Re: [DynInst_API:] New release (8.1) of Dyninst Tool Suite

[Josh Stone <jistone@xxxxxxxxxx>]

On 03/04/2013 02:16 AM, Barton Miller wrote:
> We are pleased to announce a new release of Dyninst 8.1 and its component
> tool kits. This release includes ParseAPI 8.1, SymtabAPI 8.1, InstructionAPI-`
> 8.1, StackwalkerAPI 8.1,  ProcControl 8.1, PatchAPI 8.1, and DynC_API 8.1.

Great!

I don't see new tags in either dyninst.git or docs.git - an oversight?

> IMPORTANT DEVELOPER NOTE:
> As of 8.1, Dyninst and its components are compiled with C++11x support. All
> mutator code MUST be compiled with C++11x turned on. For GCC 4.3 and later,
> this is done by including the -std=c++0x flag. For Visual Studio, this is on
> by default. Linking code compiled without C++11x support may cause undefined
> behavior. 

I wasn't aware of this new mutator requirement, even as I blissfully
tested pre-8.1.  Perhaps I was silently hitting ABI issues and just
getting lucky not to hit anything major...

But since this is mandatory, did you consider enforcing it in the
headers?  GCC before 4.7 didn't set __cplusplus correctly, but did have
an experimental flag, so something like this should work portably:

#if !((__cplusplus >= 201103L) || defined(__GXX_EXPERIMENTAL_CXX0X__))
#error "Dyninst requires C++11!"
#endif

If you're ok with that check, but don't want to redo release tarballs, I
may just patch that into the headers for our RPMs.

> And thank you to Red Hat for hosting the Linux RPMs.

Well, I haven't build 8.1 RPMs just yet. :)

I'll work on getting them soon though.  For Fedora I need to coordinate
it with a SystemTap update though, so the distro stays consistent.  I
also need to find out if it's ok to bump the library soname in F18.  If
not, then only rawhide will get this update officially, but I can still
put new builds in my personal repo as I did before for F17 and EL6.

In any case, I'll send a followup when those builds are available.


Josh