bootsnap 1.17.1 → 1.18.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 13123a80c1d6f0e73ff602768433dd998a28ab8e63caec7d0e29188990c4afe9
-  data.tar.gz: cc2e19bea1080f9ce2337095fcb161407e6def9e3412c38fdbb03c0e0ddc2a81
+  metadata.gz: 52fa7adde2726e79c8d06b4ca8a450d0ecc0fd143b377a0da3aa27965e5ecd52
+  data.tar.gz: 7029e821fbd1da561f14fce91cc3ca72f8ee80eb6ead5a099bc37b74233c9dce
 SHA512:
-  metadata.gz: 8fbd0f6652387c744f0a345d46e9d7238c1609597c8e26f93ad92cd928178f7c5a3afd2075390c5561c3ace19084c318b27ff3acec9d96bd0153c53ea8a4cb95
-  data.tar.gz: f35f769e3b39c8f28857a74c874e77cdbe04b4db0df3141fa988ba26ea55bf485e8d560b4cab65d5887508ac2b927eb97d49b6791321b0a68c896db2b617eb6d
+  metadata.gz: bce0d723ae15e1fcc8961b698977977a49587d4fe514c76dd10d97453808c6562681e9df80dd3ecec2f1eff70f8a54592027c699d37c4c39e9254c6b63507f6b
+  data.tar.gz: 8272f4541d3789e8eb390319ec04d202b90fc0745696f73d414a7dc1365380246ec7b44746b3592abf19883eae709f2daf4e5033677b7be3490dc0ad642fb6d7

data/CHANGELOG.md

@@ -1,5 +1,15 @@
 # Unreleased
 
+# 1.18.0
+
+* `Bootsnap.instrumentation` now receive `:hit` events.
+* Add `Bootsnap.log_stats!` to print hit rate statistics on process exit. Can also be enabled with `BOOTSNAP_STATS=1`.
+* Revalidate stale cache entries by digesting the source content.
+  This should significantly improve performance in environments where `mtime` isn't preserved (e.g. CI systems doing a git clone, etc).
+  See #468.
+* Open source files and cache entries with `O_NOATIME` when available to reduce disk accesses. See #469.
+* `bootsnap precompile --gemfile` now look for `.rb` files in the whole gem and not just the `lib/` directory. See #466.
+
 # 1.17.1
 
 * Fix a compatibility issue with the `prism` library that ships with Ruby 3.3. See #463.

data/README.md

@@ -81,6 +81,7 @@ well together.
 - `DISABLE_BOOTSNAP_COMPILE_CACHE` allows to disable ISeq and YAML caches.
 - `BOOTSNAP_READONLY` configure bootsnap to not update the cache on miss or stale entries.
 - `BOOTSNAP_LOG` configure bootsnap to log all caches misses to STDERR.
+- `BOOTSNAP_STATS` log hit rate statistics on exit. Can't be used if `BOOTSNAP_LOG` is enabled.
 - `BOOTSNAP_IGNORE_DIRECTORIES` a comma separated list of directories that shouldn't be scanned.
   Useful when you have large directories of non-ruby files inside `$LOAD_PATH`.
   It defaults to ignore any directory named `node_modules`.
@@ -99,8 +100,8 @@ Bootsnap cache misses can be monitored though a callback:
 Bootsnap.instrumentation = ->(event, path) { puts "#{event} #{path}" }
 ```
 
-`event` is either `:miss` or `:stale`. You can also call `Bootsnap.log!` as a shortcut to
-log all events to STDERR.
+`event` is either `:hit`, `:miss`, `:stale` or `:revalidated`.
+You can also call `Bootsnap.log!` as a shortcut to log all events to STDERR.
