%%
%% %CopyrightBegin%
%%
%% Copyright Ericsson AB 2006-2012. All Rights Reserved.
%%
%% The contents of this file are subject to the Erlang Public License,
%% Version 1.1, (the "License"); you may not use this file except in
%% compliance with the License. You should have received a copy of the
%% Erlang Public License along with this software. If not, it can be
%% retrieved online at http://www.erlang.org/.
%%
%% Software distributed under the License is distributed on an "AS IS"
%% basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
%% the License for the specific language governing rights and limitations
%% under the License.
%%
%% %CopyrightEnd%
%%
%%% @doc Common Test Framework functions handling test specifications.
%%%
%%% <p>This module exports functions that are used within CT to
%%% scan and parse test specifikations.</p>
-module(ct_testspec).
-export([prepare_tests/1, prepare_tests/2,
collect_tests_from_list/2, collect_tests_from_list/3,
collect_tests_from_file/2, collect_tests_from_file/3]).
-include("ct_util.hrl").
-define(testspec_fields, record_info(fields, testspec)).
%%%------------------------------------------------------------------
%%% NOTE:
%%% Multiple testspecs may be used as input with the result that
%%% the data is merged. It's in this case up to the user to ensure
%%% there are no clashes in any "global" variables, such as logdir.
%%%-------------------------------------------------------------------
%%%-------------------------------------------------------------------
%%% prepare_tests/2 compiles the testspec data into a list of tests
%%% to be run and a list of tests to be skipped, either for one
%%% particular node or for all nodes.
%%%-------------------------------------------------------------------
%%%-------------------------------------------------------------------
%%% Version 1 - extract and return all tests and skips for Node
%%% (incl all_nodes)
%%%-------------------------------------------------------------------
prepare_tests(TestSpec,Node) when is_record(TestSpec,testspec),
is_atom(Node) ->
case lists:keysearch(Node,1,prepare_tests(TestSpec)) of
{value,{Node,Run,Skip}} ->
{Run,Skip};
false ->
{[],[]}
end.
%%%-------------------------------------------------------------------
%%% Version 2 - create and return a list of {Node,Run,Skip} tuples,
%%% one for each node specified in the test specification.
%%% The tuples in the Run list will have the form {Dir,Suites,Cases}
%%% and the tuples in the Skip list will have the form
%%% {Dir,Suites,Comment} or {Dir,Suite,Cases,Comment}.
%%%-------------------------------------------------------------------
prepare_tests(TestSpec) when is_record(TestSpec,testspec) ->
Tests = TestSpec#testspec.tests,
%% Sort Tests into "flat" Run and Skip lists (not sorted per node).
{Run,Skip} = get_run_and_skip(Tests,[],[]),
%% Create initial list of {Node,{Run,Skip}} tuples
NodeList = lists:map(fun(N) -> {N,{[],[]}} end, list_nodes(TestSpec)),
%% Get all Run tests sorted per node basis.
NodeList1 = run_per_node(Run,NodeList,
TestSpec#testspec.merge_tests),
%% Get all Skip entries sorted per node basis.
NodeList2 = skip_per_node(Skip,NodeList1),
%% Change representation.
Result=
lists:map(fun({Node,{Run1,Skip1}}) ->
Run2 = lists:map(fun({D,{Ss,Cs}}) ->
{D,Ss,Cs}
end, Run1),
Skip2 = lists:map(fun({D,{Ss,Cmt}}) ->
{D,Ss,Cmt};
({D,{S,Cs,Cmt}}) ->
{D,S,Cs,Cmt}
end, Skip1),
{Node,Run2,Skip2}
end, NodeList2),
Result.
%% run_per_node/2 takes the Run list as input and returns a list
%% of {Node,RunPerNode,[]} tuples where the tests have been sorted
%% on a per node basis.
run_per_node([{{Node,Dir},Test}|Ts],Result, MergeTests) ->
{value,{Node,{Run,Skip}}} = lists:keysearch(Node,1,Result),
Run1 = case MergeTests of
false ->
append({Dir, Test}, Run);
true ->
merge_tests(Dir,Test,Run)
end,
run_per_node(Ts,insert_in_order({Node,{Run1,Skip}},Result),
MergeTests);
run_per_node([],Result,_) ->
Result.
merge_tests(Dir,Test={all,_},TestDirs) ->
%% overwrite all previous entries for Dir
TestDirs1 = lists:filter(fun({D,_}) when D==Dir ->
false;
(_) ->
true
end,TestDirs),
insert_in_order({Dir,Test},TestDirs1);
merge_tests(Dir,Test={Suite,all},TestDirs) ->
TestDirs1 = lists:filter(fun({D,{S,_}}) when D==Dir,S==Suite ->
false;
(_) ->
true
end,TestDirs),
TestDirs1++[{Dir,Test}];
merge_tests(Dir,Test,TestDirs) ->
merge_suites(Dir,Test,TestDirs).
merge_suites(Dir,{Suite,Cases},[{Dir,{Suite,Cases0}}|Dirs]) ->
Cases1 = insert_in_order(Cases,Cases0),
[{Dir,{Suite,Cases1}}|Dirs];
merge_suites(Dir,Test,[Other|Dirs]) ->
[Other|merge_suites(Dir,Test,Dirs)];
merge_suites(Dir,Test,[]) ->
[{Dir,Test}].
%% skip_per_node/2 takes the Skip list as input and returns a list
%% of {Node,RunPerNode,SkipPerNode} tuples where the skips have been
%% sorted on a per node basis.
skip_per_node([{{Node,Dir},Test}|Ts],Result) ->
{value,{Node,{Run,Skip}}} = lists:keysearch(Node,1,Result),
Skip1 = [{Dir,Test}|Skip],
skip_per_node(Ts,insert_in_order({Node,{Run,Skip1}},Result));
skip_per_node([],Result) ->
Result.
%% get_run_and_skip/3 takes a list of test terms as input and sorts
%% them into a list of Run tests and a list of Skip entries. The
%% elements all have the form
%%
%% {{Node,Dir},TestData}
%%
%% TestData has the form:
%%
%% Run entry: {Suite,Cases}
%%
%% Skip entry: {Suites,Comment} or {Suite,Cases,Comment}
%%
get_run_and_skip([{{Node,Dir},Suites}|Tests],Run,Skip) ->
TestDir = ct_util:get_testdir(Dir,catch element(1,hd(Suites))),
case lists:keysearch(all,1,Suites) of
{value,_} -> % all Suites in Dir
Skipped = get_skipped_suites(Node,TestDir,Suites),
%% note: this adds an 'all' test even if only skip is specified,
%% probably a good thing cause it gets logged as skipped then
get_run_and_skip(Tests,
[[{{Node,TestDir},{all,all}}]|Run],
[Skipped|Skip]);
false ->
{R,S} = prepare_suites(Node,TestDir,Suites,[],[]),
get_run_and_skip(Tests,[R|Run],[S|Skip])
end;
get_run_and_skip([],Run,Skip) ->
{lists:flatten(lists:reverse(Run)),
lists:flatten(lists:reverse(Skip))}.
