Fix a bug in Dialyzer related to call-site analysis

Dialyzer's "dataflow" module is using information from the callgraph to
determine which functions may be called at a particular call-site. Unfortunately
this information can include functions that are certainly not among the possible
choices. We don't want to emit warnings in such cases, so a "reasonable"
compromise is to stay silent if there are many possible funs and at least one of
them can succeed.

Bug reported by Dan Gudmundsson, test shrunk down by Magnus Lång.

1 files changed, 22 insertions, 0 deletions

diff --git a/lib/dialyzer/test/user_SUITE_data/src/wpc_hlines.erl b/lib/dialyzer/test/user_SUITE_data/src/wpc_hlines.erl
new file mode 100644
index 0000000000..8c205a8247
--- /dev/null
+++ b/lib/dialyzer/test/user_SUITE_data/src/wpc_hlines.erl
@@ -0,0 +1,22 @@
+%% Bug reported by Dan Gudmundsson, test shrunk down by Magnus Lång.
+
+%% The problem is that dialyzer_dep generates edges from the fun 
+%% application to both of the functions, and then during the warning pass 
+%% dialyzer_dataflow:handle_apply_or_call generates warnings for any such 
+%% edge that won't return.
+
+%% Since dialyzer_dep is currently supposed to overapproximate rather than
+%% underapproximate, the fix was to modify handle_apply_or_call to not generate
+%% warnings if some of the possible funs can succeed.
+
+-module(wpc_hlines).
+
+-export([do_export/0]).
+
+do_export() ->
+   {Proj, _} =  % The culprit seems to be putting the funs in a tuple
+     {fun good/1, fun bad/1},
+   Proj(true).
+
+good(_) -> ok.
+bad(false) -> ok.