 
 To turn instrumentation back off you can set it to nil:
 

data/ext/bootsnap/bootsnap.c

@@ -20,6 +20,10 @@
 #include <fcntl.h>
 #include <sys/stat.h>
 
+#ifndef O_NOATIME
+#define O_NOATIME 0
+#endif
+
 /* 1000 is an arbitrary limit; FNV64 plus some slashes brings the cap down to
  * 981 for the cache dir */
 #define MAX_CACHEPATH_SIZE 1000
@@ -30,7 +34,7 @@
 #define MAX_CREATE_TEMPFILE_ATTEMPT 3
 
 #ifndef RB_UNLIKELY
-  #define RB_UNLIKELY(x) (x)
+#define RB_UNLIKELY(x) (x)
 #endif
 
 /*
@@ -54,8 +58,10 @@ struct bs_cache_key {
   uint32_t ruby_revision;
   uint64_t size;
   uint64_t mtime;
-  uint64_t data_size; /* not used for equality */
-  uint8_t  pad[24];
+  uint64_t data_size; //
+  uint64_t digest;
+  uint8_t digest_set;
+  uint8_t pad[15];
 } __attribute__((packed));
 
 /*
@@ -69,7 +75,7 @@ struct bs_cache_key {
 STATIC_ASSERT(sizeof(struct bs_cache_key) == KEY_SIZE);
 
 /* Effectively a schema version. Bumping invalidates all previous caches */
-static const uint32_t current_version = 4;
+static const uint32_t current_version = 5;
 
 /* hash of e.g. "x86_64-darwin17", invalidating when ruby is recompiled on a
  * new OS ABI, etc. */
@@ -87,8 +93,7 @@ static VALUE rb_mBootsnap_CompileCache;
 static VALUE rb_mBootsnap_CompileCache_Native;
 static VALUE rb_cBootsnap_CompileCache_UNCOMPILABLE;
 static ID instrumentation_method;
-static VALUE sym_miss;
-static VALUE sym_stale;
+static VALUE sym_hit, sym_miss, sym_stale, sym_revalidated;
 static bool instrumentation_enabled = false;
 static bool readonly = false;
 
@@ -100,9 +105,18 @@ static VALUE bs_rb_fetch(VALUE self, VALUE cachedir_v, VALUE path_v, VALUE handl
 static VALUE bs_rb_precompile(VALUE self, VALUE cachedir_v, VALUE path_v, VALUE handler);
 
 /* Helpers */
+enum cache_status {
+  miss,
+  hit,
+  stale,
+};
 static void bs_cache_path(const char * cachedir, const VALUE path, char (* cache_path)[MAX_CACHEPATH_SIZE]);
 static int bs_read_key(int fd, struct bs_cache_key * key);
-static int cache_key_equal(struct bs_cache_key * k1, struct bs_cache_key * k2);
+static enum cache_status cache_key_equal_fast_path(struct bs_cache_key * k1, struct bs_cache_key * k2);
+static int cache_key_equal_slow_path(struct bs_cache_key * current_key, struct bs_cache_key * cached_key, const VALUE input_data);
+static int update_cache_key(struct bs_cache_key *current_key, int cache_fd, const char ** errno_provenance);
+
+static void bs_cache_key_digest(struct bs_cache_key * key, const VALUE input_data);
 static VALUE bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args);
 static VALUE bs_precompile(char * path, VALUE path_v, char * cache_path, VALUE handler);
 static int open_current_file(char * path, struct bs_cache_key * key, const char ** errno_provenance);
@@ -161,11 +175,10 @@ Init_bootsnap(void)
 
   instrumentation_method = rb_intern("_instrument");
 
+  sym_hit = ID2SYM(rb_intern("hit"));
   sym_miss = ID2SYM(rb_intern("miss"));
-  rb_global_variable(&sym_miss);
-
   sym_stale = ID2SYM(rb_intern("stale"));
-  rb_global_variable(&sym_stale);
+  sym_revalidated = ID2SYM(rb_intern("revalidated"));
 
   rb_define_module_function(rb_mBootsnap, "instrumentation_enabled=", bs_instrumentation_enabled_set, 1);
   rb_define_module_function(rb_mBootsnap_CompileCache_Native, "readonly=", bs_readonly_set, 1);
@@ -185,6 +198,14 @@ bs_instrumentation_enabled_set(VALUE self, VALUE enabled)
   return enabled;
 }
 
+static inline void
+bs_instrumentation(VALUE event, VALUE path)
+{
+    if (RB_UNLIKELY(instrumentation_enabled)) {
+       rb_funcall(rb_mBootsnap, instrumentation_method, 2, event, path);
+    }
+}
+
 static VALUE
 bs_readonly_set(VALUE self, VALUE enabled)
 {
@@ -290,17 +311,53 @@ bs_cache_path(const char * cachedir, const VALUE path, char (* cache_path)[MAX_C
  * The data_size member is not compared, as it serves more of a "header"
  * function.
