bootsnap 1.5.1 → 1.7.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 9a15b298603bbdda820fa4aa3d37e32c72f181aa49aac240a02f076de2dd17eb
-  data.tar.gz: bbce00645395d42d30cb3c89b35dcb910003e3f9c3b65afd3a94b7e9019b868a
+  metadata.gz: b7dac6bb1ac2018e46d1cdac77a243cc1e2d46d8dc24deb0a9248320451549ba
+  data.tar.gz: 9cd95e8c52993daffd79f1d21961c5754f88642250c9552881f30f26f9dcc462
 SHA512:
-  metadata.gz: 982d29eb952ba2c053fd78d244493fc2c75d148f11bed24c67e657c0b59d1d6cb4cfc964583087dfababe0b3f977aa9d183c67c69949a31b1dca65c5d51886a4
-  data.tar.gz: 9e862ebb9a2ddb6cebdfec4835e78c9fcf14dd1451af73a6f04bf38edb362111f27cb70a04e381cb9550cdbdf4cf83d838aaf514126a41e1c9f1c1a12ab96417
+  metadata.gz: e6cee7f1ed6c30d819e576fa79d1e627585996d139dd4add684fc22d4bf720ae1c0f778bb1ee23c2a7819d788c7956b9632f03b5861bb7f6e507f3562687b70e
+  data.tar.gz: 505de0bcbfa90ab03fde256bea4dec4f5728482b61d452887da7a49c84894daff61c1aa0b16a2a114ddab142ae2a75d09f4447de80a6b33068ea0c15ec0fdd8f

data/CHANGELOG.md

@@ -1,3 +1,46 @@
+# Unreleased
+
+# 1.7.5
+
+* Handle a regression of Ruby 2.7.3 causing Bootsnap to call the deprecated `untaint` method. (#360)
+* Gracefully handle read-only file system as well as other errors preventing to persist the load path cache. (#358)
+
+# 1.7.4
+
+* Stop raising errors when encoutering various file system errors. The cache is now best effort,
+  if somehow it can't be saved, bootsnapp will gracefully fallback to the original operation (e.g. `Kernel.require`).
+  (#353, #177, #262)
+
+# 1.7.3
+
+* Disable YAML precompilation when encountering YAML tags. (#351)
+
+# 1.7.2
+
+* Fix compatibility with msgpack < 1. (#349)
+
+# 1.7.1
+
+* Warn Ruby 2.5 users if they turn ISeq caching on. (#327, #244)
+* Disable ISeq caching for the whole 2.5.x series again.
+* Better handle hashing of Ruby strings. (#318)
+
+# 1.7.0
+
+* Fix detection of YAML files in gems.
+* Adds an instrumentation API to monitor cache misses.
+* Allow to control the behavior of `require 'bootsnap/setup'` using environment variables.
+* Deprecate the `disable_trace` option.
+* Deprecate the `ActiveSupport::Dependencies` (AKA Classic autoloader) integration. (#344) 
+
+# 1.6.0
+
+* Fix a Ruby 2.7/3.0 issue with `YAML.load_file` keyword arguments. (#342)
+* `bootsnap precompile` CLI use multiple processes to complete faster. (#341)
+* `bootsnap precompile` CLI also precompile YAML files. (#340)
+* Changed the load path cache directory from `$BOOTSNAP_CACHE_DIR/bootsnap-load-path-cache` to `$BOOTSNAP_CACHE_DIR/bootsnap/load-path-cache` for ease of use. (#334)
+* Changed the compile cache directory from `$BOOTSNAP_CACHE_DIR/bootsnap-compile-cache` to `$BOOTSNAP_CACHE_DIR/bootsnap/compile-cache` for ease of use. (#334)
+
 # 1.5.1
 
 * Workaround a Ruby bug in InstructionSequence.compile_file. (#332)

data/README.md

@@ -1,6 +1,6 @@
 # Bootsnap [![Actions Status](https://github.com/Shopify/bootsnap/workflows/ci/badge.svg)](https://github.com/Shopify/bootsnap/actions)
 
-Bootsnap is a library that plugs into Ruby, with optional support for `ActiveSupport` and `YAML`,
+Bootsnap is a library that plugs into Ruby, with optional support for `YAML`,
 to optimize and cache expensive computations. See [How Does This Work](#how-does-this-work).
 
