449532 - Measure code coverage of js shell tests

Status: RESOLVED DUPLICATE of bug 1301170

(Core :: JavaScript Engine, defect, P2)

Description

[Robert Sayre]

This might be difficult to get for the whole browser, but we can definitely get a shell build.

Comment 1

[Jason Orendorff [:jorendorff]]

I could take this.

Comment 2

[Robert Sayre]

I have this working.

Comment 3

[Paul Biggar]

I'd like to resurrect this. I'm assuming the work here is something like adding gcov to CFLAGS? If not:

- Is there a patch to make this work?
- Is this in the build system somewhere?
- Have we automated tracking/testing/arewecoveredyet.com for this?
  - if so, where do the numbers go?
  - if so, were problems like the ones reported in bug 635155 and bug 621096 detected or detectable?

Comment 4

[(not currently active) Ted Mielczarek]

We have existing weekly whole-browser code coverage builds, they're apparently failing to compile right now (in JS, conveniently enough):
http://tinderbox.mozilla.org/showlog.cgi?log=Testing/1299323050.1299324259.9880.gz

I'd have to poke to find out where the results were, but we were producing lcov output from having run all the unittest suites.

Comment 5

[(not currently active) Ted Mielczarek]

The builds are successful and producing output now, but they're timing out running "make check" / xpcshell-tests, I filed bug 657631 on that. The output isn't really going anywhere useful at the moment, there's a .tar.bz2 here with the lcov output:
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/experimental/codecoverage/

Comment 6

[(not currently active) Ted Mielczarek]

See also bug 657639.

Comment 7

[André Bargull [:anba]]

Looks like code coverage is currently tracked in bug 1301170.

Comment 8

[Marco Castelluccio [:marco]]

We are currently running the coverage buily daily. You can see a report of the coverage at https://codecov.io/gh/marco-c/gecko-dev.