prepare_suites(Node,Dir,[{Suite,Cases}|Suites],Run,Skip) ->
case lists:member(all,Cases) of
true -> % all Cases in Suite
Skipped = get_skipped_cases(Node,Dir,Suite,Cases),
%% note: this adds an 'all' test even if only skip is specified
prepare_suites(Node,Dir,Suites,
[[{{Node,Dir},{Suite,all}}]|Run],
[Skipped|Skip]);
false ->
{RL,SL} = prepare_cases(Node,Dir,Suite,Cases),
prepare_suites(Node,Dir,Suites,[RL|Run],[SL|Skip])
end;
prepare_suites(_Node,_Dir,[],Run,Skip) ->
{lists:flatten(lists:reverse(Run)),
lists:flatten(lists:reverse(Skip))}.
prepare_cases(Node,Dir,Suite,Cases) ->
case get_skipped_cases(Node,Dir,Suite,Cases) of
SkipAll=[{{Node,Dir},{Suite,_Cmt}}] -> % all cases to be skipped
%% note: this adds an 'all' test even if only skip is specified
{[{{Node,Dir},{Suite,all}}],SkipAll};
Skipped ->
%% note: this adds a test even if only skip is specified
PrepC = lists:foldr(fun({{G,Cs},{skip,_Cmt}}, Acc) when
is_atom(G) ->
case lists:keymember(G, 1, Cases) of
true ->
Acc;
false ->
[{skipped,G,Cs}|Acc]
end;
({C,{skip,_Cmt}},Acc) ->
case lists:member(C,Cases) of
true ->
Acc;
false ->
[C|Acc]
end;
(C,Acc) -> [C|Acc]
end, [], Cases),
{{{Node,Dir},{Suite,PrepC}},Skipped}
end.
get_skipped_suites(Node,Dir,Suites) ->
lists:flatten(get_skipped_suites1(Node,Dir,Suites)).
get_skipped_suites1(Node,Dir,[{Suite,Cases}|Suites]) ->
SkippedCases = get_skipped_cases(Node,Dir,Suite,Cases),
[SkippedCases|get_skipped_suites1(Node,Dir,Suites)];
get_skipped_suites1(_,_,[]) ->
[].
get_skipped_cases(Node,Dir,Suite,Cases) ->
case lists:keysearch(all,1,Cases) of
{value,{all,{skip,Cmt}}} ->
[{{Node,Dir},{Suite,Cmt}}];
_ ->
get_skipped_cases1(Node,Dir,Suite,Cases)
end.
get_skipped_cases1(Node,Dir,Suite,[{Case,{skip,Cmt}}|Cs]) ->
[{{Node,Dir},{Suite,Case,Cmt}}|get_skipped_cases1(Node,Dir,Suite,Cs)];
get_skipped_cases1(Node,Dir,Suite,[_Case|Cs]) ->
get_skipped_cases1(Node,Dir,Suite,Cs);
get_skipped_cases1(_,_,_,[]) ->
[].
%%% collect_tests_from_file reads a testspec file and returns a record
%%% containing the data found.
collect_tests_from_file(Specs, Relaxed) ->
collect_tests_from_file(Specs,[node()],Relaxed).
collect_tests_from_file(Specs,Nodes,Relaxed) when is_list(Nodes) ->
NodeRefs = lists:map(fun(N) -> {undefined,N} end, Nodes),
catch collect_tests_from_file1(Specs,#testspec{nodes=NodeRefs},Relaxed).
collect_tests_from_file1([Spec|Specs],TestSpec,Relaxed) ->
SpecDir = filename:dirname(filename:absname(Spec)),
case file:consult(Spec) of
{ok,Terms} ->
case collect_tests(Terms,
TestSpec#testspec{spec_dir=SpecDir},
Relaxed) of
TS = #testspec{tests=Tests, logdir=LogDirs} when Specs == [] ->
LogDirs1 = lists:delete(".",LogDirs) ++ ["."],
TS#testspec{tests=lists:flatten(Tests), logdir=LogDirs1};
TS = #testspec{alias = As, nodes = Ns} ->
TS1 = TS#testspec{alias = lists:reverse(As),
nodes = lists:reverse(Ns)},
collect_tests_from_file1(Specs,TS1,Relaxed)
end;
{error,Reason} ->
ReasonStr =
lists:flatten(io_lib:format("~s",
[file:format_error(Reason)])),
throw({error,{Spec,ReasonStr}})
end.
collect_tests_from_list(Terms,Relaxed) ->
collect_tests_from_list(Terms,[node()],Relaxed).
collect_tests_from_list(Terms,Nodes,Relaxed) when is_list(Nodes) ->
{ok,Cwd} = file:get_cwd(),
NodeRefs = lists:map(fun(N) -> {undefined,N} end, Nodes),
case catch collect_tests(Terms,#testspec{nodes=NodeRefs,
spec_dir=Cwd},
Relaxed) of
E = {error,_} ->
E;
TS ->
#testspec{tests=Tests, logdir=LogDirs} = TS,
LogDirs1 = lists:delete(".",LogDirs) ++ ["."],
TS#testspec{tests=lists:flatten(Tests), logdir=LogDirs1}
end.
collect_tests(Terms,TestSpec,Relaxed) ->
put(relaxed,Relaxed),
Terms1 = replace_names(Terms),
TestSpec1 = get_global(Terms1,TestSpec),
TestSpec2 = get_all_nodes(Terms1,TestSpec1),
{Terms2, TestSpec3} = filter_init_terms(Terms1, [], TestSpec2),
add_tests(Terms2,TestSpec3).
%% replace names (atoms) in the testspec matching those in 'define' terms by
%% searching recursively through tuples and lists
replace_names(Terms) ->
Defs =
lists:flatmap(fun(Def={define,Name,_Replacement}) ->
%% check that name follows convention
if not is_atom(Name) ->
throw({illegal_name_in_testspec,Name});
true ->
[First|_] = atom_to_list(Name),
if ((First == $?) or (First == $$)
or (First == $_)
or ((First >= $A)
and (First =< $Z))) ->
[Def];
true ->
throw({illegal_name_in_testspec,
Name})
end
end;
(_) -> []
end, Terms),
DefProps = replace_names_in_defs(Defs,[]),
replace_names(Terms,[],DefProps).
replace_names_in_defs([Def|Left],ModDefs) ->
[{define,Name,Replacement}] = replace_names([Def],[],ModDefs),
replace_names_in_defs(Left,[{Name,Replacement}|ModDefs]);
replace_names_in_defs([],ModDefs) ->
ModDefs.
replace_names([Term|Ts],Modified,Defs) when is_tuple(Term) ->
[TypeTag|Data] = tuple_to_list(Term),
Term1 = list_to_tuple([TypeTag|replace_names_in_elems(Data,[],Defs)]),
replace_names(Ts,[Term1|Modified],Defs);
replace_names([Term|Ts],Modified,Defs) when is_atom(Term) ->
case proplists:get_value(Term,Defs) of
undefined ->
replace_names(Ts,[Term|Modified],Defs);
Replacement ->
replace_names(Ts,[Replacement|Modified],Defs)
end;
replace_names([Term=[Ch|_]|Ts],Modified,Defs) when is_integer(Ch) ->
%% Term *could* be a string, attempt to search through it
Term1 = replace_names_in_string(Term,Defs),
replace_names(Ts,[Term1|Modified],Defs);
replace_names([Term|Ts],Modified,Defs) ->
replace_names(Ts,[Term|Modified],Defs);
replace_names([],Modified,_Defs) ->
lists:reverse(Modified).