 #### Performance
@@ -11,7 +11,7 @@ to optimize and cache expensive computations. See [How Does This Work](#how-does
 - The core Shopify platform -- a rather large monolithic application -- boots about 75% faster,
   dropping from around 25s to 6.5s.
 * In Shopify core (a large app), about 25% of this gain can be attributed to `compile_cache_*`
-  features; 75% to path caching, and ~1% to `disable_trace`. This is fairly representative.
+  features; 75% to path caching. This is fairly representative.
 
 ## Usage
 
@@ -29,7 +29,8 @@ If you are using Rails, add this to `config/boot.rb` immediately after `require
 require 'bootsnap/setup'
 ```
 
-Note that bootsnap writes to `tmp/cache`, and that directory *must* be writable. Rails will fail to
+Note that bootsnap writes to `tmp/cache` (or the path specified by `ENV['BOOTSNAP_CACHE_DIR']`),
+and that directory *must* be writable. Rails will fail to
 boot if it is not. If this is unacceptable (e.g. you are running in a read-only container and
 unwilling to mount in a writable tmpdir), you should remove this line or wrap it in a conditional.
 
@@ -53,15 +54,11 @@ Bootsnap.setup(
   cache_dir:            'tmp/cache',          # Path to your cache
   development_mode:     env == 'development', # Current working environment, e.g. RACK_ENV, RAILS_ENV, etc
   load_path_cache:      true,                 # Optimize the LOAD_PATH with a cache
-  autoload_paths_cache: true,                 # Optimize ActiveSupport autoloads with cache
-  disable_trace:        true,                 # Set `RubyVM::InstructionSequence.compile_option = { trace_instruction: false }`
   compile_cache_iseq:   true,                 # Compile Ruby code into ISeq cache, breaks coverage reporting.
   compile_cache_yaml:   true                  # Compile YAML into a cache
 )
 ```
 
-**Note that `disable_trace` will break debuggers and tracing.**
-
 **Protip:** You can replace `require 'bootsnap'` with `BootLib::Require.from_gem('bootsnap',
 'bootsnap')` using [this trick](https://github.com/Shopify/bootsnap/wiki/Bootlib::Require). This
 will help optimize boot time further if you have an extremely large `$LOAD_PATH`.
@@ -71,12 +68,39 @@ speeds up the loading of individual source files, Spring keeps a copy of a pre-b
 on hand to completely skip parts of the boot process the next time it's needed. The two tools work
 well together, and are both included in a newly-generated Rails applications by default.
 
+### Environment variables
+
+`require 'bootsnap/setup'` behavior can be changed using environment variables:
+
+- `BOOTSNAP_CACHE_DIR` allows to define the cache location.
+- `DISABLE_BOOTSNAP` allows to entirely disable bootsnap.
+- `DISABLE_BOOTSNAP_LOAD_PATH_CACHE` allows to disable load path caching.
+- `DISABLE_BOOTSNAP_COMPILE_CACHE` allows to disable ISeq and YAML caches.
+- `BOOTSNAP_LOG` configure bootsnap to log all caches misses to STDERR.
+
 ### Environments
 
 All Bootsnap features are enabled in development, test, production, and all other environments according to the configuration in the setup. At Shopify, we use this gem safely in all environments without issue.
 
 If you would like to disable any feature for a certain environment, we suggest changing the configuration to take into account the appropriate ENV var or configuration according to your needs.
 
+### Instrumentation
+
+Bootsnap cache misses can be monitored though a callback:
+
+```ruby
+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.
+
+To turn instrumentation back off you can set it to nil:
+
+```ruby
+Bootsnap.instrumentation = nil
+```
+
 ## How does this work?
 
 Bootsnap optimizes methods to cache results of expensive computations, and can be grouped
@@ -84,8 +108,6 @@ into two broad categories:
 
