Re: musl and gcc-4.8 (sabotage)

[Szabolcs Nagy <nsz@port70.net>]

* Rich Felker <dalias@aerifal.cx> [2013-03-28 20:07:58 -0400]:
> On Fri, Mar 29, 2013 at 12:58:43AM +0100, Daniel Cegie??ka wrote:
> > game over. libsanitizer requires pthread_getattr_np() but this
> > function isn't implemented in musl.
> 
> There's almost certainly a way to disable libsanitizer, no? It seems
> like gcc folks keep inventing new low-quality alternatives to
> valgrind...

asan+tsan are useful tools, but now looking into the code
it has unreliable c++ code and invalid assumptions about
the c runtime internals

but i still think it may worth trying to port it to musl
maybe musl can provide some hooks for similar tools
(weak symbol callbacks at the right places)

asan hooks into thread creation, so they should be able
to get the bottom of the stack and they can assume a
guard page at the end which can be found by vmfill or
parsing /proc smaps which they already do

tsan tries to get the tls as well with
void *get_tls_static_info_ptr = dlsym(RTLD_NEXT, "_dl_get_tls_static_info");
and then copy this onto a function pointer with
__attribute__((regparm(3), stdcall)) on i386

the rest should be portable