  */
-static int
-cache_key_equal(struct bs_cache_key * k1, struct bs_cache_key * k2)
+static enum cache_status cache_key_equal_fast_path(struct bs_cache_key *k1,
+                                     struct bs_cache_key *k2) {
+  if (k1->version == k2->version &&
+          k1->ruby_platform == k2->ruby_platform &&
+          k1->compile_option == k2->compile_option &&
+          k1->ruby_revision == k2->ruby_revision && k1->size == k2->size) {
+    return (k1->mtime == k2->mtime) ? hit : stale;
+  }
+  return miss;
+}
+
+static int cache_key_equal_slow_path(struct bs_cache_key *current_key,
+                                     struct bs_cache_key *cached_key,
+                                     const VALUE input_data)
+{
+  bs_cache_key_digest(current_key, input_data);
+  return current_key->digest == cached_key->digest;
+}
+
+static int update_cache_key(struct bs_cache_key *current_key, int cache_fd, const char ** errno_provenance)
 {
-  return (
-    k1->version        == k2->version        &&
-    k1->ruby_platform  == k2->ruby_platform  &&
-    k1->compile_option == k2->compile_option &&
-    k1->ruby_revision  == k2->ruby_revision  &&
-    k1->size           == k2->size           &&
-    k1->mtime          == k2->mtime
-  );
+  lseek(cache_fd, 0, SEEK_SET);
+  ssize_t nwrite = write(cache_fd, current_key, KEY_SIZE);
+  if (nwrite < 0) {
+      *errno_provenance = "update_cache_key:write";
+      return -1;
+  }
+
+#ifdef HAVE_FDATASYNC
+  if (fdatasync(cache_fd) < 0) {
+      *errno_provenance = "update_cache_key:fdatasync";
+      return -1;
+  }
+#endif
+
+  return 0;
+}
+
+/*
+ * Fills the cache key digest.
+ */
+static void bs_cache_key_digest(struct bs_cache_key *key,
+                                const VALUE input_data) {
+  if (key->digest_set)
+    return;
+  key->digest = fnv1a_64(input_data);
+  key->digest_set = 1;
 }
 
 /*
@@ -366,7 +423,7 @@ open_current_file(char * path, struct bs_cache_key * key, const char ** errno_pr
   struct stat statbuf;
   int fd;
 
-  fd = open(path, O_RDONLY);
+  fd = open(path, O_RDONLY | O_NOATIME);
   if (fd < 0) {
     *errno_provenance = "bs_fetch:open_current_file:open";
     return fd;
@@ -389,6 +446,7 @@ open_current_file(char * path, struct bs_cache_key * key, const char ** errno_pr
   key->ruby_revision  = current_ruby_revision;
   key->size           = (uint64_t)statbuf.st_size;
   key->mtime          = (uint64_t)statbuf.st_mtime;
+  key->digest_set     = false;
 
   return fd;
 }
@@ -432,7 +490,12 @@ open_cache_file(const char * path, struct bs_cache_key * key, const char ** errn
 {
   int fd, res;
 
-  fd = open(path, O_RDONLY);
+  if (readonly) {
+    fd = open(path, O_RDONLY | O_NOATIME);
+  } else {
+    fd = open(path, O_RDWR | O_NOATIME);
+  }
+
   if (fd < 0) {
     *errno_provenance = "bs_fetch:open_cache_file:open";
     return CACHE_MISS;
@@ -677,7 +740,8 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
   int res, valid_cache = 0, exception_tag = 0;
   const char * errno_provenance = NULL;
 
-  VALUE input_data;   /* data read from source file, e.g. YAML or ruby source */
+  VALUE status = Qfalse;
+  VALUE input_data = Qfalse;   /* data read from source file, e.g. YAML or ruby source */
   VALUE storage_data; /* compiled data, e.g. msgpack / binary iseq */