 * [Path Pre-Scanning](#path-pre-scanning)
     * `Kernel#require` and `Kernel#load` are modified to eliminate `$LOAD_PATH` scans.
-    * `ActiveSupport::Dependencies.{autoloadable_module?,load_missing_constant,depend_on}` are
-      overridden to eliminate scans of `ActiveSupport::Dependencies.autoload_paths`.
 * [Compilation caching](#compilation-caching)
     * `RubyVM::InstructionSequence.load_iseq` is implemented to cache the result of ruby bytecode
       compilation.
@@ -124,10 +146,6 @@ open y/foo.rb
 ...
 ```
 
-Exactly the same strategy is employed for methods that traverse
-`ActiveSupport::Dependencies.autoload_paths` if the `autoload_paths_cache` option is given to
-`Bootsnap.setup`.
-
 The following diagram flowcharts the overrides that make the `*_path_cache` features work.
 
 ![Flowchart explaining

data/ext/bootsnap/bootsnap.c

@@ -14,6 +14,7 @@
 #include "bootsnap.h"
 #include "ruby.h"
 #include <stdint.h>
+#include <stdbool.h>
 #include <sys/types.h>
 #include <errno.h>
 #include <fcntl.h>
@@ -34,6 +35,10 @@
 
 #define MAX_CREATE_TEMPFILE_ATTEMPT 3
 
+#ifndef RB_UNLIKELY
+  #define RB_UNLIKELY(x) (x)
+#endif
+
 /*
  * An instance of this key is written as the first 64 bytes of each cache file.
  * The mtime and size members track whether the file contents have changed, and
@@ -70,7 +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 = 2;
+static const uint32_t current_version = 3;
 
 /* hash of e.g. "x86_64-darwin17", invalidating when ruby is recompiled on a
  * new OS ABI, etc. */
@@ -88,17 +93,23 @@ static VALUE rb_mBootsnap_CompileCache;
 static VALUE rb_mBootsnap_CompileCache_Native;
 static VALUE rb_eBootsnap_CompileCache_Uncompilable;
 static ID uncompilable;
+static ID instrumentation_method;
+static VALUE sym_miss;
+static VALUE sym_stale;
+static bool instrumentation_enabled = false;
 
 /* Functions exposed as module functions on Bootsnap::CompileCache::Native */
+static VALUE bs_instrumentation_enabled_set(VALUE self, VALUE enabled);
 static VALUE bs_compile_option_crc32_set(VALUE self, VALUE crc32_v);
 static VALUE bs_rb_fetch(VALUE self, VALUE cachedir_v, VALUE path_v, VALUE handler, VALUE args);
+static VALUE bs_rb_precompile(VALUE self, VALUE cachedir_v, VALUE path_v, VALUE handler);
 
 /* Helpers */
-static uint64_t fnv1a_64(const char *str);
-static void bs_cache_path(const char * cachedir, const char * path, const char * extra, char (* cache_path)[MAX_CACHEPATH_SIZE]);
+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 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);
 static int fetch_cached_data(int fd, ssize_t data_size, VALUE handler, VALUE args, VALUE * output_data, int * exception_tag, const char ** errno_provenance);
 static uint32_t get_ruby_revision(void);
@@ -146,15 +157,31 @@ Init_bootsnap(void)
   current_ruby_platform = get_ruby_platform();
 
   uncompilable = rb_intern("__bootsnap_uncompilable__");
+  instrumentation_method = rb_intern("_instrument");
+
+  sym_miss = ID2SYM(rb_intern("miss"));
+  rb_global_variable(&sym_miss);
+
+  sym_stale = ID2SYM(rb_intern("stale"));
+  rb_global_variable(&sym_stale);
 
+  rb_define_module_function(rb_mBootsnap, "instrumentation_enabled=", bs_instrumentation_enabled_set, 1);
   rb_define_module_function(rb_mBootsnap_CompileCache_Native, "coverage_running?", bs_rb_coverage_running, 0);
   rb_define_module_function(rb_mBootsnap_CompileCache_Native, "fetch", bs_rb_fetch, 4);
+  rb_define_module_function(rb_mBootsnap_CompileCache_Native, "precompile", bs_rb_precompile, 3);
   rb_define_module_function(rb_mBootsnap_CompileCache_Native, "compile_option_crc32=", bs_compile_option_crc32_set, 1);
 
   current_umask = umask(0777);
   umask(current_umask);
 }
 
+static VALUE
+bs_instrumentation_enabled_set(VALUE self, VALUE enabled)
+{
+  instrumentation_enabled = RTEST(enabled);