replace_names_in_elems([Elem|Es],Modified,Defs) when is_tuple(Elem) ->
Elem1 = list_to_tuple(replace_names_in_elems(tuple_to_list(Elem),[],Defs)),
replace_names_in_elems(Es,[Elem1|Modified],Defs);
replace_names_in_elems([Elem|Es],Modified,Defs) when is_atom(Elem) ->
case proplists:get_value(Elem,Defs) of
undefined ->
%% if Term is a node name, check it for replacements as well
Elem1 = replace_names_in_node(Elem,Defs),
replace_names_in_elems(Es,[Elem1|Modified],Defs);
Replacement ->
replace_names_in_elems(Es,[Replacement|Modified],Defs)
end;
replace_names_in_elems([Elem=[Ch|_]|Es],Modified,Defs) when is_integer(Ch) ->
%% Term *could* be a string, attempt to search through it
case replace_names_in_string(Elem,Defs) of
Elem ->
List = replace_names_in_elems(Elem,[],Defs),
replace_names_in_elems(Es,[List|Modified],Defs);
Elem1 ->
replace_names_in_elems(Es,[Elem1|Modified],Defs)
end;
replace_names_in_elems([Elem|Es],Modified,Defs) when is_list(Elem) ->
List = replace_names_in_elems(Elem,[],Defs),
replace_names_in_elems(Es,[List|Modified],Defs);
replace_names_in_elems([Elem|Es],Modified,Defs) ->
replace_names_in_elems(Es,[Elem|Modified],Defs);
replace_names_in_elems([],Modified,_Defs) ->
lists:reverse(Modified).
replace_names_in_string(Term,Defs=[{Name,Replacement=[Ch|_]}|Ds])
when is_integer(Ch) ->
try re:replace(Term,[$'|atom_to_list(Name)]++"'",
Replacement,[{return,list}]) of
Term -> % no match, proceed
replace_names_in_string(Term,Ds);
Term1 ->
replace_names_in_string(Term1,Defs)
catch
_:_ -> Term % Term is not a string
end;
replace_names_in_string(Term,[_|Ds]) ->
replace_names_in_string(Term,Ds);
replace_names_in_string(Term,[]) ->
Term.
replace_names_in_node(Node,Defs) ->
String = atom_to_list(Node),
case lists:member($@,String) of
true ->
list_to_atom(replace_names_in_node1(String,Defs));
false ->
Node
end.
replace_names_in_node1(NodeStr,Defs=[{Name,Replacement}|Ds]) ->
ReplStr = case Replacement of
[Ch|_] when is_integer(Ch) -> Replacement;
_ when is_atom(Replacement) -> atom_to_list(Replacement);
_ -> false
end,
if ReplStr == false ->
replace_names_in_node1(NodeStr,Ds);
true ->
case re:replace(NodeStr,atom_to_list(Name),
ReplStr,[{return,list}]) of
NodeStr -> % no match, proceed
replace_names_in_node1(NodeStr,Ds);
NodeStr1 ->
replace_names_in_node1(NodeStr1,Defs)
end
end;
replace_names_in_node1(NodeStr,[]) ->
NodeStr.
%% global terms that will be used for analysing all other terms in the spec
get_global([{merge_tests,Bool} | Ts], Spec) ->
get_global(Ts,Spec#testspec{merge_tests=Bool});
%% the 'define' term replaces the 'alias' and 'node' terms, but we need to keep
%% the latter two for backwards compatibility...
get_global([{alias,Ref,Dir}|Ts],Spec=#testspec{alias=Refs}) ->
get_global(Ts,Spec#testspec{alias=[{Ref,get_absdir(Dir,Spec)}|Refs]});
get_global([{node,Ref,Node}|Ts],Spec=#testspec{nodes=Refs}) ->
get_global(Ts,Spec#testspec{nodes=[{Ref,Node} |
lists:keydelete(Node,2,Refs)]});
get_global([_|Ts],Spec) ->
get_global(Ts,Spec);
get_global([],Spec=#testspec{nodes=Ns, alias=As}) ->
Spec#testspec{nodes=lists:reverse(Ns), alias=lists:reverse(As)}.
get_absfile(Callback,FullName,#testspec{spec_dir=SpecDir}) ->
% we need to temporary switch to new cwd here, because
% otherwise config files cannot be found
{ok, OldWd} = file:get_cwd(),
ok = file:set_cwd(SpecDir),
R = Callback:check_parameter(FullName),
ok = file:set_cwd(OldWd),
case R of
{ok, {file, FullName}}->
File = filename:basename(FullName),
Dir = get_absname(filename:dirname(FullName),SpecDir),
filename:join(Dir,File);
{ok, {config, FullName}}->
FullName;
{error, {nofile, FullName}}->
FullName;
{error, {wrong_config, FullName}}->
FullName
end.
get_absfile(FullName,#testspec{spec_dir=SpecDir}) ->
File = filename:basename(FullName),
Dir = get_absname(filename:dirname(FullName),SpecDir),
filename:join(Dir,File).
get_absdir(Dir,#testspec{spec_dir=SpecDir}) ->
get_absname(Dir,SpecDir).
get_absname(Dir,SpecDir) ->
AbsName = filename:absname(Dir,SpecDir),
shorten_path(AbsName,SpecDir).
shorten_path(Path,SpecDir) ->
case shorten_split_path(filename:split(Path),[]) of
[] ->
[Root|_] = filename:split(SpecDir),
Root;
Short ->
filename:join(Short)
end.
shorten_split_path([".."|Path],SoFar) ->
shorten_split_path(Path,tl(SoFar));
shorten_split_path(["."|Path],SoFar) ->
shorten_split_path(Path,SoFar);
shorten_split_path([Dir|Path],SoFar) ->
shorten_split_path(Path,[Dir|SoFar]);
shorten_split_path([],SoFar) ->
lists:reverse(SoFar).
