On Wed, Mar 12, 2014 at 10:54 PM, Antti Kantee <[email protected]> wrote: > On 12/03/14 22:33, Justin Cormack wrote: >> On Wed, Mar 12, 2014 at 9:36 PM, Antti Kantee <[email protected]> wrote: >>> On 12/03/14 21:28, Justin Cormack wrote: >>> I'm not sure if I mind mixing static and dynamic rump kernel libs >>> failing. However, with the current way of things, I think it will fail >>> extremely non-obviously, as in link successfully, but fail to recognize >>> components/modules. That I do mind. At least now things honestly fail >>> to link. >> >> Yes I can't immediately think of any way to make it fail correctly in >> that situation. > > Hmm, what if rumpuser_dl.c just provided weak aliases for dlinfo() and > dlsym()? I think it could work, at least unless weak aliases now start > shooting us in the foot ("dance, dance!"). Earlier I forgot that the > kernel side of things already should correctly handle components/modules > in the static linking situation due to memories of hacks in an even > earlier situation when it didn't.
Should work I think - just checked and glibc which is somewhat perverse about dl stuff and static linking doesnt actually have dlsym in libc.a. Yes glibc allows dlopen in staticly linked binaries, as it uses it for its own internal stuff. It might even work sometimes, allegedly. Justin ------------------------------------------------------------------------------ Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/13534_NeoTech _______________________________________________ rumpkernel-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/rumpkernel-users
