I was trying to generate a code coverage report from trace1.json. The trace1.json was not generated by gcovr, it was generated using Lauterbach software from the real hardware trace data. From Lauterbach spec, it has the ability to export coverage information about functions and lines to a file in JSON format compatible to Gcov. So after I got the JSON file and tried to use gcovr to generate the code coverage report: gcovr --add-tracefile result.json --html-details result.html --verbose
I got an empty report and the gcovr log shows "Gathered coveraged data for 0 files". So I'm wondering after I get the Json file, do I still need to compile the source with --coverage? since even I compile the source with coverage flags, the executable is running on another real hardware, which will not able to collect any gcda.
As of gcovr 5.2, there is no support for the gcov JSON format. Gcovr's internal JSON format is closely based on the gcov JSON format, but differs in some details. Importantly, gcovr validates that the input JSON document has a matching gcovr/format_version. This means gcovr should die with an error in your scenario, and shouldn't even get to the “Gathered coveraged data for 0 files” message.
This suggests that your result.json is an empty JSON report generated by a previous gcovr --json result.json run, and NOT the trace1.json generated by your Lauterbach tool!
You may be able to write a script to modify the JSON file to conform to gcovr's expected format, but you're on your own there. The JSON format is documented in the Gcovr User Guide, though there's a pending update.
Assuming things work as expected, the GCC --coverage flag is unnecessary. While --coverage is necessary for producing .gcda and .gcno files that are needed by gcov to work, gcovr's --add-tracefile mode only consumes the JSON file (and perhaps the source code files) and no other data.