From 60fe70ee5ad85eccfcdb6d33511e4a95f50e3e33 Mon Sep 17 00:00:00 2001
From: Fitz Elliott <fitz@cos.io>
Date: Mon, 17 Aug 2026 10:53:04 -0400
Subject: [PATCH] slurp small downloads from oraclecloud

 * WB has a weird bug (also encountered with googledrive) where trying
   to download small files indirectly will hang after reading one
   chunk. "Indirectly" in this case means from MFR, which uses WB as a
   library to download cached file conversions. This commit applies
   the same fix to oraclecloud as we used for googlecloud. If the
   download response has a content-length header AND the header is
   less than a configurable number of bytes (default: 100k), then we
   slurp the file contents into memory and wrap it in a StringStream
   object.

   No idea why this is happening, but the divine Cargomides seems to
   appreciate it.
---
 tests/providers/oraclecloud/test_provider.py  | 9 +++++++--
 waterbutler/providers/oraclecloud/provider.py | 9 ++++++++-
 waterbutler/providers/oraclecloud/settings.py | 3 +++
 3 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/tests/providers/oraclecloud/test_provider.py b/tests/providers/oraclecloud/test_provider.py
index 3114bf662..e8c32d47c 100644
--- a/tests/providers/oraclecloud/test_provider.py
+++ b/tests/providers/oraclecloud/test_provider.py
@@ -228,7 +228,9 @@ async def test_download_file(self, mock_provider, file_wb_path, file_content):
         ):
             stream = await mock_provider.download(file_wb_path)
 
-        assert isinstance(stream, ResponseStreamReader)
+        # workaround: for small files, we slurp into memory and return a StringStream
+        # assert isinstance(stream, ResponseStreamReader)
+        assert isinstance(stream, StringStream)
 
     @pytest.mark.asyncio
     async def test_download_folder_raises(self, mock_provider, folder_wb_path):
@@ -268,7 +270,10 @@ async def test_download_with_range(self, mock_provider, file_wb_path):
         ) as mocked:
             stream = await mock_provider.download(file_wb_path, range=(0, 6))
 
-        assert isinstance(stream, ResponseStreamReader)
+        # workaround: for small files, we slurp into memory and return a StringStream
+        # assert isinstance(stream, ResponseStreamReader)
+        assert isinstance(stream, StringStream)
+
         # Verify Range header was included in the signed headers
         call_kwargs = mocked.call_args
         assert call_kwargs.kwargs["headers"]["Range"] == "bytes=0-6"
diff --git a/waterbutler/providers/oraclecloud/provider.py b/waterbutler/providers/oraclecloud/provider.py
index 663d5b5bf..f10c84950 100644
--- a/waterbutler/providers/oraclecloud/provider.py
+++ b/waterbutler/providers/oraclecloud/provider.py
@@ -20,8 +20,9 @@
 )
 from waterbutler.core.path import WaterButlerPath
 from waterbutler.core.provider import BaseProvider
-from waterbutler.core.streams import BaseStream, HashStreamWriter, ResponseStreamReader
+from waterbutler.core.streams import BaseStream, HashStreamWriter, ResponseStreamReader, StringStream
 from waterbutler.core.utils import make_disposition
+from waterbutler.providers.googlecloud import settings as pd_settings
 from waterbutler.providers.oraclecloud.metadata import (
     BaseOracleCloudMetadata,
     OracleCloudFileMetadata,
@@ -307,6 +308,12 @@ async def download(
                 f"Object not found: {path}", code=HTTPStatus.NOT_FOUND
             )
 
+        resp_size = resp.headers.get('Content-Length', None)
+        if resp_size is not None and int(resp_size) < pd_settings.MAX_SLURP_SIZE:
+            stream = StringStream(await resp.read())
+            await resp.release()
+            return stream
+
         return ResponseStreamReader(resp)
 
     async def delete(self, path: WaterButlerPath, *args, **kwargs) -> None:  # type: ignore[override]
diff --git a/waterbutler/providers/oraclecloud/settings.py b/waterbutler/providers/oraclecloud/settings.py
index 6577c2bf5..8750714dd 100644
--- a/waterbutler/providers/oraclecloud/settings.py
+++ b/waterbutler/providers/oraclecloud/settings.py
@@ -1,3 +1,6 @@
 from waterbutler import settings
 
 config = settings.child("ORACLECLOUD_PROVIDER_SETTINGS")
+
+# slurp downloads below this threshhold (in bytes)
+MAX_SLURP_SIZE = 100 * 1000