On 24/07/2011 22:29, Andrei Alexandrescu wrote:
On 7/24/11 4:06 PM, Robert Clipsham wrote:
Hi all,
I took the liberty of running dmd through clang's static analysis, and
it's turned up quite a few (486) potential bugs in dmd:
http://octarineparrot.com/assets/dmd/
There's bound to be a few false positives in there (please feel free to
report them to the nice folk over at llvm), and the dead code won't
cause any problems (and most of the "dead code" in the backend is
probably in use with dmc or something), but there are still a few
hundred potential crashes in there.
Very interesting. I selected randomly one of each category and found
only false positives though, mostly because the tool does not understand
that assert(p) will subsequently guarantee p is non-null, and assert(0)
terminates the program.
Andrei
That's true for quite a few of the null pointer dereferences, there are
some where there are no assertions though. It might be worth reporting
these false positives to the folk working on it. In the other categories
there are far fewer where they're false positives due to the assertion.
It tends to be the ones with shorter path lengths that exhibit this problem.
--
Robert
http://octarineparrot.com/