%% go through all tests and register all nodes found
get_all_nodes([{suites,Nodes,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{suites,Node,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{groups,[Char|_],_,_,_}|Ts],Spec) when is_integer(Char) ->
get_all_nodes(Ts,Spec);
get_all_nodes([{groups,Nodes,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{groups,Nodes,_,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{groups,_,_,_,{cases,_}}|Ts],Spec) ->
get_all_nodes(Ts,Spec);
get_all_nodes([{groups,Node,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{groups,Node,_,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{cases,Nodes,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{cases,Node,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{skip_suites,Nodes,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{skip_suites,Node,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{skip_groups,[Char|_],_,_,_,_}|Ts],Spec) when is_integer(Char) ->
get_all_nodes(Ts,Spec);
get_all_nodes([{skip_groups,Nodes,_,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{skip_groups,Node,_,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{skip_groups,Nodes,_,_,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{skip_groups,Node,_,_,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([{skip_cases,Nodes,_,_,_,_}|Ts],Spec) when is_list(Nodes) ->
get_all_nodes(Ts,save_nodes(Nodes,Spec));
get_all_nodes([{skip_cases,Node,_,_,_,_}|Ts],Spec) ->
get_all_nodes(Ts,save_nodes([Node],Spec));
get_all_nodes([_Other|Ts],Spec) ->
get_all_nodes(Ts,Spec);
get_all_nodes([],Spec) ->
Spec.
filter_init_terms([{init,InitOptions}|Ts],NewTerms,Spec) ->
filter_init_terms([{init,list_nodes(Spec),InitOptions}|Ts],
NewTerms,Spec);
filter_init_terms([{init,all_nodes,InitOptions}|Ts],NewTerms,Spec) ->
filter_init_terms([{init,list_nodes(Spec),InitOptions}|Ts],
NewTerms,Spec);
filter_init_terms([{init,NodeRef,InitOptions}|Ts],
NewTerms,Spec) when is_atom(NodeRef) ->
filter_init_terms([{init,[NodeRef],InitOptions}|Ts],NewTerms,Spec);
filter_init_terms([{init,NodeRefs,InitOption}|Ts],
NewTerms,Spec) when is_tuple(InitOption) ->
filter_init_terms([{init,NodeRefs,[InitOption]}|Ts],NewTerms,Spec);
filter_init_terms([{init,[NodeRef|NodeRefs],InitOptions}|Ts],
NewTerms,Spec=#testspec{init=InitData}) ->
NodeStartOptions =
case lists:keyfind(node_start,1,InitOptions) of
{node_start,NSOptions}->
case lists:keyfind(callback_module,1,NSOptions) of
{callback_module,_Callback}->
NSOptions;
false->
[{callback_module,ct_slave}|NSOptions]
end;
false->
[]
end,
EvalTerms = case lists:keyfind(eval,1,InitOptions) of
{eval,MFA} when is_tuple(MFA) ->
[MFA];
{eval,MFAs} when is_list(MFAs) ->
MFAs;
false->
[]
end,
Node = ref2node(NodeRef,Spec#testspec.nodes),
InitData2 = add_option({node_start,NodeStartOptions},Node,InitData,true),
InitData3 = add_option({eval,EvalTerms},Node,InitData2,false),
filter_init_terms([{init,NodeRefs,InitOptions}|Ts],
NewTerms,Spec#testspec{init=InitData3});
filter_init_terms([{init,[],_}|Ts],NewTerms,Spec) ->
filter_init_terms(Ts,NewTerms,Spec);
filter_init_terms([Term|Ts],NewTerms,Spec) ->
filter_init_terms(Ts,[Term|NewTerms],Spec);
filter_init_terms([],NewTerms,Spec) ->
{lists:reverse(NewTerms),Spec}.
add_option({Key,Value},Node,List,WarnIfExists) when is_list(Value) ->
OldOptions = case lists:keyfind(Node,1,List) of
{Node,Options}->
Options;
false->
[]
end,
NewOption = case lists:keyfind(Key,1,OldOptions) of
{Key,OldOption} when WarnIfExists,OldOption/=[]->
io:format("There is an option ~w=~w already "
"defined for node ~p, skipping new ~w~n",
[Key,OldOption,Node,Value]),
OldOption;
{Key,OldOption}->
OldOption ++ Value;
false->
Value
end,
lists:keystore(Node,1,List,
{Node,lists:keystore(Key,1,OldOptions,{Key,NewOption})});
add_option({Key,Value},Node,List,WarnIfExists) ->
add_option({Key,[Value]},Node,List,WarnIfExists).
save_nodes(Nodes,Spec=#testspec{nodes=NodeRefs}) ->
NodeRefs1 =
lists:foldr(fun(all_nodes,NR) ->
NR;
(Node,NR) ->
case lists:keymember(Node,1,NR) of
true ->
NR;
false ->
case lists:keymember(Node,2,NR) of
true ->
NR;
false ->
[{undefined,Node}|NR]
end
end
end,NodeRefs,Nodes),
Spec#testspec{nodes=NodeRefs1}.
list_nodes(#testspec{nodes=NodeRefs}) ->
lists:map(fun({_Ref,Node}) -> Node end, NodeRefs).
%% -----------------------------------------------------
%% / \
%% | When adding test/config terms, remember to update |
%% | valid_terms/0 also! |
%% \ /
%% -----------------------------------------------------
%% --- suites ---
add_tests([{suites,all_nodes,Dir,Ss}|Ts],Spec) ->
add_tests([{suites,list_nodes(Spec),Dir,Ss}|Ts],Spec);
add_tests([{suites,Dir,Ss}|Ts],Spec) ->
add_tests([{suites,all_nodes,Dir,Ss}|Ts],Spec);
add_tests([{suites,Nodes,Dir,Ss}|Ts],Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,suites,[Dir,Ss],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{suites,Node,Dir,Ss}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = insert_suites(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Ss,Tests, Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
%% --- groups ---
%% Later make it possible to specify group execution properties
%% that will override thse in the suite. Also make it possible
%% create dynamic groups in specification, i.e. to group test cases
%% by means of groups defined only in the test specification.