   VALUE output_data;  /* return data, e.g. ruby hash or loaded iseq */
 
@@ -695,20 +759,44 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
   cache_fd = open_cache_file(cache_path, &cached_key, &errno_provenance);
   if (cache_fd == CACHE_MISS || cache_fd == CACHE_STALE) {
     /* This is ok: valid_cache remains false, we re-populate it. */
-    if (RB_UNLIKELY(instrumentation_enabled)) {
-      rb_funcall(rb_mBootsnap, instrumentation_method, 2, cache_fd == CACHE_MISS ? sym_miss : sym_stale, path_v);
-    }
+    bs_instrumentation(cache_fd == CACHE_MISS ? sym_miss : sym_stale, path_v);
   } else if (cache_fd < 0) {
     exception_message = rb_str_new_cstr(cache_path);
     goto fail_errno;
   } else {
     /* True if the cache existed and no invalidating changes have occurred since
      * it was generated. */
-    valid_cache = cache_key_equal(&current_key, &cached_key);
-    if (RB_UNLIKELY(instrumentation_enabled)) {
-      if (!valid_cache) {
-        rb_funcall(rb_mBootsnap, instrumentation_method, 2, sym_stale, path_v);
+
+    switch(cache_key_equal_fast_path(&current_key, &cached_key)) {
+    case hit:
+      status = sym_hit;
+      valid_cache = true;
+      break;
+    case miss:
+      valid_cache = false;
+      break;
+    case stale:
+      valid_cache = false;
+      if ((input_data = bs_read_contents(current_fd, current_key.size,
+                                     &errno_provenance)) == Qfalse) {
+        exception_message = path_v;
+        goto fail_errno;
+      }
+      valid_cache = cache_key_equal_slow_path(&current_key, &cached_key, input_data);
+      if (valid_cache) {
+        if (!readonly) {
+          if (update_cache_key(&current_key, cache_fd, &errno_provenance)) {
+              exception_message = path_v;
+              goto fail_errno;
+          }
+        }
+        status = sym_revalidated;
       }
+      break;
+    };
+
+    if (!valid_cache) {
+      status = sym_stale;
     }
   }
 
@@ -722,7 +810,7 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
     else if (res == CACHE_UNCOMPILABLE) {
       /* If fetch_cached_data returned `Uncompilable` we fallback to `input_to_output`
         This happens if we have say, an unsafe YAML cache, but try to load it in safe mode */
-      if ((input_data = bs_read_contents(current_fd, current_key.size, &errno_provenance)) == Qfalse){
+      if (input_data == Qfalse && (input_data = bs_read_contents(current_fd, current_key.size, &errno_provenance)) == Qfalse) {
         exception_message = path_v;
         goto fail_errno;
       }
@@ -741,7 +829,7 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
   /* Cache is stale, invalid, or missing. Regenerate and write it out. */
 
   /* Read the contents of the source file into a buffer */
-  if ((input_data = bs_read_contents(current_fd, current_key.size, &errno_provenance)) == Qfalse){
+  if (input_data == Qfalse && (input_data = bs_read_contents(current_fd, current_key.size, &errno_provenance)) == Qfalse) {
     exception_message = path_v;
     goto fail_errno;
   }
@@ -763,6 +851,7 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
    * We do however ignore any failures to persist the cache, as it's better
    * to move along, than to interrupt the process.
