Error on html report generation

[evil-shrike]

I have a coverage.json file with coverage info extracted from within browser after qunit tests completed (as value of window.coverage). No I'm trying to generate a html report:

Then I execute:

istanbul report html coverage.json 

and get the following error:

C:\...\Client\node_modules\istanbul\lib\report\html.js:156
        structuredText[lineNumber].covered = count > 0 ? 'yes' : 'no';
                                                 ^
TypeError: Cannot set property 'covered' of undefined
    at C:\...\Client\node_modules\istanbul\lib\report\html.js:156:66
    at Array.forEach (native)
    at annotateLines (C:\...\Client\node_modules\istanbul\lib\report\html.js:154:28)
    at HtmlReport.Report.mix.writeDetailPage (C:\...\Client\node_modules\istanbul\lib\report\html.js:370:9)
    at C:\...\Client\node_modules\istanbul\lib\report\html.js:435:26
    at AsyncFileWriter.extend.processFile (C:\...\Client\node_modules\istanbul\lib\util\file-writer.js:93:9)
    at Object.q.process [as _onImmediate] (C:\...\Client\node_modules\istanbul\node_modules\async\lib\async.js:728:21)
    at processImmediate [as _immediateCallback] (timers.js:317:15)

"text" report generation works.

[gotwarlost]

Only the HTML report uses the actual sources of your files for annotating the uncovered lines and branches.

If for any reason your source files have changed between the time they were instrumented and the reporting, you will see this error. So, is it possible that your source files may have changed in the interim?

[evil-shrike]

It's absolutely impossible. I have a grunt build workflow which :

• copy sources into temp dir
• run istanbul to instrument files in this dir:

grunt.file.write(filepath, instrumenter.instrumentSync(String(fs.readFileSync(filepath)), filepath));

• start connect local server with instrumented sources folder mounted
• run qunit tests against local server via phantomjs
• catch window.coverage inside browser and return it to the host (grunt task)
• save reported coverage info into a file (coverage.json)
• run 'istanbul report html'

I added some logging into report/html.js:

Object.keys(lineStats):
1,3,8,22,24,25,26,27,28,29,30,31,33,34,35,36,37,38,39,40,41,42,43,44,45,46,48,49,55,56,57,68,70,72,74,75,77,80,81,83,94,104,106,107,108,109,110,111,114,117,119,124,128,130,131,134
Object.keys(structuredText): 0,1,2,3,4,5,6,7

So structuredText array really doesn't contain values from lineStats.
What did I miss?

[gotwarlost]

Thanks for the details. I think I know what is happening. Basically structured text contains the source code of the instrumented file and not the original source code. It needs the original source code.

The fix is to not instrument files in place. Structure your grunt task so that the file is instrumented from the original path and written to the tmp/ directory. That way the instrumented file in the tmp/ directory correctly refers to the original source code. As it stands now, it is referring to the same file which now contains the instrumented code.

[evil-shrike]

@gotwarlost Thanks a lot. I've done as you described and everything works fine. Cool.

Is it possible to flatten files in html report?
I'd like to have a flatten list of files sorted by coverage percentage.

[gotwarlost]

No, not right now. The HTML report was designed to be 100% compatible with the genhtml program in the lcov package.

That said people have asked for different things:

1. Truly hierarchical reports right now if you have lib/foo and lib/bar directories they appear at the top level as 2 entries rather than a lib/ entry with subdirectories under it.
2. Your ask of a flattened list of all files so that coverage can be inspected for all of them together and sorted

These are good requirements and will need changes to the html report to support additional flags. I'll see what I can do once I get some time to work on this.

[evil-shrike]

Thanks. I'm looking forward for the lib progress, it's awesome!

btw: I published my grunt taks in grunt-croc-qunit plugin (I know it's a yet another grunt plugin for Istanbul/qunit/phantomjs but anyway)

[jnikles]

I also had that error. However, I was using the hbsfy transform for browserify at that time, ignoring the templates folder fixed that issue as well.

[SarahFowler73]

Using mochify-istanbul, I found that anything that I was using transform for would mess up the source mapping. Adding exclude['**/*.html', '**/*.css', and '**.*.json'] completely cleared up the problem for me.