Hidden tags specified as "[.foo]" aren't matched

[echuber2]

Sorry, I hit "enter" and submitted this one too soon. Edited:

Describe the bug
If you have a test case that had a hidden tag specified as [.foo], in the past, you could specify it on the executable as in:

./program [.foo]

However, now it reports no matching test. Other forms of specifying the tag do work, such as:

./program [.]
./program [foo]

In much older versions, I think the last one here actually hadn't worked, though.

Expected behavior
./program [.foo] should launch the hidden [.foo] (or [.][foo]) test case, as in some previous versions of Catch.

Reproduction steps
Using catch.hpp v2.10.2.
Create foo.cpp:

#define CATCH_CONFIG_MAIN

#include "catch.hpp"

TEST_CASE("Test", "[.foo]") {

  REQUIRE(true);

}

Compiling and testing:

$ g++ --std=c++11 foo.cpp -o foo
$ ls
catch.hpp  foo  foo.cpp
$ ./foo
===============================================================================
No tests ran

$ ./foo [.]
Filters: [.]
===============================================================================
All tests passed (1 assertion in 1 test case)

$ ./foo [foo]
Filters: [foo]
===============================================================================
All tests passed (1 assertion in 1 test case)

$ ./foo [.foo]
Filters: [.foo]
No test cases matched '[.foo]'
===============================================================================
No tests ran

Platform information:

• OS: Linux
• Compiler+version: g++ 5.4.0 and/or 7.2, clang 3.8
• Catch version: v2.10.2

Used to work as expected in Catch version v2.2.3.

[echuber2]

(Sorry, I hit "enter" and submitted this one too soon. Edited.)

[horenmar]

Right, the problem is that we recently started normalizing [.foo] into [.][foo] as was intended according to the documentation, but which we did not do for a long time. As you noted, this meant that [foo] would not call test cases tagged as [.foo].

I will not revert the changes, but I think we could make the test spec parser also normalize [.foo] into [.][foo]...

[echuber2]

Thank you. I can adjust to these changes, but I'd previously distributed some scripts to students that tried to invoke it the "[.foo]" way from the shell, which has caused me a problem. It's not disastrous but it may affect some existing system configurations.