    */
+  bs_cache_key_digest(&current_key, input_data);
   atomic_write_cache_file(cache_path, &current_key, storage_data, &errno_provenance);
 
   /* Having written the cache, now convert storage_data to output_data */
@@ -792,6 +881,7 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
   goto succeed; /* output_data is now the correct return. */
 
 #define CLEANUP \
+  if (status != Qfalse) bs_instrumentation(status, path_v); \
   if (current_fd >= 0)  close(current_fd); \
   if (cache_fd >= 0)    close(cache_fd);
 
@@ -818,13 +908,16 @@ invalid_type_storage_data:
 static VALUE
 bs_precompile(char * path, VALUE path_v, char * cache_path, VALUE handler)
 {
+  if (readonly) {
+    return Qfalse;
+  }
+
   struct bs_cache_key cached_key, current_key;
-  char * contents = NULL;
   int cache_fd = -1, current_fd = -1;
   int res, valid_cache = 0, exception_tag = 0;
   const char * errno_provenance = NULL;
 
-  VALUE input_data;   /* data read from source file, e.g. YAML or ruby source */
+  VALUE input_data = Qfalse;   /* data read from source file, e.g. YAML or ruby source */
   VALUE storage_data; /* compiled data, e.g. msgpack / binary iseq */
 
   /* Open the source file and generate a cache key for it */
@@ -840,7 +933,26 @@ bs_precompile(char * path, VALUE path_v, char * cache_path, VALUE handler)
   } else {
     /* True if the cache existed and no invalidating changes have occurred since
      * it was generated. */
-    valid_cache = cache_key_equal(&current_key, &cached_key);
+    switch(cache_key_equal_fast_path(&current_key, &cached_key)) {
+    case hit:
+      valid_cache = true;
+      break;
+    case miss:
+      valid_cache = false;
+      break;
+    case stale:
+      valid_cache = false;
+      if ((input_data = bs_read_contents(current_fd, current_key.size, &errno_provenance)) == Qfalse) {
+        goto fail;
+      }
+      valid_cache = cache_key_equal_slow_path(&current_key, &cached_key, input_data);
+       if (valid_cache) {
+         if (update_cache_key(&current_key, cache_fd, &errno_provenance)) {
+             goto fail;
+         }
+      }
+      break;
+    };
   }
 
   if (valid_cache) {
@@ -867,6 +979,7 @@ bs_precompile(char * path, VALUE path_v, char * cache_path, VALUE handler)
   if (!RB_TYPE_P(storage_data, T_STRING)) goto fail;
 
   /* Write the cache key and storage_data to the cache directory */
+  bs_cache_key_digest(&current_key, input_data);
   res = atomic_write_cache_file(cache_path, &current_key, storage_data, &errno_provenance);
   if (res < 0) goto fail;
 

data/ext/bootsnap/extconf.rb

@@ -3,21 +3,28 @@
 require "mkmf"
 
 if %w[ruby truffleruby].include?(RUBY_ENGINE)
-  $CFLAGS << " -O3 "
-  $CFLAGS << " -std=c99"
+  have_func "fdatasync", "fcntl.h"
+
+  unless RUBY_PLATFORM.match?(/mswin|mingw|cygwin/)
+    append_cppflags ["_GNU_SOURCE"] # Needed of O_NOATIME
+  end
+
+  append_cflags ["-O3", "-std=c99"]
 
   # ruby.h has some -Wpedantic fails in some cases
   # (e.g. https://github.com/Shopify/bootsnap/issues/15)
   unless ["0", "", nil].include?(ENV["BOOTSNAP_PEDANTIC"])
-    $CFLAGS << " -Wall"
-    $CFLAGS << " -Werror"
-    $CFLAGS << " -Wextra"
-    $CFLAGS << " -Wpedantic"
+    append_cflags([
+      "-Wall",
+      "-Werror",
+      "-Wextra",
+      "-Wpedantic",
 
-    $CFLAGS << " -Wno-unused-parameter" # VALUE self has to be there but we don't care what it is.
