On 05/09/2011 01:20 PM, Mark Wielaard wrote:
I have to agree with roland, that this doesn't really test that
much,
except if you can flip the bits.
Yep, that is true, I've already thrown it away. I just didn't know
how else/better can this be tested.
How are you measuring coverage?
Using the lcov. Here's what I do:
$ ./configure --enable-gcov
$ make -j4 check
$ lcov --directory . --capture --output-file blah.info
$ genhtml blah.info
$ firefox index.html
Here's how the result looks like (current git):
http://people.redhat.com/mpolacek/tmp/eu-lcov/
Which reminds me, do you (or anyone) please have a suggestion what should I test
next?
libebl isn't stable, libelf is quite well-covered, and in libdw the most
important functions are also covered already. Maybe the libdwfl could use some
additional test, for instance core_file.c or link_map.c. But I'm not sure
if it is even worth it. Is it?
Shouldn't these functions be exercised by some the the unstrip
tests?
I don't know. elf_flag{data,ehdr,elf,phdr} are used in update{2,3,4}.c,
but the coverage is only ~40%. Well, it is probably fine as it is.
Marek