add_tests([{groups,all_nodes,Dir,Suite,Gs}|Ts],Spec) ->
add_tests([{groups,list_nodes(Spec),Dir,Suite,Gs}|Ts],Spec);
add_tests([{groups,all_nodes,Dir,Suite,Gs,{cases,TCs}}|Ts],Spec) ->
add_tests([{groups,list_nodes(Spec),Dir,Suite,Gs,{cases,TCs}}|Ts],Spec);
add_tests([{groups,Dir,Suite,Gs}|Ts],Spec) ->
add_tests([{groups,all_nodes,Dir,Suite,Gs}|Ts],Spec);
add_tests([{groups,Dir,Suite,Gs,{cases,TCs}}|Ts],Spec) ->
add_tests([{groups,all_nodes,Dir,Suite,Gs,{cases,TCs}}|Ts],Spec);
add_tests([{groups,Nodes,Dir,Suite,Gs}|Ts],Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,groups,[Dir,Suite,Gs],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{groups,Nodes,Dir,Suite,Gs,{cases,TCs}}|Ts],
Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,groups,[Dir,Suite,Gs,{cases,TCs}],Ts,
Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{groups,Node,Dir,Suite,Gs}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = insert_groups(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Suite,Gs,all,Tests,
Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
add_tests([{groups,Node,Dir,Suite,Gs,{cases,TCs}}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = insert_groups(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Suite,Gs,TCs,Tests,
Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
%% --- cases ---
add_tests([{cases,all_nodes,Dir,Suite,Cs}|Ts],Spec) ->
add_tests([{cases,list_nodes(Spec),Dir,Suite,Cs}|Ts],Spec);
add_tests([{cases,Dir,Suite,Cs}|Ts],Spec) ->
add_tests([{cases,all_nodes,Dir,Suite,Cs}|Ts],Spec);
add_tests([{cases,Nodes,Dir,Suite,Cs}|Ts],Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,cases,[Dir,Suite,Cs],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{cases,Node,Dir,Suite,Cs}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = insert_cases(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Suite,Cs,Tests, Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
%% --- skip_suites ---
add_tests([{skip_suites,all_nodes,Dir,Ss,Cmt}|Ts],Spec) ->
add_tests([{skip_suites,list_nodes(Spec),Dir,Ss,Cmt}|Ts],Spec);
add_tests([{skip_suites,Dir,Ss,Cmt}|Ts],Spec) ->
add_tests([{skip_suites,all_nodes,Dir,Ss,Cmt}|Ts],Spec);
add_tests([{skip_suites,Nodes,Dir,Ss,Cmt}|Ts],Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,skip_suites,[Dir,Ss,Cmt],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{skip_suites,Node,Dir,Ss,Cmt}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = skip_suites(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Ss,Cmt,Tests,
Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
%% --- skip_groups ---
add_tests([{skip_groups,all_nodes,Dir,Suite,Gs,Cmt}|Ts],Spec) ->
add_tests([{skip_groups,list_nodes(Spec),Dir,Suite,Gs,Cmt}|Ts],Spec);
add_tests([{skip_groups,all_nodes,Dir,Suite,Gs,{cases,TCs},Cmt}|Ts],Spec) ->
add_tests([{skip_groups,list_nodes(Spec),Dir,Suite,Gs,{cases,TCs},Cmt}|Ts],
Spec);
add_tests([{skip_groups,Dir,Suite,Gs,Cmt}|Ts],Spec) ->
add_tests([{skip_groups,all_nodes,Dir,Suite,Gs,Cmt}|Ts],Spec);
add_tests([{skip_groups,Dir,Suite,Gs,{cases,TCs},Cmt}|Ts],Spec) ->
add_tests([{skip_groups,all_nodes,Dir,Suite,Gs,{cases,TCs},Cmt}|Ts],Spec);
add_tests([{skip_groups,Nodes,Dir,Suite,Gs,Cmt}|Ts],Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,skip_groups,[Dir,Suite,Gs,Cmt],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{skip_groups,Nodes,Dir,Suite,Gs,{cases,TCs},Cmt}|Ts],
Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,skip_groups,[Dir,Suite,Gs,{cases,TCs},Cmt],Ts,
Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{skip_groups,Node,Dir,Suite,Gs,Cmt}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = skip_groups(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Suite,Gs,all,Cmt,Tests,
Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
add_tests([{skip_groups,Node,Dir,Suite,Gs,{cases,TCs},Cmt}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = skip_groups(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Suite,Gs,TCs,Cmt,Tests,
Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
%% --- skip_cases ---
add_tests([{skip_cases,all_nodes,Dir,Suite,Cs,Cmt}|Ts],Spec) ->
add_tests([{skip_cases,list_nodes(Spec),Dir,Suite,Cs,Cmt}|Ts],Spec);
add_tests([{skip_cases,Dir,Suite,Cs,Cmt}|Ts],Spec) ->
add_tests([{skip_cases,all_nodes,Dir,Suite,Cs,Cmt}|Ts],Spec);
add_tests([{skip_cases,Nodes,Dir,Suite,Cs,Cmt}|Ts],Spec) when is_list(Nodes) ->
Ts1 = separate(Nodes,skip_cases,[Dir,Suite,Cs,Cmt],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
add_tests([{skip_cases,Node,Dir,Suite,Cs,Cmt}|Ts],Spec) ->
Tests = Spec#testspec.tests,
Tests1 = skip_cases(ref2node(Node,Spec#testspec.nodes),
ref2dir(Dir,Spec),
Suite,Cs,Cmt,Tests,Spec#testspec.merge_tests),
add_tests(Ts,Spec#testspec{tests=Tests1});
%% --- various configuration terms ---
add_tests([{config,Nodes,CfgDir,Files}|Ts],Spec) when is_list(Nodes);
Nodes == all_nodes ->
add_tests([{config,Nodes,{CfgDir,Files}}|Ts],Spec);
add_tests([{config,Node,CfgDir,FileOrFiles}|Ts],Spec) ->
add_tests([{config,Node,{CfgDir,FileOrFiles}}|Ts],Spec);
add_tests([{config,CfgDir=[Ch|_],Files}|Ts],Spec) when is_integer(Ch) ->
add_tests([{config,all_nodes,{CfgDir,Files}}|Ts],Spec);
add_tests([{event_handler,Nodes,Hs,Args}|Ts],Spec) when is_list(Nodes);
Nodes == all_nodes ->
add_tests([{event_handler,Nodes,{Hs,Args}}|Ts],Spec);
add_tests([{event_handler,Node,HOrHs,Args}|Ts],Spec) ->
add_tests([{event_handler,Node,{HOrHs,Args}}|Ts],Spec);
add_tests([{enable_builtin_hooks,Bool}|Ts],Spec) ->
add_tests(Ts, Spec#testspec{enable_builtin_hooks = Bool});
add_tests([{release_shell,Bool}|Ts],Spec) ->
add_tests(Ts, Spec#testspec{release_shell = Bool});
%% --- handled/errors ---
add_tests([{define,_,_}|Ts],Spec) -> % handled
add_tests(Ts,Spec);
add_tests([{alias,_,_}|Ts],Spec) -> % handled
add_tests(Ts,Spec);
add_tests([{node,_,_}|Ts],Spec) -> % handled
add_tests(Ts,Spec);
add_tests([{merge_tests, _} | Ts], Spec) -> % handled
add_tests(Ts,Spec);
%% --------------------------------------------------
%% / \
%% | General add_tests/2 clauses below will work for |
%% | most test spec configuration terms |
%% \ /
%% --------------------------------------------------
%% create one test entry per known node and reinsert
add_tests([Term={Tag,all_nodes,Data}|Ts],Spec) ->
case check_term(Term) of
valid ->
Tests = [{Tag,Node,Data} || Node <- list_nodes(Spec),
should_be_added(Tag,Node,Data,Spec)],
add_tests(Tests++Ts,Spec);
invalid -> % ignore term
add_tests(Ts,Spec)
end;
%% create one test entry per node in Nodes and reinsert
add_tests([{Tag,[],Data}|Ts],Spec) ->
add_tests([{Tag,all_nodes,Data}|Ts],Spec);
add_tests([{Tag,String=[Ch|_],Data}|Ts],Spec) when is_integer(Ch) ->
add_tests([{Tag,all_nodes,{String,Data}}|Ts],Spec);
add_tests([{Tag,NodesOrOther,Data}|Ts],Spec) when is_list(NodesOrOther) ->
case lists:all(fun(Test) -> is_node(Test,Spec#testspec.nodes)
end, NodesOrOther) of
true ->
Ts1 = separate(NodesOrOther,Tag,[Data],Ts,Spec#testspec.nodes),
add_tests(Ts1,Spec);
false ->
add_tests([{Tag,all_nodes,{NodesOrOther,Data}}|Ts],Spec)
end;
%% update data for testspec term of type Tag
add_tests([Term={Tag,NodeOrOther,Data}|Ts],Spec) ->
case is_node(NodeOrOther,Spec#testspec.nodes) of
true ->
case check_term(Term) of
valid ->
Node = ref2node(NodeOrOther,Spec#testspec.nodes),
NodeIxData =
update_recorded(Tag,Node,Spec) ++
handle_data(Tag,Node,Data,Spec),
add_tests(Ts,mod_field(Spec,Tag,NodeIxData));
invalid -> % ignore term
add_tests(Ts,Spec)
end;
false ->
add_tests([{Tag,all_nodes,{NodeOrOther,Data}}|Ts],Spec)
end;
%% this test should be added for all known nodes
add_tests([Term={Tag,Data}|Ts],Spec) ->
case check_term(Term) of
valid ->
add_tests([{Tag,all_nodes,Data}|Ts],Spec);
invalid ->
add_tests(Ts,Spec)
end;
%% some other data than a tuple
add_tests([Other|Ts],Spec) ->
case get(relaxed) of
true ->
add_tests(Ts,Spec);
false ->
throw({error,{undefined_term_in_spec,Other}})
end;
add_tests([],Spec) -> % done
Spec.