-    $CFLAGS << " -Wno-keyword-macro" # hiding return
-    $CFLAGS << " -Wno-gcc-compat" # ruby.h 2.6.0 on macos 10.14, dunno
-    $CFLAGS << " -Wno-compound-token-split-by-macro"
+      "-Wno-unused-parameter", # VALUE self has to be there but we don't care what it is.
+      "-Wno-keyword-macro", # hiding return
+      "-Wno-gcc-compat", # ruby.h 2.6.0 on macos 10.14, dunno
+      "-Wno-compound-token-split-by-macro",
+    ])
   end
 
   create_makefile("bootsnap/bootsnap")

data/lib/bootsnap/cli.rb

@@ -60,14 +60,16 @@ module Bootsnap
         precompile_json_files(main_sources)
 
         if compile_gemfile
-          # Some gems embed their tests, they're very unlikely to be loaded, so not worth precompiling.
-          gem_exclude = Regexp.union([exclude, "/spec/", "/test/"].compact)
-          precompile_ruby_files($LOAD_PATH.map { |d| File.expand_path(d) }, exclude: gem_exclude)
-
           # Gems that include JSON or YAML files usually don't put them in `lib/`.
           # So we look at the gem root.
+          # Similarly, gems that include Rails engines generally file Ruby files in `app/`.
+          # However some gems embed their tests, they're very unlikely to be loaded, so not worth precompiling.
+          gem_exclude = Regexp.union([exclude, "/spec/", "/test/", "/features/"].compact)
+
           gem_pattern = %r{^#{Regexp.escape(Bundler.bundle_path.to_s)}/?(?:bundler/)?gems/[^/]+}
-          gem_paths = $LOAD_PATH.map { |p| p[gem_pattern] }.compact.uniq
+          gem_paths = $LOAD_PATH.map { |p| p[gem_pattern] || p }.uniq
+
+          precompile_ruby_files(gem_paths, exclude: gem_exclude)
           precompile_yaml_files(gem_paths, exclude: gem_exclude)
           precompile_json_files(gem_paths, exclude: gem_exclude)
         end

data/lib/bootsnap/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Bootsnap
-  VERSION = "1.17.1"
+  VERSION = "1.18.0"
 end

data/lib/bootsnap.rb

@@ -11,6 +11,16 @@ module Bootsnap
   class << self
     attr_reader :logger
 
+    def log_stats!
+      stats = {hit: 0, revalidated: 0, miss: 0, stale: 0}
+      self.instrumentation = ->(event, _path) { stats[event] += 1 }
+      Kernel.at_exit do
+        stats.each do |event, count|
+          $stderr.puts "bootsnap #{event}: #{count}"
+        end
+      end
+    end
+
     def log!
       self.logger = $stderr.method(:puts)
     end
@@ -18,9 +28,9 @@ module Bootsnap
     def logger=(logger)
       @logger = logger
       self.instrumentation = if logger.respond_to?(:debug)
-        ->(event, path) { @logger.debug("[Bootsnap] #{event} #{path}") }
+        ->(event, path) { @logger.debug("[Bootsnap] #{event} #{path}") unless event == :hit }
       else
-        ->(event, path) { @logger.call("[Bootsnap] #{event} #{path}") }
+        ->(event, path) { @logger.call("[Bootsnap] #{event} #{path}") unless event == :hit }
       end
     end
 
@@ -110,6 +120,8 @@ module Bootsnap
 
         if ENV["BOOTSNAP_LOG"]
           log!
+        elsif ENV["BOOTSNAP_STATS"]
+          log_stats!
         end
       end
     end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: bootsnap
 version: !ruby/object:Gem::Version
-  version: 1.17.1
+  version: 1.18.0
 platform: ruby
 authors:
 - Burke Libbey
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2024-01-12 00:00:00.000000000 Z
+date: 2024-01-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: msgpack
@@ -83,7 +83,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.5.4
+rubygems_version: 3.5.5
 signing_key: 
 specification_version: 4
 summary: Boot large ruby/rails apps faster