Use Base16 hash for script path.

Issue #9334 shows that `%` characters on Windows result in invalid
paths, also `/` characters on Linux create invalid paths.

This changes from using base64 to using base16 with the same length
we use for unit-ids.

cabal-install/cabal-install.cabal

@@ -209,7 +209,6 @@ library
         async      >= 2.0      && < 2.3,
         array      >= 0.4      && < 0.6,
         base16-bytestring >= 0.1.1 && < 1.1.0.0,
-        base64-bytestring >= 1.0 && < 1.3,
         binary     >= 0.7.3    && < 0.9,
         bytestring >= 0.10.6.0 && < 0.13,
         containers >= 0.5.6.2  && < 0.8,

cabal-install/src/Distribution/Client/HashValue.hs

@@ -7,7 +7,6 @@ module Distribution.Client.HashValue
   , hashValue
   , truncateHash
   , showHashValue
-  , showHashValueBase64
   , readFileHashValue
   , hashFromTUF
   ) where
@@ -19,7 +18,6 @@ import qualified Hackage.Security.Client as Sec
 
 import qualified Crypto.Hash.SHA256 as SHA256
 import qualified Data.ByteString.Base16 as Base16
-import qualified Data.ByteString.Base64 as Base64
 import qualified Data.ByteString.Char8 as BS
 import qualified Data.ByteString.Lazy.Char8 as LBS
 
@@ -57,9 +55,6 @@ hashValue = HashValue . SHA256.hashlazy
 showHashValue :: HashValue -> String
 showHashValue (HashValue digest) = BS.unpack (Base16.encode digest)
 
-showHashValueBase64 :: HashValue -> String
-showHashValueBase64 (HashValue digest) = BS.unpack (Base64.encode digest)
-
 -- | Hash the content of a file. Uses SHA256.
 readFileHashValue :: FilePath -> IO HashValue
 readFileHashValue tarball =

cabal-install/src/Distribution/Client/ScriptUtils.hs

@@ -37,7 +37,8 @@ import Distribution.Client.DistDirLayout
   )
 import Distribution.Client.HashValue
   ( hashValue
-  , showHashValueBase64
+  , showHashValue
+  , truncateHash
   )
 import Distribution.Client.HttpUtils
   ( HttpTransport
@@ -218,33 +219,20 @@ import qualified Text.Parsec as P
 --    repl to deal with the fact that the repl is relative to the working directory and not
 --    the project root.
 
--- | Get the hash of a script's absolute path)
---
--- Two hashes will be the same as long as the absolute paths
--- are the same.
+-- | Get the hash of a script's contents.
 getScriptHash :: FilePath -> IO String
 getScriptHash script =
-  -- Base64 is shorter than Base16, which helps avoid long path issues on windows
-  -- but it can contain /'s which aren't valid in file paths so replace them with
-  -- %'s. 26 chars / 130 bits is enough to practically avoid collisions.
-  map (\c -> if c == '/' then '%' else c)
-    . take 26
-    . showHashValueBase64
+  showHashValue
+    . truncateHash 26
     . hashValue
     . fromString
     <$> canonicalizePath script
 
 -- | Get the directory for caching a script build.
---
--- The only identity of a script is it's absolute path, so append the
--- hashed path to the @script-builds@ dir to get the cache directory.
 getScriptCacheDirectory :: FilePath -> IO FilePath
 getScriptCacheDirectory script = (</>) <$> defaultScriptBuildsDir <*> getScriptHash script
 
 -- | Get the directory for caching a script build and ensure it exists.
---
--- The only identity of a script is it's absolute path, so append the
--- hashed path to the @script-builds@ dir to get the cache directory.
 ensureScriptCacheDirectory :: Verbosity -> FilePath -> IO FilePath
 ensureScriptCacheDirectory verbosity script = do
   cacheDir <- getScriptCacheDirectory script

cabal-testsuite/cabal-testsuite.cabal

@@ -60,7 +60,7 @@ library
     , aeson                 ^>= 1.4.2.0 || ^>=1.5.0.0 || ^>= 2.0.0.0 || ^>= 2.1.0.0 || ^>= 2.2.1.0
     , async                 ^>= 2.2.1
     , attoparsec            ^>= 0.13.2.2 || ^>=0.14.1
-    , base64-bytestring     ^>= 1.0.0.0 || ^>= 1.1.0.0 || ^>= 1.2.0.0
+    , base16-bytestring     ^>= 0.1.1.5 || ^>= 1.0
     , bytestring            ^>= 0.10.0.2 || ^>= 0.11.0.0 || ^>= 0.12.0.0
     , containers            ^>= 0.5.0.0 || ^>= 0.6.0.1
     , cryptohash-sha256     ^>= 0.11.101.0
@@ -120,6 +120,7 @@ executable setup
 
 -- If you require an external dependency for a test it must be listed here.
 executable test-runtime-deps
+  default-language: Haskell2010
   build-depends: cabal-testsuite,
                  base,
                  directory,

cabal-testsuite/src/Test/Cabal/Prelude.hs

@@ -53,7 +53,7 @@ import Control.Monad (unless, when, void, forM_, liftM2, liftM4)
 import Control.Monad.Trans.Reader (withReaderT, runReaderT)
 import Control.Monad.IO.Class (MonadIO (..))
 import qualified Crypto.Hash.SHA256 as SHA256
-import qualified Data.ByteString.Base64 as Base64
+import qualified Data.ByteString.Base16 as Base16
 import qualified Data.ByteString.Char8 as C
 import Data.List (isInfixOf, stripPrefix, isPrefixOf, intercalate)
 import Data.List.NonEmpty (NonEmpty (..))
@@ -840,8 +840,7 @@ getScriptCacheDirectory :: FilePath -> TestM FilePath
 getScriptCacheDirectory script = do
     cabalDir <- testCabalDir `fmap` getTestEnv
     hashinput <- liftIO $ canonicalizePath script
-    let hash = map (\c -> if c == '/' then '%' else c) . take 26
-             . C.unpack . Base64.encode . SHA256.hash . C.pack $ hashinput
+    let hash = C.unpack . Base16.encode . C.take 26 . SHA256.hash . C.pack $ hashinput
     return $ cabalDir </> "script-builds" </> hash
 
 ------------------------------------------------------------------------

changelog.d/base16-script-cache

@@ -0,0 +1,9 @@
+synopsis: Script cache dir is the base16 hash of the canonical path of the script.
+prs: #9459
+packages: cabal-install
+
+description: {
+
+Script cache dir is the base16 hash of the canonical path of the script.
+
+}