From 8d1ff6eb864b0e9a2cc0a2c8e0602e4e5d42c3a2 Mon Sep 17 00:00:00 2001
From: James Brechtel <git@jamesbrechtel.com>
Date: Wed, 20 May 2026 14:53:11 -0400
Subject: [PATCH] fix: read subprocess output as raw bytes to avoid locale
 encoding errors

readProcessWithExitCode returns stdout/stderr as String, which forces GHC
to decode raw bytes using the current locale encoding. In a Docker
container with C locale (default), non-ASCII UTF-8 bytes (like 0xE2)
cause 'invalid argument' errors.

Replace with readProcessBytes which uses createProcess + hSetBinaryMode
to capture stdout/stderr as raw lazy ByteStrings, bypassing locale
decoding entirely. JSON is decoded directly from raw bytes via A.decode.
---
 src/Roux/Server.hs | 35 ++++++++++++++++++++++++-----------
 1 file changed, 24 insertions(+), 11 deletions(-)

diff --git a/src/Roux/Server.hs b/src/Roux/Server.hs
index 8ba1412..7d20e83 100644
--- a/src/Roux/Server.hs
+++ b/src/Roux/Server.hs
@@ -50,7 +50,8 @@ import qualified Roux.Html as Html
 import qualified Roux.RecipeIndex as Idx
 import Roux.SchemaOrg (SchemaOrgRecipe (..), parseSchemaOrgRecipe, schemaOrgToCooklang)
 import System.Exit (ExitCode (..))
-import System.Process (readProcessWithExitCode)
+import System.IO (hSetBinaryMode)
+import System.Process (CreateProcess (..), StdStream (CreatePipe), createProcess, proc, waitForProcess)
 
 -- | Read the full request body as a lazy ByteString.
 readRequestBody :: Wai.Request -> IO LB.ByteString
@@ -234,6 +235,22 @@ parseFormBody body =
          in toEnum hexVal : go rest
     go (c : rest) = c : go rest
 
+-- | Run a subprocess and capture stdout/stderr as raw bytes (no locale decoding).
+readProcessBytes :: FilePath -> [String] -> IO (ExitCode, LB.ByteString, LB.ByteString)
+readProcessBytes cmd args = do
+    let process =
+            (proc cmd args)
+                { std_out = CreatePipe
+                , std_err = CreatePipe
+                }
+    (Just outH, Just errH, _, ph) <- createProcess process
+    hSetBinaryMode outH True
+    hSetBinaryMode errH True
+    out <- LB.hGetContents outH
+    err <- LB.hGetContents errH
+    ec <- waitForProcess ph
+    return (ec, out, err)
+
 -- | Locate the scrape-recipe script: check PATH first, then relative to project.
 findScrapeScript :: IO FilePath
 findScrapeScript = do
@@ -250,23 +267,19 @@ runImportPipeline recipeDir url = do
     -- Step 1: locate and run the scraper
     script <- findScrapeScript
     putStrLn $ "[roux] importing from " <> T.unpack url <> " using script " <> script
-    (exitCode, stdout, stderr) <-
-        readProcessWithExitCode
-            script
-            ["--pretty", T.unpack url]
-            ""
+    (exitCode, stdoutBytes, stderrBytes) <-
+        readProcessBytes script ["--pretty", T.unpack url]
     case exitCode of
         ExitFailure _ -> do
             putStrLn $ "[roux] scrape-recipe exited with code " <> show exitCode <> " for " <> T.unpack url
-            putStrLn $ "[roux] scrape-recipe stderr: " <> stderr
+            putStrLn $ "[roux] scrape-recipe stderr: " <> T.unpack (TE.decodeUtf8 (LB.toStrict stderrBytes))
             return $
                 Left $
                     Html.ImportError $
-                        "Could not scrape URL (" <> url <> "). The scraper exited with: " <> T.pack stderr
+                        "Could not scrape URL (" <> url <> "). The scraper exited with: " <> TE.decodeUtf8 (LB.toStrict stderrBytes)
         ExitSuccess -> do
-            -- Step 2: decode stdout (String) as JSON Value
-            let jsonBytes = LB.fromStrict (TE.encodeUtf8 (T.pack stdout))
-                jsonVal = A.decode jsonBytes :: Maybe A.Value
+            -- Step 2: decode stdout (raw bytes) as JSON Value
+            let jsonVal = A.decode stdoutBytes :: Maybe A.Value
             case jsonVal of
                 Nothing -> do
                     putStrLn $ "[roux] failed to decode JSON from scrape-recipe stdout for " <> T.unpack url