On 17 May 2015 at 19:05, Antti Kantee <[email protected]> wrote: > On 17/05/15 17:41, Justin Cormack wrote: >> >> There is another problem - unlike the __aeabi_* symbols and the >> __sync_ ones, these do not seem to be standard, so Linux on arm does >> not supply them as far as I can see. So programs linked against >> rump+linux libc rather than rump_netbsd libc are going to have issues >> if they need these symbols - we havent run into the issue as nothing >> has actually needed them yet. > > > I'm not 100% sure what "these" refers to, but at any rate, in case __symbols > are used internally (i.e. _NOT_ via the toolchain), they should be used via > an internal alias instead of via the toolchain alias. > > ... which is not to say I'm denying a problem exists, just saying that if > one exists, fixing it requires rototilling code to conform to what the code > already should be conforming to. >
Actually, it does seem to be that the aliases are being defined in the rump kernel as well as libc, and they are not being used in the kernel under the aliases, so I think the patch is correct and the usage is correct. There are a couple of other cases I need to look at which are not in the patch yet. Justin
