Number of JS files in crawler reports looks too small

[marco-c]

The number of files is too small to be true, there must be something wrong.
@rhcu could you check what the number of files is before running the lcov rewriter?

[rhcu]

Ok, I'll check this.

[marco-c]

Basically you just need to pass jsvm_dir to grcov (

coverage-crawler/crawler.py

Line 263 in faa178e

jsvm_files = [os.path.join(jsvm_dir, e) for e in os.listdir(jsvm_dir)]

) to generate a JSON, and then parse it.

[rhcu]

I tried running the crawler with different websites, and the max number in output.json was ~110 JS files. However, still no such files after diffing.

[marco-c]

There's likely a problem somewhere, either in the coverage generation, or in the lcov rewriter or in grcov. I bet on the lcov rewriter (so checking what the output would be before the rewriting might help).

[rhcu]

I tried running the crawler with different websites, and the max number in output.json was ~110 JS files. However, still no such files after diffing.

This was an output before rewriting.

[marco-c]

Can you upload the list here?

[rhcu]

https://pastebin.com/ee3aUCVw - 104 files

[marco-c]

Can you upload the list before rewriting and the list after rewriting? So we can see which files get removed.

[rhcu]

https://pastebin.com/9TMb4bTM

The list with source file names before rewriting (content of .info files from jsvm_dir) and after - from jsvm_output_file. All files that have prefixes chrome and resource://gre/modules/ are present in the file after rewriting. Other files are not. There are source files with similar names that don't end with .js/.jsm before rewriting. I've attached one of the files that have such kind of source files.

1531736203-11705-0.info.zip

[marco-c]

OK, so it looks like the lcov rewriter is behaving correctly. The next thing to look at would be the logs from Firefox.
Two options: 1) try setting webdriver.firefox.logfile; 2) try running the Firefox build manually, after setting the necessary env variables (especially JS_CODE_COVERAGE_OUTPUT_DIR) and then use grcov to parse the info files in the JS_CODE_COVERAGE_OUTPUT_DIR directory

In the case 2 the logs should be outputted to the console. Parsing the info files is not actually required (but it's just a sanity check).

[rhcu]

I enabled trace logs from geckodriver, and have information about actions stored in geckodriver.log. Does it have the same effect as setting webdriver.firefox.logfile?
I see there all actions performed by the crawler with 200 OK status in this file.

[marco-c]

I don't think so, but can you upload the file so I can check?

[rhcu]

geckodriver.log.zip