.o_gc;
+      log_d("GC runtime state @ %p", self->gc_state_raddr);
+    }
   }
   else if (self->interp_head_raddr != NULL) {
     if (py_proc__get_type(self, self->interp_head_raddr, interp_head_raddr)) {
@@ -706,7 +715,8 @@ _py_proc__run(py_proc_t * self, int try_once) {
   if (
     self->tstate_curr_raddr == NULL &&
     self->py_runtime_raddr  == NULL &&
-    self->interp_head_raddr == NULL
+    self->interp_head_raddr == NULL &&
+    self->gc_state_raddr    == NULL
   )
     log_w("No remote symbol references have been set.");
   #endif
@@ -746,6 +756,7 @@ py_proc_new() {
     return NULL;
 
   py_proc->min_raddr = (void *) -1;
+  py_proc->gc_state_raddr = NULL;
 
   // Pre-hash symbol names
   if (_dynsym_hash_array[0] == 0) {
@@ -1069,6 +1080,22 @@ _py_proc__get_memory_delta(py_proc_t * self) {
 }
 
 
+// ----------------------------------------------------------------------------
+int
+py_proc__is_gc_collecting(py_proc_t * self) {
+  if (!isvalid(self->gc_state_raddr))
+    return FALSE;
+
+  GCRuntimeState gc_state;
+  if (fail(py_proc__get_type(self, self->gc_state_raddr, gc_state))) {
+    log_d("Failed to get GC runtime state");
+    return -1;
+  }
+
+  return V_FIELD(int, gc_state, py_gc, o_collecting);
+}
+
+
 // ----------------------------------------------------------------------------
 int
 py_proc__sample(py_proc_t * self) {
@@ -1080,8 +1107,9 @@ py_proc__sample(py_proc_t * self) {
   if (fail(py_proc__get_type(self, self->is_raddr, is)))
     FAIL;
 
-  if (is.tstate_head != NULL) {
-    raddr_t raddr = { .pid = PROC_REF, .addr = is.tstate_head };
+  void * tstate_head = V_FIELD(void *, is, py_is, o_tstate_head);
+  if (isvalid(tstate_head)) {
+    raddr_t raddr = { .pid = PROC_REF, .addr = tstate_head };
     py_thread_t py_thread;
     if (fail(py_thread__fill_from_raddr(&py_thread, &raddr, self)))
       FAIL;

src/py_proc.h

@@ -64,6 +64,7 @@ typedef struct {
   void          * tstate_curr_raddr;
   void          * py_runtime_raddr;
   void          * interp_head_raddr;
+  void          * gc_state_raddr;
 
   void          * is_raddr;
 
@@ -148,6 +149,20 @@ int
 py_proc__is_python(py_proc_t *);
 
 
+/**
+ * Check whether the GC is collecting for the given process.
+ * 
+ * NOTE: This method makes sense only for Python>=3.7.
+ * 
+ * @param py_proc_t * the process object.
+ * 
+ * @return TRUE if the GC is collecting, FALSE otherwise.
+ * 
+ */
+int
+py_proc__is_gc_collecting(py_proc_t *);
+
+
 /**
  * Sample the frame stack of each thread of the given Python process.
  *

src/py_thread.c

@@ -30,7 +30,6 @@
 #include "logging.h"
 #include "mem.h"
 #include "platform.h"
-#include "pthread.h"
 #include "timing.h"
 #include "version.h"
 
@@ -489,6 +488,11 @@ py_thread__print_collapsed_stack(py_thread_t * self, ctime_t time_delta, ssize_t
     }
   }
 
+  if (pargs.gc && py_proc__is_gc_collecting(self->proc) == TRUE) {
+    fprintf(pargs.output_file, ";:GC:");
+    stats_gc_time(time_delta);
+  }
+
   // Finish off sample with the metric(s)
   if (pargs.full) {
     fprintf(pargs.output_file, " " TIME_METRIC METRIC_SEP IDLE_METRIC METRIC_SEP MEM_METRIC "\n",