|
From: Paul P. <pp...@mo...> - 2004-03-05 16:52:08
|
Hello, I am curious if any of the developers have looked into using valgrind to gather code coverage information? When valgrind executes a basic block it seems like it would be almost trivial to add this information to some kind of coverage file, similar to the .da files created by gcov(the gnu code coverage tool). I think, the more difficult part would be mapping the basic block execution info back to the actual source code, especially in the case of Java, or other translated/interpreted languages. Please let me know if anybody has input into the feasablity of this and why it would or wouldn't work. Thanks Paul Pearcy Motorola Urbana Design Center |
|
From: Nicholas N. <nj...@ca...> - 2004-03-08 15:35:26
|
On Fri, 5 Mar 2004, Paul Pearcy wrote: > I am curious if any of the developers have looked into using valgrind to > gather code coverage information? When valgrind executes a basic block it > seems like it would be almost trivial to add this information to some kind > of coverage file, similar to the .da files created by gcov(the gnu code > coverage tool). I think, the more difficult part would be mapping the basic > block execution info back to the actual source code, especially in the case > of Java, or other translated/interpreted languages. Please let me know if > anybody has input into the feasablity of this and why it would or wouldn't > work. See vgprof, from valgrind.kde.org/related.html. Mapping back to source code in Java would be tricky, I think. Well, Valgrind currently reads stabs and DWARF2 debug info in order to map from instructions back to source code; if Java used them too that shouldn't pose problems, but I don't know if Java uses them. N |