+  return enabled;
+}
+
 /*
  * Bootsnap's ruby code registers a hook that notifies us via this function
  * when compile_option changes. These changes invalidate all existing caches.
@@ -183,7 +210,7 @@ bs_compile_option_crc32_set(VALUE self, VALUE crc32_v)
  *   - 32 bits doesn't feel collision-resistant enough; 64 is nice.
  */
 static uint64_t
-fnv1a_64_iter(uint64_t h, const char *str)
+fnv1a_64_iter_cstr(uint64_t h, const char *str)
 {
   unsigned char *s = (unsigned char *)str;
 
@@ -196,7 +223,21 @@ fnv1a_64_iter(uint64_t h, const char *str)
 }
 
 static uint64_t
-fnv1a_64(const char *str)
+fnv1a_64_iter(uint64_t h, const VALUE str)
+{
+  unsigned char *s = (unsigned char *)RSTRING_PTR(str);
+  unsigned char *str_end = (unsigned char *)RSTRING_PTR(str) + RSTRING_LEN(str);
+
+  while (s < str_end) {
+    h ^= (uint64_t)*s++;
+    h += (h << 1) + (h << 4) + (h << 5) + (h << 7) + (h << 8) + (h << 40);
+  }
+
+  return h;
+}
+
+static uint64_t
+fnv1a_64(const VALUE str)
 {
   uint64_t h = (uint64_t)0xcbf29ce484222325ULL;
   return fnv1a_64_iter(h, str);
@@ -217,7 +258,7 @@ get_ruby_revision(void)
   } else {
     uint64_t hash;
 
-    hash = fnv1a_64(StringValueCStr(ruby_revision));
+    hash = fnv1a_64(ruby_revision);
     return (uint32_t)(hash >> 32);
   }
 }
@@ -237,19 +278,19 @@ get_ruby_platform(void)
   VALUE ruby_platform;
 
   ruby_platform = rb_const_get(rb_cObject, rb_intern("RUBY_PLATFORM"));
-  hash = fnv1a_64(RSTRING_PTR(ruby_platform));
+  hash = fnv1a_64(ruby_platform);
 
 #ifdef _WIN32
   return (uint32_t)(hash >> 32) ^ (uint32_t)GetVersion();
 #elif defined(__GLIBC__)
-  hash = fnv1a_64_iter(hash, gnu_get_libc_version());
+  hash = fnv1a_64_iter_cstr(hash, gnu_get_libc_version());
   return (uint32_t)(hash >> 32);
 #else
   struct utsname utsname;
 
   /* Not worth crashing if this fails; lose extra cache invalidation potential */
   if (uname(&utsname) >= 0) {
-    hash = fnv1a_64_iter(hash, utsname.version);
+    hash = fnv1a_64_iter_cstr(hash, utsname.version);
   }
 
   return (uint32_t)(hash >> 32);
@@ -264,17 +305,13 @@ get_ruby_platform(void)
  * The path will look something like: <cachedir>/12/34567890abcdef
  */
 static void
-bs_cache_path(const char * cachedir, const char * path, const char * extra, char (* cache_path)[MAX_CACHEPATH_SIZE])
+bs_cache_path(const char * cachedir, const VALUE path, char (* cache_path)[MAX_CACHEPATH_SIZE])
 {
   uint64_t hash = fnv1a_64(path);
-  if (extra) {
-    hash ^= fnv1a_64(extra);
-  }
-
   uint8_t first_byte = (hash >> (64 - 8));
   uint64_t remainder = hash & 0x00ffffffffffffff;
 
-  sprintf(*cache_path, "%s/%02x/%014llx", cachedir, first_byte, remainder);
+  sprintf(*cache_path, "%s/%02"PRIx8"/%014"PRIx64, cachedir, first_byte, remainder);
 }
 
 /*
@@ -318,18 +355,39 @@ bs_rb_fetch(VALUE self, VALUE cachedir_v, VALUE path_v, VALUE handler, VALUE arg
   char * cachedir = RSTRING_PTR(cachedir_v);
   char * path     = RSTRING_PTR(path_v);
   char cache_path[MAX_CACHEPATH_SIZE];
-  char * extra    = NULL;
-  if (!NIL_P(args)) {
-    VALUE args_serial = rb_marshal_dump(args, Qnil);
-    extra = RSTRING_PTR(args_serial);
-  }
 
   /* generate cache path to cache_path */
-  bs_cache_path(cachedir, path, extra, &cache_path);
+  bs_cache_path(cachedir, path_v, &cache_path);
 
   return bs_fetch(path, path_v, cache_path, handler, args);
 }
 
+/*
+ * Entrypoint for Bootsnap::CompileCache::Native.precompile.
+ * Similar to fetch, but it only generate the cache if missing
+ * and doesn't return the content.
+ */
+static VALUE
+bs_rb_precompile(VALUE self, VALUE cachedir_v, VALUE path_v, VALUE handler)
+{
+  FilePathValue(path_v);
+
+  Check_Type(cachedir_v, T_STRING);
+  Check_Type(path_v, T_STRING);
+
+  if (RSTRING_LEN(cachedir_v) > MAX_CACHEDIR_SIZE) {
+    rb_raise(rb_eArgError, "cachedir too long");