%% check if it's a CT term that has bad format or if the user seems to
%% have added something of his/her own, which we'll let pass if relaxed
%% mode is enabled.
check_term(Term) ->
Size = size(Term),
[Name|_] = tuple_to_list(Term),
Valid = valid_terms(),
case lists:member({Name,Size},Valid) of
true ->
valid;
false ->
case lists:keymember(Name,1,Valid) of
true -> % halt
throw({error,{bad_term_in_spec,Term}});
false -> % ignore
case get(relaxed) of
true ->
%% warn if name resembles a CT term
case resembles_ct_term(Name,size(Term)) of
true ->
io:format("~nSuspicious term, "
"please check:~n"
"~p~n", [Term]),
invalid;
false ->
invalid
end;
false ->
throw({error,{undefined_term_in_spec,Term}})
end
end
end.
%% specific data handling before saving in testspec record, e.g.
%% converting relative paths to absolute for directories and files
%% (introduce a clause *only* if the data value needs processing)
handle_data(logdir,Node,Dir,Spec) ->
[{Node,ref2dir(Dir,Spec)}];
handle_data(cover,Node,File,Spec) ->
[{Node,get_absfile(File,Spec)}];
handle_data(include,Node,Dirs=[D|_],Spec) when is_list(D) ->
[{Node,ref2dir(Dir,Spec)} || Dir <- Dirs];
handle_data(include,Node,Dir=[Ch|_],Spec) when is_integer(Ch) ->
handle_data(include,Node,[Dir],Spec);
handle_data(config,Node,File=[Ch|_],Spec) when is_integer(Ch) ->
handle_data(config,Node,[File],Spec);
handle_data(config,Node,{CfgDir,File=[Ch|_]},Spec) when is_integer(Ch) ->
handle_data(config,Node,{CfgDir,[File]},Spec);
handle_data(config,Node,Files=[F|_],Spec) when is_list(F) ->
[{Node,get_absfile(File,Spec)} || File <- Files];
handle_data(config,Node,{CfgDir,Files=[F|_]},Spec) when is_list(F) ->
[{Node,filename:join(ref2dir(CfgDir,Spec),File)} || File <- Files];
handle_data(userconfig,Node,CBs,Spec) when is_list(CBs) ->
[{Node,{Callback,get_absfile(Callback,Config,Spec)}} ||
{Callback,Config} <- CBs];
handle_data(userconfig,Node,CB,Spec) when is_tuple(CB) ->
handle_data(userconfig,Node,[CB],Spec);
handle_data(event_handler,Node,H,Spec) when is_atom(H) ->
handle_data(event_handler,Node,{[H],[]},Spec);
handle_data(event_handler,Node,{H,Args},Spec) when is_atom(H) ->
handle_data(event_handler,Node,{[H],Args},Spec);
handle_data(event_handler,Node,Hs,_Spec) when is_list(Hs) ->
[{Node,EvH,[]} || EvH <- Hs];
handle_data(event_handler,Node,{Hs,Args},_Spec) when is_list(Hs) ->
[{Node,EvH,Args} || EvH <- Hs];
handle_data(ct_hooks,Node,Hooks,_Spec) when is_list(Hooks) ->
[{Node,Hook} || Hook <- Hooks ];
handle_data(ct_hooks,Node,Hook,_Spec) ->
[{Node,Hook}];
handle_data(stylesheet,Node,CSSFile,Spec) ->
[{Node,get_absfile(CSSFile,Spec)}];
handle_data(verbosity,Node,VLvls,_Spec) when is_integer(VLvls) ->
[{Node,[{'$unspecified',VLvls}]}];
handle_data(verbosity,Node,VLvls,_Spec) when is_list(VLvls) ->
VLvls1 = lists:map(fun(VLvl = {_Cat,_Lvl}) -> VLvl;
(Lvl) -> {'$unspecified',Lvl} end, VLvls),
[{Node,VLvls1}];
handle_data(silent_connections,Node,all,_Spec) ->
[{Node,[all]}];
handle_data(silent_connections,Node,Conn,_Spec) when is_atom(Conn) ->
[{Node,[Conn]}];
handle_data(silent_connections,Node,Conns,_Spec) ->
[{Node,Conns}];
handle_data(_Tag,Node,Data,_Spec) ->
[{Node,Data}].
%% check if duplicates should be saved or not
should_be_added(Tag,Node,_Data,Spec) ->
if
%% list terms *without* possible duplicates here
Tag == logdir; Tag == logopts;
Tag == basic_html; Tag == label;
Tag == auto_compile; Tag == stylesheet;
Tag == verbosity; Tag == silent_connections ->
lists:keymember(ref2node(Node,Spec#testspec.nodes),1,
read_field(Spec,Tag)) == false;
%% for terms *with* possible duplicates
true ->
true
end.
%% check if previous elements for Node should be deleted
update_recorded(Tag,Node,Spec) ->
if Tag == config; Tag == userconfig; Tag == event_handler;
Tag == ct_hooks; Tag == include ->
read_field(Spec,Tag);
true ->
%% delete previous value for Tag
lists:keydelete(Node,1,read_field(Spec,Tag))
end.
%% create one test term per node
separate(Nodes,Tag,Data,Tests,Refs) ->
Separated = separate(Nodes,Tag,Data,Refs),
Separated ++ Tests.
separate([N|Ns],Tag,Data,Refs) ->
[list_to_tuple([Tag,ref2node(N,Refs)|Data])|separate(Ns,Tag,Data,Refs)];
separate([],_,_,_) ->
[].
%% read the value for FieldName in record Rec#testspec
read_field(Rec, FieldName) ->
catch lists:foldl(fun(F, Pos) when F == FieldName ->
throw(element(Pos, Rec));
(_,Pos) ->
Pos+1
end,2,?testspec_fields).
%% modify the value for FieldName in record Rec#testspec
mod_field(Rec, FieldName, NewVal) ->
[_testspec|RecList] = tuple_to_list(Rec),
RecList1 =
(catch lists:foldl(fun(F, {Prev,[_OldVal|Rest]}) when F == FieldName ->
throw(lists:reverse(Prev) ++ [NewVal|Rest]);
(_,{Prev,[Field|Rest]}) ->
{[Field|Prev],Rest}
end,{[],RecList},?testspec_fields)),
list_to_tuple([testspec|RecList1]).
