immediate_mode execution order?

[michaelquigley]

In reading the documention, it sounds like the following code example:

int main(int argc, char** argv) {
    try {
        CLI::App app("immediate_callback example");
        app.require_subcommand(1, 1);
        app.failure_message(CLI::FailureMessage::help);

        app.callback([&](){
            std::cout << "main callback" << std::endl;
        })->immediate_callback(true);

        app.add_subcommand("hello")->callback([&](){
            std::cout << "hello callback" << std::endl;
        })->immediate_callback(false);

        CLI11_PARSE(app, argc, argv);

        return 0;
    }
    catch(const std::runtime_error& re) {
        std::cerr << "exception [" << typeid(re).name() << "]: " << re.what() << std::endl;
    }
}

...should result in showing:

main callback
hello callback

...when I execute example hello, but instead, I'm getting:

hello callback
main callback

What am I missing? Or, is this a bug?

[phlptp]

There were some fixes related to immediate_callback and the main app in PR #292. So you need to make sure you are on the latest master vs the latest release.

On the latest master I ran a test

app.require_subcommand(1, 1);
    std::vector<int> v;
    app.callback([&v]() { v.push_back(1); })->immediate_callback(true);

    auto sub = app.add_subcommand("hello")->callback([&v]() { v.push_back(2); })->immediate_callback(false);
    args = {"hello"};
    run();
    // immediate_callback inherited
    ASSERT_EQ(v.size(), 2u);
    EXPECT_EQ(v[0], 1);
    EXPECT_EQ(v[1], 2);
    v.clear();
    sub->immediate_callback(true);
    run();
    // immediate_callback is now triggered for the main first
    ASSERT_EQ(v.size(), 2u);
    EXPECT_EQ(v[0], 2);
    EXPECT_EQ(v[1], 1);

And this has the expected behavior.
In the "hopefully not to distant future" we are reworking the callbacks a little to add a third one which will hopefully make it clearer which callbacks get executed when, at which point immediate_callback will be deprecated.

[michaelquigley]

Makes perfect sense. How can I build a single CLI11.hpp from master?

[michaelquigley]

Found the CLI11_SINGLE_FILE cmake option. Should be all set.