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.

(cherry picked from commit 97f9917)

Co-authored-by: Hécate Moonlight <[email protected]>
Co-autohred-by: Javier Sagredo <[email protected]>

Author: jasagredo

cabal-install/cabal-install.cabal

@@ -207,7 +207,6 @@ 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  && < 0.7,

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

@@ -6,7 +6,6 @@ module Distribution.Client.HashValue (
     hashValue,
     truncateHash,
     showHashValue,
-    showHashValueBase64,
     readFileHashValue,
     hashFromTUF,
     ) where
@@ -16,10 +15,9 @@ import Prelude ()
 
 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 Crypto.Hash.SHA256 as SHA256
+import qualified Data.ByteString.Base16 as Base16
+import qualified Data.ByteString.Char8 as BS
 import qualified Data.ByteString.Lazy.Char8 as LBS
 
 import System.IO         (IOMode (..), withBinaryFile)
@@ -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

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

@@ -27,7 +27,7 @@ import Distribution.Client.Config
 import Distribution.Client.DistDirLayout
     ( DistDirLayout(..), DistDirParams(..) )
 import Distribution.Client.HashValue
-    ( hashValue, showHashValueBase64 )
+    ( hashValue, showHashValue, truncateHash )
 import Distribution.Client.HttpUtils
          ( HttpTransport, configureTransport )
 import Distribution.Client.NixStyleOptions
@@ -131,17 +131,18 @@ 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)
+-- | Get the hash of a script's absolute path.
 --
 -- Two hashes will be the same as long as the absolute paths
 -- are the same.
 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 . hashValue . fromString <$> canonicalizePath script
+getScriptHash script =
+  -- Truncation here tries to help with long path issues on Windows.
+  showHashValue
+    . truncateHash 26
+    . hashValue
+    . fromString
+    <$> canonicalizePath script
 
 -- | Get the directory for caching a script build.
 --

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

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

@@ -52,7 +52,7 @@ 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 (..))
@@ -842,8 +842,7 @@ 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.
+
+}