+  }
+
+  char * cachedir = RSTRING_PTR(cachedir_v);
+  char * path     = RSTRING_PTR(path_v);
+  char cache_path[MAX_CACHEPATH_SIZE];
+
+  /* generate cache path to cache_path */
+  bs_cache_path(cachedir, path_v, &cache_path);
+
+  return bs_precompile(path, path_v, cache_path, handler);
+}
 /*
  * Open the file we want to load/cache and generate a cache key for it if it
  * was loaded.
@@ -366,7 +424,8 @@ open_current_file(char * path, struct bs_cache_key * key, const char ** errno_pr
 }
 
 #define ERROR_WITH_ERRNO -1
-#define CACHE_MISSING_OR_INVALID -2
+#define CACHE_MISS -2
+#define CACHE_STALE -3
 
 /*
  * Read the cache key from the given fd, which must have position 0 (e.g.
@@ -374,15 +433,16 @@ open_current_file(char * path, struct bs_cache_key * key, const char ** errno_pr
  *
  * Possible return values:
  *   - 0 (OK, key was loaded)
- *   - CACHE_MISSING_OR_INVALID (-2)
  *   - ERROR_WITH_ERRNO (-1, errno is set)
+ *   - CACHE_MISS (-2)
+ *   - CACHE_STALE (-3)
  */
 static int
 bs_read_key(int fd, struct bs_cache_key * key)
 {
   ssize_t nread = read(fd, key, KEY_SIZE);
   if (nread < 0)        return ERROR_WITH_ERRNO;
-  if (nread < KEY_SIZE) return CACHE_MISSING_OR_INVALID;
+  if (nread < KEY_SIZE) return CACHE_STALE;
   return 0;
 }
 
@@ -392,7 +452,8 @@ bs_read_key(int fd, struct bs_cache_key * key)
  *
  * Possible return values:
  *   - 0 (OK, key was loaded)
- *   - CACHE_MISSING_OR_INVALID (-2)
+ *   - CACHE_MISS (-2)
+ *   - CACHE_STALE (-3)
  *   - ERROR_WITH_ERRNO (-1, errno is set)
  */
 static int
@@ -403,8 +464,7 @@ open_cache_file(const char * path, struct bs_cache_key * key, const char ** errn
   fd = open(path, O_RDONLY);
   if (fd < 0) {
     *errno_provenance = "bs_fetch:open_cache_file:open";
-    if (errno == ENOENT) return CACHE_MISSING_OR_INVALID;
-    return ERROR_WITH_ERRNO;
+    return CACHE_MISS;
   }
   #ifdef _WIN32
   setmode(fd, O_BINARY);
@@ -458,7 +518,7 @@ fetch_cached_data(int fd, ssize_t data_size, VALUE handler, VALUE args, VALUE *
     goto done;
   }
   if (nread != data_size) {
-    ret = CACHE_MISSING_OR_INVALID;
+    ret = CACHE_STALE;
     goto done;
   }
 
@@ -652,14 +712,22 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
 
   /* Open the cache key if it exists, and read its cache key in */
   cache_fd = open_cache_file(cache_path, &cached_key, &errno_provenance);
-  if (cache_fd == CACHE_MISSING_OR_INVALID) {
+  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);
+    }
   } else if (cache_fd < 0) {
     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);
+      }
+    }
   }
 
   if (valid_cache) {
@@ -668,10 +736,10 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
       cache_fd, (ssize_t)cached_key.data_size, handler, args,
       &output_data, &exception_tag, &errno_provenance
     );
-    if (exception_tag != 0)                   goto raise;
-    else if (res == CACHE_MISSING_OR_INVALID) valid_cache = 0;
-    else if (res == ERROR_WITH_ERRNO)         goto fail_errno;
-    else if (!NIL_P(output_data))             goto succeed; /* fast-path, goal */
+    if (exception_tag != 0) goto raise;
+    else if (res == CACHE_MISS || res == CACHE_STALE) valid_cache = 0;
+    else if (res == ERROR_WITH_ERRNO) goto fail_errno;
+    else if (!NIL_P(output_data)) goto succeed; /* fast-path, goal */
   }
   close(cache_fd);
   cache_fd = -1;
@@ -694,9 +762,11 @@ bs_fetch(char * path, VALUE path_v, char * cache_path, VALUE handler, VALUE args
   /* If storage_data isn't a string, we can't cache it */
   if (!RB_TYPE_P(storage_data, T_STRING)) goto invalid_type_storage_data;
 
-  /* Write the cache key and storage_data to the cache directory */
-  res = atomic_write_cache_file(cache_path, &current_key, storage_data, &errno_provenance);
-  if (res < 0) goto fail_errno;
