boost warnings when using clang

When I compile apps that use boost classes (boost 1.49), I get a number of compiler warnings. I looked at one of them and it looks like a legitimate warning as they are doing an operation that results in a long being being converted to an int on 64 bit. I'm a bit puzzled how such a problem would not get fixed during the development cycle. Perhaps clang catches more issues than gcc or perhaps their macro usage causes the correct behavior on gcc but not on clang.

Does anyone on the clang team communicate with the boost people to let them know when clang reports a warning that might be going unseen with gcc or VC++? If not, perhaps I should report them as their presence in my build logs bugs me.


In terms of the main public LLVM buildbots, I don't believe we have
boost running on any of them. It's sort of that tricky intersection -
whether Clang should be responsible/devote resources to validating
Boost or the other way around. Ideally we want to make sure we can
compile Boost code, but finding bugs in Boost isn't really where we'd
be focused.

Basically I'd expect this to be Boost's responsibility to use Clang to
find bugs if they want to. (or their users to contribute patches/file
bugs/etc found with Clang)

- David

If you see issues which seem like bugs in boost, please report them
(it looks like has their bug
reporting guidelines). We have a buildbot running the boost
regression tests, but we wouldn't notice any compiler warnings coming
out of it as long as the tests pass.


in the specific instance I looked at (in the date time library), the conversion would not fail unless the test was code input something like 4294967296 hours.

test was codeD TO input *