%% Representation:
%% {{Node,Dir},[{Suite1,[GrOrCase11,GrOrCase12,...]},
%% {Suite2,[GrOrCase21,GrOrCase22,...]},...]}
%% {{Node,Dir},[{Suite1,{skip,Cmt}},
%% {Suite2,[{GrOrCase21,{skip,Cmt}},GrOrCase22,...]},...]}
%% GrOrCase = {GroupSpec,[Case1,Case2,...]} | Case
%% GroupSpec = {GroupName,OverrideProps} |
%% {GroupName,OverrideProps,SubGroupSpec}
%% OverrideProps = Props | default
%% SubGroupSpec = GroupSpec | []
insert_suites(Node,Dir,[S|Ss],Tests, MergeTests) ->
Tests1 = insert_cases(Node,Dir,S,all,Tests,MergeTests),
insert_suites(Node,Dir,Ss,Tests1,MergeTests);
insert_suites(_Node,_Dir,[],Tests,_MergeTests) ->
Tests;
insert_suites(Node,Dir,S,Tests,MergeTests) ->
insert_suites(Node,Dir,[S],Tests,MergeTests).
insert_groups(Node,Dir,Suite,Group,Cases,Tests,MergeTests)
when is_atom(Group); is_tuple(Group) ->
insert_groups(Node,Dir,Suite,[Group],Cases,Tests,MergeTests);
insert_groups(Node,Dir,Suite,Groups,Cases,Tests,false) when
((Cases == all) or is_list(Cases)) and is_list(Groups) ->
Groups1 = [if is_list(Gr) -> % preserve group path
{[Gr],Cases};
true ->
{Gr,Cases} end || Gr <- Groups],
append({{Node,Dir},[{Suite,Groups1}]},Tests);
insert_groups(Node,Dir,Suite,Groups,Cases,Tests,true) when
((Cases == all) or is_list(Cases)) and is_list(Groups) ->
Groups1 = [if is_list(Gr) -> % preserve group path
{[Gr],Cases};
true ->
{Gr,Cases} end || Gr <- Groups],
case lists:keysearch({Node,Dir},1,Tests) of
{value,{{Node,Dir},[{all,_}]}} ->
Tests;
{value,{{Node,Dir},Suites0}} ->
Suites1 = insert_groups1(Suite,Groups1,Suites0),
insert_in_order({{Node,Dir},Suites1},Tests);
false ->
insert_in_order({{Node,Dir},[{Suite,Groups1}]},Tests)
end;
insert_groups(Node,Dir,Suite,Groups,Case,Tests, MergeTests)
when is_atom(Case) ->
Cases = if Case == all -> all; true -> [Case] end,
insert_groups(Node,Dir,Suite,Groups,Cases,Tests, MergeTests).
insert_groups1(_Suite,_Groups,all) ->
all;
insert_groups1(Suite,Groups,Suites0) ->
case lists:keysearch(Suite,1,Suites0) of
{value,{Suite,all}} ->
Suites0;
{value,{Suite,GrAndCases0}} ->
GrAndCases = insert_groups2(Groups,GrAndCases0),
insert_in_order({Suite,GrAndCases},Suites0);
false ->
insert_in_order({Suite,Groups},Suites0)
end.
insert_groups2(_Groups,all) ->
all;
insert_groups2([Group={Gr,Cases}|Groups],GrAndCases) ->
case lists:keysearch(Gr,1,GrAndCases) of
{value,{Gr,all}} ->
GrAndCases;
{value,{Gr,Cases0}} ->
Cases1 = insert_in_order(Cases,Cases0),
insert_groups2(Groups,insert_in_order({Gr,Cases1},GrAndCases));
false ->
insert_groups2(Groups,insert_in_order(Group,GrAndCases))
end;
insert_groups2([],GrAndCases) ->
GrAndCases.
insert_cases(Node,Dir,Suite,Cases,Tests,false) when is_list(Cases) ->
append({{Node,Dir},[{Suite,Cases}]},Tests);
insert_cases(Node,Dir,Suite,Cases,Tests,true) when is_list(Cases) ->
case lists:keysearch({Node,Dir},1,Tests) of
{value,{{Node,Dir},[{all,_}]}} ->
Tests;
{value,{{Node,Dir},Suites0}} ->
Suites1 = insert_cases1(Suite,Cases,Suites0),
insert_in_order({{Node,Dir},Suites1},Tests);
false ->
insert_in_order({{Node,Dir},[{Suite,Cases}]},Tests)
end;
insert_cases(Node,Dir,Suite,Case,Tests,MergeTests) when is_atom(Case) ->
insert_cases(Node,Dir,Suite,[Case],Tests,MergeTests).
insert_cases1(_Suite,_Cases,all) ->
all;
insert_cases1(Suite,Cases,Suites0) ->
case lists:keysearch(Suite,1,Suites0) of
{value,{Suite,all}} ->
Suites0;
{value,{Suite,Cases0}} ->
Cases1 = insert_in_order(Cases,Cases0),
insert_in_order({Suite,Cases1},Suites0);
false ->
insert_in_order({Suite,Cases},Suites0)
end.
skip_suites(Node,Dir,[S|Ss],Cmt,Tests,MergeTests) ->
Tests1 = skip_cases(Node,Dir,S,all,Cmt,Tests,MergeTests),
skip_suites(Node,Dir,Ss,Cmt,Tests1,MergeTests);
skip_suites(_Node,_Dir,[],_Cmt,Tests,_MergeTests) ->
Tests;
skip_suites(Node,Dir,S,Cmt,Tests,MergeTests) ->
skip_suites(Node,Dir,[S],Cmt,Tests,MergeTests).
skip_groups(Node,Dir,Suite,Group,all,Cmt,Tests,MergeTests)
when is_atom(Group) ->
skip_groups(Node,Dir,Suite,[Group],all,Cmt,Tests,MergeTests);
skip_groups(Node,Dir,Suite,Group,Cases,Cmt,Tests,MergeTests)
when is_atom(Group) ->
skip_groups(Node,Dir,Suite,[Group],Cases,Cmt,Tests,MergeTests);
skip_groups(Node,Dir,Suite,Groups,Case,Cmt,Tests,MergeTests)
when is_atom(Case),Case =/= all ->
skip_groups(Node,Dir,Suite,Groups,[Case],Cmt,Tests,MergeTests);
skip_groups(Node,Dir,Suite,Groups,Cases,Cmt,Tests,false) when
((Cases == all) or is_list(Cases)) and is_list(Groups) ->
Suites1 = skip_groups1(Suite,[{Gr,Cases} || Gr <- Groups],Cmt,[]),
append({{Node,Dir},Suites1},Tests);
skip_groups(Node,Dir,Suite,Groups,Cases,Cmt,Tests,true) when
((Cases == all) or is_list(Cases)) and is_list(Groups) ->
Suites =
case lists:keysearch({Node,Dir},1,Tests) of
{value,{{Node,Dir},Suites0}} ->
Suites0;
false ->
[]
end,
Suites1 = skip_groups1(Suite,[{Gr,Cases} || Gr <- Groups],Cmt,Suites),
insert_in_order({{Node,Dir},Suites1},Tests);
skip_groups(Node,Dir,Suite,Groups,Case,Cmt,Tests,MergeTests)
when is_atom(Case) ->
Cases = if Case == all -> all; true -> [Case] end,
skip_groups(Node,Dir,Suite,Groups,Cases,Cmt,Tests,MergeTests).
