When cover is enabled in rebar.config after test meck failed with enoent

[jonasrichard]

```=ERROR REPORT==== 27-Aug-2013::13:36:22 ===
** Generic server my_module_meck terminating
** Last message in was {'EXIT',<0.150.0>,normal}
** When Server state == {state,my_module,
                               {dict,0,16,16,8,80,48,
                                     {[],[],[],[],[],[],[],[],[],[],[],[],[],
                                      [],[],[]},
                                     {{[],[],[],[],[],[],[],[],[],[],[],[],[],
                                       [],[],[]}}},
                               true,[],
                               {{"/xxxxxx/apps/my_app/.eunit/my_module.beam",
                                 "my_module.coverdata",
                                 [{outdir,".eunit"},
                                  debug_info,
                                  {d,'TEST'},
                                  fail_on_warning,debug_info,
                                  {i,"../"},
                                  {i,"include"}]},
                                <<70,79,82,49```

The test using:

    meck:new(my_module, [passthrough]).
    meck:expect(my_module, time, ...).
    meck:unload(my_module).
    meck:new(my_module).

The reason behind the unload/new again is to leave the module loaded. Otherwise consecutive tests will fail in error:undef my_module:time/1.
Am I doing something wrong or meck is confused?

[eproxus]

Could you attach a (preferably) minimal test module so I can reproduce this?

[jonasrichard]

I managed to reproduce but the bug is just partly related to the cover thing, it came up with lager_transform only.

rebar.config

{erl_opts, [debug_info,
            {parse_transform, lager_transform}]}.

{cover_enabled, true}.

{deps, [
    {lager, ".*",
        {git, "git://github.com/basho/lager.git", {tag, "2.0.0"}}},
    {meck, ".*",
        {git, "git://github.com/eproxus/meck.git", {tag, "0.7.2"}}}
]}.

src/meckbug.app.src: just a placeholder in order that rebar starts work

{application, meckbug,
 [
  {description, ""},
  {vsn, "1"},
  {registered, []},
  {applications, [
                  kernel,
                  stdlib
                 ]},
  {mod, { meckbug_app, []}},
  {env, []}
 ]}.

src/my_module.erl:

-module(my_module).

-export([calculate/2]).

calculate(A, B) ->
    lager:info("Addig ~p and ~p", [A, B]),   %% without this, everything is fine
    A + B.

src/my_other.erl:

-module(my_other).

-ifdef(TEST).
-include_lib("eunit/include/eunit.hrl").
-endif.

-ifdef(TEST).

calculate_test() ->
    meck:new(my_module, [passthrough]),
    meck:expect(my_module, calculate,
        fun(1, 2) ->
            5
        end),

    ?debugVal(my_module:calculate(1, 2)),
    ?assertEqual(5, my_module:calculate(1, 2)),

    meck:unload(my_module),
    meck:new(my_module).

-endif.

./rebar compile
./rebar eunit apps=meckbug

So the problem is to meck:new() a module which is lager transformed.

[rlipscombe]

I'm seeing the same problem. Was there any resolution?

[rlipscombe]

*Ignore this, this is a separate issue #114 *

Further investigation seems to show meck_cover:read_cover_file being called with a relative path "my_module_meck_original.coverdata", but with file:get_cwd() returning "/path/to/my/apps/app", rather than "/path/to/my/apps/app/.eunit".

If I hack on meck_cover:read_cover_file to make it also look in the .eunit folder, it works -- at least until it barfs in meck_proc:restore_original instead.

[shino]

@jonasrichard You call meck:new(my_module) again at the end of my_other:calculate_test/0.
It seems unnecessary, isn't it? When I use meck/cover/lager at once, I encountered the same error but went well after proper use of meck:unload/0,1.

[jonasrichard]

Yes, it seems that removing extra meck:new solves the problem. I wrote that because somehow I experienced that after meck:unload I wasn't able to use the non-mocked module.

    meck:unload(my_module),
    my_module:calculate(1, 2).
    %%meck:new(my_module).

After changing the eunit test I didn't get any error, and I am able to use my_module. So I am closing this issue.