CQ shared: Fix off-by-nine error leading to lost messages

And `eof` crashes.

The problem is that we may end up trying to read more data
from the file when scanning, despite being at the end of the
file. This results in the current Acc to be returned instead
of the remaining data being parsed.

This results in some messages at the end of the file being
truncated off despite still being in memory (and still pointing
to the end of the original file, well past the truncation point).

(cherry picked from commit f363854)

Author: lhoguin

deps/rabbit/src/rabbit_msg_store.erl

@@ -1599,7 +1599,7 @@ scan_data(<<Size:64, MsgIdInt:128, _Rest/bits>> = Data, Fd, Fun, Offset, FileSiz
     end;
 %% This might be the start of a message.
 scan_data(<<Size:64, Rest/bits>> = Data, Fd, Fun, Offset, FileSize, MsgIdsFound, Acc)
-          when byte_size(Rest) < Size + 1, Size < FileSize - Offset ->
+          when byte_size(Rest) < Size + 1, Size + 9 =< FileSize - Offset ->
     scan(Data, Fd, Fun, Offset, FileSize, MsgIdsFound, Acc);
 scan_data(Data, Fd, Fun, Offset, FileSize, MsgIdsFound, Acc)
           when byte_size(Data) < 8 ->

deps/rabbit/test/backing_queue_SUITE.erl

@@ -710,6 +710,17 @@ msg_store_file_scan1(Config) ->
         [{bin, <<0, 0:48, 17, 17, "idididididididid", 255, 0:4352/unit:8, 255>>}],
         {ok, [{<<"idididididididid">>, 4378, 1}]},
         fun(Obj = {<<"idididididididid">>, 4378, 1}) -> {valid, Obj}; (_) -> invalid end),
+    %% Off-by-nine regression testing. The file scanning could miss
+    %% some messages if previous data looked like a message but its
+    %% size went past the end of the file.
+    lists:foreach(fun(N) ->
+        ok = Scan([
+            {bin, <<(4194304 + N):64, 0:(4194304 - 8 - 25 - 10)/unit:8>>},
+            {msg, gen_id(), <<>>},
+            %% Padding ensures there's no 255 at the end of the size indicated by 'bin'.
+            {pad, 10}
+        ])
+    end, lists:seq(-9, -1)),
     %% All good!!
     passed.