|
From: Brad H. <br...@fr...> - 2008-01-05 03:46:55
|
There is a bit of interest[0] in running a test coverage BoF session at linux.conf.au 2008[1] (Melbourne, Australia). Are any of the valgrind developers going to be at the conference? Brad [0] http://lists.mel8ourne.org/pipermail/chat/2008-January/thread.html#37 [1] http://linux.conf.au |
|
From: Nicholas N. <nj...@cs...> - 2008-01-05 05:16:16
|
On Sat, 5 Jan 2008, Brad Hards wrote: > There is a bit of interest[0] in running a test coverage BoF session at > linux.conf.au 2008[1] (Melbourne, Australia). > > Are any of the valgrind developers going to be at the conference? I work on the Melbourne Uni campus so it would be very easy for me to go. I didn't realise it was in Melbourne this year. Unfortunately, it seems like registrations are full -- http://www.itwire.com/content/view/15889/1090/ says they sold out of registrations in late December. http://linux.conf.au/register doesn't say they've sold out, but I also can't find a link that would let a new registration occur. As for test coverage with Valgrind, I've tried writing a tool that does this (as has at least one other person). The problem I encountered was that Valgrind works at the binary level, whereas test coverage is all about a source level concept (lines). You can't get percentages when working at the binary level, except by using debug info, and I found GCC's output debug info to be sufficiently unreliable that this didn't work well. Nick |
|
From: Benjamin M. <be...@me...> - 2008-01-27 00:50:40
|
On Jan 5, 2008, at 6:15 AM, Nicholas Nethercote wrote: > On Sat, 5 Jan 2008, Brad Hards wrote: > >> There is a bit of interest[0] in running a test coverage BoF >> session at >> linux.conf.au 2008[1] (Melbourne, Australia). >> >> Are any of the valgrind developers going to be at the conference? > > I work on the Melbourne Uni campus so it would be very easy for me > to go. > I didn't realise it was in Melbourne this year. > > Unfortunately, it seems like registrations are full -- > http://www.itwire.com/content/view/15889/1090/ says they sold out of > registrations in late December. http://linux.conf.au/register > doesn't say > they've sold out, but I also can't find a link that would let a new > registration occur. > > As for test coverage with Valgrind, I've tried writing a tool that > does this > (as has at least one other person). The problem I encountered was > that > Valgrind works at the binary level, whereas test coverage is all > about a > source level concept (lines). You can't get percentages when > working at the > binary level, except by using debug info, and I found GCC's output > debug > info to be sufficiently unreliable that this didn't work well. > > Nick I solved that by combining callgrind's data with rpp which for my C++ source is able to give me all the coverage data I want. http://benjamin-meyer.blogspot.com/2007/12/valgrind-callgrind-tools- part-3-code.html -Benjamin Meyer |
|
From: Nicholas N. <nj...@cs...> - 2008-01-27 22:56:33
|
On Sun, 27 Jan 2008, Benjamin Meyer wrote: > I solved that by combining callgrind's data with rpp which for my C++ source > is able to give me all the coverage data I want. > > http://benjamin-meyer.blogspot.com/2007/12/valgrind-callgrind-tools- > part-3-code.html What is 'rpp'? I Googled but couldn't find it. Is it a C++ parser? In the comments on your post there are complaints that the Valgrind docs encourage people to write coverage tools but it's been done several times before. First, I believe this is no longer in the docs -- at least, I just looked and couldn't find it in the section linked to or any other sections. Second, any suggestions you have for improvements are much more likely to be addressed if you report them to this list or Bugzilla rather than writing them in your blog. Nick |