Re: [PATCH] Linux: Add execveat system call wrapper

[Joseph Myers <joseph@codesourcery.com>]

On Tue, 28 Apr 2020, Florian Weimer wrote:

> * Alexandra Hájková via Libc-alpha:
> 
> > diff --git a/sysdeps/mach/hurd/i386/libc.abilist b/sysdeps/mach/hurd/i386/libc.abilist
> > index 60696d827f..571c281b06 100644
> > --- a/sysdeps/mach/hurd/i386/libc.abilist
> > +++ b/sysdeps/mach/hurd/i386/libc.abilist
> > @@ -2182,6 +2182,7 @@ GLIBC_2.3.4 xdr_quad_t F
> >  GLIBC_2.3.4 xdr_u_quad_t F
> >  GLIBC_2.30 twalk_r F
> >  GLIBC_2.32 mach_print F
> > +GLIBC_2.32 execveat F
> >  GLIBC_2.32 thrd_current F
> >  GLIBC_2.32 thrd_equal F
> >  GLIBC_2.32 thrd_sleep F
> > diff --git a/sysdeps/unix/sysv/linux/Makefile b/sysdeps/unix/sysv/linux/Makefile
> 
> This is not correct because you are not adding execveat for Hurd.  You
> can just remove that change.

First, the commit message should discuss the rationale for this being a 
Linux-specific interface in glibc, as opposed to an OS-independent 
interface with an ENOSYS stub for OSes that don't implement it.  The 
interface looks quite generic like the POSIX *at interfaces, so there's a 
reasonable case for it being an OS-independent interface (which doesn't 
require adding an implementation for Hurd, an ENOSYS stub in a suitable 
OS-independent directory would suffice, along with the declaration, test 
etc. being OS-independent).

-- 
Joseph S. Myers
joseph@codesourcery.com