fix(src): avoid possible infinite loop in LoadAll().

[FedeDP]

Leave at first empty root.

This avoids an infinite loop in case of HandleNextDocument returning true even if the document has errors, basically because no token.type gets matched in SingleDocParser::HandleNode, leading to eventHandler.OnNull(mark, anchor); being called indefinitely at each iteration, pushing a null node to documents vector.

Another solution would be to add a m_scanner.pop(); in the default switch case in SingleDocParser::HandleNode to make sure we always consume the current token.

The behavior causes this issue: falcosecurity/falco#3281

Easily reproducible with a simple c++ example:

#include <yaml-cpp/yaml.h>

int main() {
    static const std::string yml = R"(
,
    )";
    YAML::LoadAll(yml);
    return 0;
}

[jbeder]

Can you add a test (e.g., the code in the PR description)? Something that infinite-loops before this PR but doesn't after.

[FedeDP]

Added the test; please note that if #1318 is merged first, i'll need to rebase this PR on master and fix up the test (that will then throw a ParserException obviously).
Or if this is merged first, i will need to rebase the other ;)

[jbeder]

Merged the other one. Go ahead and rebase this and then ping me.

[FedeDP]

@jbeder done!
Thanks for the quick response!

[FedeDP]

Sorry for the ping @jbeder , any news on this one? :)
Thanks!

[jbeder]

Sorry for the delay. It looks like I forgot to hit "send" on the review itself - github's UI is terrible :(

[jbeder]

This test file is just for the spec tests, ie, examples literally in the spec.

Can you add it to node_test?

[FedeDP]

Done :)
Thanks for the feedback! (and agree about github's UI :/ )

[FedeDP]

I see 2 CI failures that don't seem to be caused by my commits:

/usr/include/gtest/internal/gtest-port.h:279:2: error: #error C++ versions less than C++14 are not supported.
279 | #error C++ versions less than C++14 are not supported.

I think they bumped gtest version and it now requires c++14, therefore all jobs that use deps from system and building against c++11 will fail.

[leogr]

Hey @jbeder

Any update on this?