1 files changed, 80 insertions, 0 deletions
diff --git a/packages/meshbay-node/tests/test_linkpreview.py b/packages/meshbay-node/tests/test_linkpreview.py
index a14b173..9fca186 100644
--- a/packages/meshbay-node/tests/test_linkpreview.py
+++ b/packages/meshbay-node/tests/test_linkpreview.py
@@ -193,3 +193,83 @@ async def test_fetch_image_refuses_a_decompression_bomb(resolves_public, monkeyp
         return httpx.Response(200, headers={"content-type": "image/png"}, content=bomb)
     async with _client(handler) as c:
         assert await linkpreview.fetch_image("https://example.com/x.png", client=c) is None
+
+
+# ── Size caps bind while reading, not after ─────────────────────────────────
+#
+# `client.get` read and decoded the whole body before the caps looked at it, so
+# a link posted in chat could make the node hold any amount of data. These
+# count what the server actually had to hand over.
+
+_CHUNK = 64 * 1024
+
+
+def _endless(counter, head=b""):
+    async def body():
+        if head:
+            counter["sent"] += len(head)
+            yield head
+        for _ in range(2000):                    # 125 MiB if nobody stops
+            counter["sent"] += _CHUNK
+            yield b"x" * _CHUNK
+    return body()
+
+
+async def test_a_huge_page_is_not_read_past_the_cap(resolves_public):
+    counter = {"sent": 0}
+    head = b"<html><head><title>T</title></head><body>"
+
+    def handler(request):
+        return httpx.Response(200, headers={"content-type": "text/html"},
+                              content=_endless(counter, head))
+    async with _client(handler) as c:
+        meta = await linkpreview.fetch_preview("https://example.com/", client=c)
+    assert meta is not None and meta["title"] == "T"
+    assert counter["sent"] <= linkpreview._MAX_HTML_BYTES + 2 * _CHUNK
+
+
+async def test_a_huge_image_is_refused_without_being_read(resolves_public):
+    counter = {"sent": 0}
+
+    def handler(request):
+        return httpx.Response(200, headers={"content-type": "image/png"},
+                              content=_endless(counter))
+    async with _client(handler) as c:
+        assert await linkpreview.fetch_image("https://example.com/x.png", client=c) is None
+    assert counter["sent"] <= linkpreview._MAX_IMAGE_BYTES + 2 * _CHUNK
+
+
+async def test_a_compressed_body_is_capped_after_decoding(resolves_public):
+    import zlib
+    comp = zlib.compressobj(9, zlib.DEFLATED, 31)          # gzip
+    counter = {"inflated": 0}
+
+    async def body():
+        yield comp.compress(b"<html><head><title>T</title></head><body>")
+        for _ in range(2000):                               # 125 MiB inflated
+            counter["inflated"] += _CHUNK
+            # Flushed per chunk, so what is counted is what went on the wire.
+            yield comp.compress(b"x" * _CHUNK) + comp.flush(zlib.Z_SYNC_FLUSH)
+        yield comp.flush()
+
+    def handler(request):
+        return httpx.Response(200, headers={"content-type": "text/html",
+                                            "content-encoding": "gzip"},
+                              content=body())
+    async with _client(handler) as c:
+        meta = await linkpreview.fetch_preview("https://example.com/", client=c)
+    assert meta is not None
+    assert counter["inflated"] < 50 * _CHUNK      # stopped early, not at 125 MiB
+
+
+async def test_a_declared_oversized_image_is_not_read(resolves_public):
+    counter = {"sent": 0}
+
+    def handler(request):
+        return httpx.Response(
+            200, headers={"content-type": "image/png",
+                          "content-length": str(linkpreview._MAX_IMAGE_BYTES + 1)},
+            content=_endless(counter))
+    async with _client(handler) as c:
+        assert await linkpreview.fetch_image("https://example.com/x.png", client=c) is None
+    assert counter["sent"] <= _CHUNK