Use actual file blocks rather than apparent size; doc updates

Makefile

@@ -19,7 +19,7 @@ ctrl: test-backup
 .PHONY: test-backup
 test-backup: .venv
 	.venv/bin/mypy backup.py
-	./backup.py -n >/dev/null
+	./backup.py -n
 
 .PHONY: test-setup
 test-setup:

backup.py

@@ -18,7 +18,7 @@ import wcmatch.glob # type: ignore
 import humanfriendly # type: ignore
 
 class Config:
-    root: str
+    root: bytes
     max_file_size: typing.Optional[int]
     one_file_system: bool
     exclude: list[bytes]
@@ -91,9 +91,8 @@ class Config:
         return yaml.dump(d, default_flow_style=False)
 
 class Backup:
-    def __init__(self, config: Config, dry_run: bool, out: typing.BinaryIO):
+    def __init__(self, config: Config, dry_run: bool):
         self.config = config
-        self.outfile = out
         self.dry_run = dry_run
 
         # All logged messages, with severity
@@ -111,7 +110,8 @@ class Backup:
         sys.stderr.write(f"\033[1;{c}m{letter}:\033[22m {msg}\033[0m\n")
         self.logs.append((letter, msg))
 
-    def run(self):
+    def run(self, outfile: typing.BinaryIO):
+        self.outfile = outfile
         self.scan(self.config.root)
 
     def scan(self, path: bytes, parent_st: os.stat_result=None):
@@ -140,16 +140,16 @@ class Backup:
                 # Config file says to exclude
                 exclude_reason = ('I', f"skipping, excluded by config file")
 
-            elif (stat.S_ISDIR(st().st_mode)
-                  and self.config.one_file_system
+            elif (self.config.one_file_system
                   and parent_st is not None
+                  and stat.S_ISDIR(st().st_mode)
                   and st().st_dev != parent_st.st_dev):
                 # Crosses a mount point
                 exclude_reason = ('I', "skipping, on different filesystem")
 
-            elif (stat.S_ISREG(st().st_mode)
-                  and self.config.max_file_size
-                  and st().st_size > self.config.max_file_size):
+            elif (self.config.max_file_size
+                  and stat.S_ISREG(st().st_mode)
+                  and (st().st_blocks * 512) > self.config.max_file_size):
                 # Too big
                 def format_size(n):
                     return humanfriendly.format_size(
@@ -199,8 +199,13 @@ def main(argv: list[str]):
 
     args = parser.parse_args()
     config = Config(args.config)
-    backup = Backup(config, args.dry_run, sys.stdout.buffer)
-    backup.run()
+
+    backup = Backup(config, args.dry_run)
+    if args.dry_run:
+        with open(os.devnull, "wb") as out:
+            backup.run(out)
+    else:
+        backup.run(sys.stdout.buffer)
 
 if __name__ == "__main__":
     import sys

config.yaml

@@ -1,18 +1,20 @@
-root: "/tmp"
+root: "/"
 one-file-system: true
 
 # Files larger than this are excluded.  If a large file isn't
 # explicitly mentioned in "excludes" below, it also generates a
-# warning.
+# warning.  Note that this counts used blocks, so files with large
+# holes will still be considered small (since they'll compress easily)
 max-file-size: 500MiB
 
 # Files/dirs to exclude from backup.
 # Paths should be absolute, or start with **/
 exclude: |
+  /var/tmp
+  /tmp
+  /var/cache/apt/archives
   **/Steam/steamapps
   **/Steam/ubuntu*
-  /tmp/bigfile
-  /tmp/out.ps
 
 # Files that are always included, even if they would have been
 # excluded due to file size or the "exclude" list.

initial-setup.sh

@@ -324,6 +324,10 @@ notice "     Notes: (paste the following key)"
 sed -ne '/BORG/,/^$/{/./p}' "${BORG_DIR}/key.txt"
 notice ""
 notice ""
+notice "Then, test the backup file list with"
+notice "    sudo ${BORG_DIR}/backup.py --dry-run"
+notice "and make any necessary adjustments to:"
+notice "    ${BORG_DIR}/config.yaml"
 
 echo