skip_groups1(Suite,Groups,Cmt,Suites0) ->
SkipGroups = lists:map(fun(Group) ->
{Group,{skip,Cmt}}
end,Groups),
case lists:keysearch(Suite,1,Suites0) of
{value,{Suite,GrAndCases0}} ->
GrAndCases1 = GrAndCases0 ++ SkipGroups,
insert_in_order({Suite,GrAndCases1},Suites0);
false ->
insert_in_order({Suite,SkipGroups},Suites0)
end.
skip_cases(Node,Dir,Suite,Cases,Cmt,Tests,false) when is_list(Cases) ->
Suites1 = skip_cases1(Suite,Cases,Cmt,[]),
append({{Node,Dir},Suites1},Tests);
skip_cases(Node,Dir,Suite,Cases,Cmt,Tests,true) when is_list(Cases) ->
Suites =
case lists:keysearch({Node,Dir},1,Tests) of
{value,{{Node,Dir},Suites0}} ->
Suites0;
false ->
[]
end,
Suites1 = skip_cases1(Suite,Cases,Cmt,Suites),
insert_in_order({{Node,Dir},Suites1},Tests);
skip_cases(Node,Dir,Suite,Case,Cmt,Tests,MergeTests) when is_atom(Case) ->
skip_cases(Node,Dir,Suite,[Case],Cmt,Tests,MergeTests).
skip_cases1(Suite,Cases,Cmt,Suites0) ->
SkipCases = lists:map(fun(C) ->
{C,{skip,Cmt}}
end,Cases),
case lists:keysearch(Suite,1,Suites0) of
{value,{Suite,Cases0}} ->
Cases1 = Cases0 ++ SkipCases,
insert_in_order({Suite,Cases1},Suites0);
false ->
insert_in_order({Suite,SkipCases},Suites0)
end.
append(Elem, List) ->
List ++ [Elem].
insert_in_order([E|Es],List) ->
List1 = insert_elem(E,List,[]),
insert_in_order(Es,List1);
insert_in_order([],List) ->
List;
insert_in_order(E,List) ->
insert_elem(E,List,[]).
%% replace an existing entry (same key) or add last in list
insert_elem({Key,_}=E,[{Key,_}|Rest],SoFar) ->
lists:reverse([E|SoFar]) ++ Rest;
insert_elem({E,_},[E|Rest],SoFar) ->
lists:reverse([E|SoFar]) ++ Rest;
insert_elem(E,[E|Rest],SoFar) ->
lists:reverse([E|SoFar]) ++ Rest;
insert_elem(E,[E1|Rest],SoFar) ->
insert_elem(E,Rest,[E1|SoFar]);
insert_elem(E,[],SoFar) ->
lists:reverse([E|SoFar]).
ref2node(all_nodes,_Refs) ->
all_nodes;
ref2node(master,_Refs) ->
master;
ref2node(RefOrNode,Refs) ->
case lists:member($@,atom_to_list(RefOrNode)) of
false -> % a ref
case lists:keysearch(RefOrNode,1,Refs) of
{value,{RefOrNode,Node}} ->
Node;
false ->
throw({error,{noderef_missing,RefOrNode}})
end;
true -> % a node
RefOrNode
end.
ref2dir(Ref,Spec) ->
ref2dir(Ref,Spec#testspec.alias,Spec).
ref2dir(Ref,Refs,Spec) when is_atom(Ref) ->
case lists:keysearch(Ref,1,Refs) of
{value,{Ref,Dir}} ->
get_absdir(Dir,Spec);
false ->
throw({error,{alias_missing,Ref}})
end;
ref2dir(Dir,_,Spec) when is_list(Dir) ->
get_absdir(Dir,Spec);
ref2dir(What,_,_) ->
throw({error,{invalid_directory_name,What}}).
is_node(What,Nodes) when is_atom(What) ->
is_node([What],Nodes);
is_node([master|_],_Nodes) ->
true;
is_node(What={N,H},Nodes) when is_atom(N), is_atom(H) ->
is_node([What],Nodes);
is_node([What|_],Nodes) ->
case lists:keymember(What,1,Nodes) or
lists:keymember(What,2,Nodes) of
true ->
true;
false ->
false
end;
is_node([],_) ->
false.
valid_terms() ->
[
{define,3},
{node,3},
{cover,2},
{cover,3},
{config,2},
{config,3},
{config,4},
{userconfig,2},
{userconfig,3},
{alias,3},
{merge_tests,2},
{logdir,2},
{logdir,3},
{logopts,2},
{logopts,3},
{basic_html,2},
{basic_html,3},
{verbosity,2},
{verbosity,3},
{silent_connections,2},
{silent_connections,3},
{label,2},
{label,3},
{event_handler,2},
{event_handler,3},
{event_handler,4},
{ct_hooks,2},
{ct_hooks,3},
{enable_builtin_hooks,2},
{release_shell,2},
{multiply_timetraps,2},
{multiply_timetraps,3},
{scale_timetraps,2},
{scale_timetraps,3},
{include,2},
{include,3},
{auto_compile,2},
{auto_compile,3},
{stylesheet,2},
{stylesheet,3},
{suites,3},
{suites,4},
{groups,4},
{groups,5},
{groups,6},
{cases,4},
{cases,5},
{skip_suites,4},
{skip_suites,5},
{skip_groups,5},
{skip_groups,6},
{skip_groups,7},
{skip_cases,5},
{skip_cases,6},
{create_priv_dir,2},
{create_priv_dir,3}
].
%% this function "guesses" if the user has misspelled a term name
resembles_ct_term(Name,Size) when is_atom(Name) ->
resembles_ct_term2(atom_to_list(Name),Size);
resembles_ct_term(_Name,_) ->
false.
resembles_ct_term2(Name,Size) when length(Name) > 3 ->
CTTerms = [{atom_to_list(Tag),Sz} || {Tag,Sz} <- valid_terms()],
compare_names(Name,Size,CTTerms);
resembles_ct_term2(_,_) ->
false.
compare_names(Name,Size,[{Term,Sz}|Ts]) ->
if abs(Size-Sz) > 0 ->
compare_names(Name,Size,Ts);
true ->
Diff = abs(length(Name)-length(Term)),
if Diff > 1 ->
compare_names(Name,Size,Ts);
true ->
Common = common_letters(Name,Term,0),
Bad = abs(length(Name)-Common),
if Bad > 2 ->
compare_names(Name,Size,Ts);
true ->
true
end
end
end;
compare_names(_,_,[]) ->
false.
common_letters(_,[],Count) ->
Count;
common_letters([L|Ls],Term,Count) ->
case lists:member(L,Term) of
true ->
Term1 = lists:delete(L,Term),
common_letters(Ls,Term1,Count+1);
false ->
common_letters(Ls,Term,Count)
end;
common_letters([],_,Count) ->
Count.