Date: Wed, 16 Jul 2014 19:29:01 -0400 From: Rich Felker <dalias@...c.org> To: Stefan Kristiansson <stefan.kristiansson@...nalahti.fi>, musl@...ts.openwall.com Subject: Re: [PATCH] add or1k (OpenRISC 1000) architecture port On Thu, Jul 17, 2014 at 12:51:59AM +0200, Szabolcs Nagy wrote: > * Stefan Kristiansson <stefan.kristiansson@...nalahti.fi> [2014-07-16 20:55:57 +0300]: > > With the exception of a fenv implementation, the port is fully featured. > > i assume or1k is soft-float only > > (for a working soft-float fenv musl will need cooperation with libgcc) Actually I already asked about this. It has optional hard float, but does not use dedicated float registers, so the ABI is the same regardless of whether you build for a device with/without fpu. Without fpu, fenv presumably won't work, but right now it's not supported anyway. > > +++ b/arch/or1k/bits/syscall.h > > @@ -0,0 +1,523 @@ > .... > > +#define __NR_kcmp 272 > > +#define __NR_finit_module 273 > > +#define __NR_syscalls 274 > > other archs don't define __NR_syscalls for the number of syscalls > (but it does not hurt) > > .... > > +#define SYS_kcmp __NR_kcmp > > +#define SYS_finit_module __NR_finit_module > > +#define SYS_syscalls __NR_syscalls > > the syscall numbers after this are out-of-order > (this will be confusing when new syscalls will be added) > > i'd prefer if __NR_* and SYS_* were in the same order Yes, I found this somewhat confusing too. > (can be fixed after the great commit) Might be nicer to do it first to avoid gratuitous big commits. Of course the upcoming bits refactoring will make big changes here anyway, and at the same time we might arrange to remove the duplicate __NR_/SYS_ macros at the source level and generate one from the other at build time. But that topic is independent of getting or1k merged. Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.