+  /* Attempt to write the cache key and storage_data to the cache directory.
+   * We do however ignore any failures to persist the cache, as it's better
+   * to move along, than to interrupt the process.
+   */
+  atomic_write_cache_file(cache_path, &current_key, storage_data, &errno_provenance);
 
   /* Having written the cache, now convert storage_data to output_data */
   exception_tag = bs_storage_to_output(handler, args, storage_data, &output_data);
@@ -740,6 +810,79 @@ invalid_type_storage_data:
 #undef CLEANUP
 }
 
+static VALUE
+bs_precompile(char * path, VALUE path_v, char * cache_path, VALUE handler)
+{
+  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 storage_data; /* compiled data, e.g. msgpack / binary iseq */
+
+  /* Open the source file and generate a cache key for it */
+  current_fd = open_current_file(path, &current_key, &errno_provenance);
+  if (current_fd < 0) goto fail;
+
+  /* Open the cache key if it exists, and read its cache key in */
+  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. */
+  } else if (cache_fd < 0) {
+    goto fail;
+  } 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 (valid_cache) {
+    goto succeed;
+  }
+
+  close(cache_fd);
+  cache_fd = -1;
+  /* Cache is stale, invalid, or missing. Regenerate and write it out. */
+
+  /* Read the contents of the source file into a buffer */
+  if (bs_read_contents(current_fd, current_key.size, &contents, &errno_provenance) < 0) goto fail;
+  input_data = rb_str_new(contents, current_key.size);
+
+  /* Try to compile the input_data using input_to_storage(input_data) */
+  exception_tag = bs_input_to_storage(handler, Qnil, input_data, path_v, &storage_data);
+  if (exception_tag != 0) goto fail;
+
+  /* If input_to_storage raised Bootsnap::CompileCache::Uncompilable, don't try
+   * to cache anything; just return false */
+  if (storage_data == uncompilable) {
+    goto fail;
+  }
+  /* If storage_data isn't a string, we can't cache it */
+  if (!RB_TYPE_P(storage_data, T_STRING)) goto fail;
+
+  /* Write the cache key and storage_data to the cache directory */
+  res = atomic_write_cache_file(cache_path, &current_key, storage_data, &errno_provenance);
+  if (res < 0) goto fail;
+
+  goto succeed;
+
+#define CLEANUP \
+  if (contents != NULL) xfree(contents);   \
+  if (current_fd >= 0)  close(current_fd); \
+  if (cache_fd >= 0)    close(cache_fd);
+
+succeed:
+  CLEANUP;
+  return Qtrue;
+fail:
+  CLEANUP;
+  return Qfalse;
+#undef CLEANUP
+}
+
+
 /*****************************************************************************/
 /********************* Handler Wrappers **************************************/
 /*****************************************************************************
@@ -771,7 +914,6 @@ struct i2o_data {
 
 struct i2s_data {
   VALUE handler;
-  VALUE args;
   VALUE input_data;
   VALUE pathval;
 };
@@ -789,7 +931,7 @@ bs_storage_to_output(VALUE handler, VALUE args, VALUE storage_data, VALUE * outp
   int state;
   struct s2o_data s2o_data = {
     .handler      = handler,
-    .args       = args,
+    .args         = args,
     .storage_data = storage_data,
   };
   *output_data = rb_protect(prot_storage_to_output, (VALUE)&s2o_data, &state);
@@ -818,7 +960,7 @@ static VALUE
 try_input_to_storage(VALUE arg)
 {
   struct i2s_data * data = (struct i2s_data *)arg;
-  return rb_funcall(data->handler, rb_intern("input_to_storage"), 3, data->input_data, data->pathval, data->args);
+  return rb_funcall(data->handler, rb_intern("input_to_storage"), 2, data->input_data, data->pathval);
 }
 
 static VALUE
@@ -843,7 +985,6 @@ bs_input_to_storage(VALUE handler, VALUE args, VALUE input_data, VALUE pathval,
   int state;
   struct i2s_data i2s_data = {
     .handler    = handler,
-    .args       = args,
     .input_data = input_data,
     .pathval    = pathval,
   };