fix inner_active_ test in variant handler's on_array_end

[Ramya-9353]

Debug build, parse_into into std::variant<X, int> from {"z":[[0]]}:

Assertion failed: (object_depth_ > 0), function on_object_end, file parse_into.hpp, line 1038.

inner_active_ in the variant handler is an int initialised to -1, and every other test of it uses < 0, as do the tuple and described-class handlers that share that convention. on_array_end tests !inner_active_, which is true for alternative 0 and false when no alternative is selected, so while the first alternative is active the closing bracket of its own array is diverted to signal_end instead of being forwarded to it.

Release build, same revision:

variant<vector<vector<int>>,int>   [[1,2]]       accepted, parses to []
variant<vector<vector<int>>,int>   [[1],[2]]     error::exhausted_variants
variant<X,int>                     {"z":[[0]]}   accepted, X default-constructed

The third line is the one that bites. ignoring_handler never receives the array_end that would balance its depth counters, so they wrap, the described-class handler stays pinned on the ignore slot, and seen_ stops being updated; X's three required members are all absent and the document is still accepted. With inner_active_ < 0 the three cases give [[1,2]], [[1],[2]] and error::exhausted_variants.

Tests added to testVariant cover a nested-array alternative and the described-struct case. They fail on develop.

[cppalliance-bot]

An automated preview of the documentation is available at https://1203.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

If more commits are pushed to the pull request, the docs will rebuild at the same URL.

2026-10-01 07:20:00 UTC

[cppalliance-bot]

GCOVR code coverage report https://1203.json.prtest2.cppalliance.org/gcovr/index.html
LCOV code coverage report https://1203.json.prtest2.cppalliance.org/genhtml/index.html
Coverage Diff Report https://1203.json.prtest2.cppalliance.org/diff-report/index.html

Build time: 2026-10-01 07:47:17 UTC

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/1203/pullrequest.html