Re: Remove pre-2.4 Linux kernel support

[Roland McGrath <roland at hack dot frob dot com>]

> To confirm:
> 
> /* Consider moving to syscalls.list.  */
> 
> would be a suitable comment.

Sure.  Please also file a bug that mentions the exact string to grep for
(unless you plan to replace them all with syscalls.list entries very soon
yourself).

> This also applies to files where there are calls to bounds-checking macros 
> for arguments to INLINE_SYSCALL, since the bounded-pointers code is 
> bitrotten.

Right.  (Even if it weren't, syscalls.list supposedly handles those too.)

> This also applies to files where the only boilerplate around 
> INLINE_SYSCALL is for cancellation (none in this patch, but 
> sysdeps/unix/sysv/linux/poll.c and the powerpc64 
> pread/pwrite/pread64/pwrite64 from the 2.2 patch).

Right.

> Files using INTERNAL_SYSCALL or INLINE_SETXID_SYSCALL cannot be converted 
> to syscalls.list (the former don't set errno, the latter have extra 
> processing in the threaded case).

Wrong(ish) and right.  A syscalls.list fourth column starting with E
requests a no-errno stub as for getpid (SYSCALL_NOERRNO set for
syscall-template.S) and one starting with V requests an errno-return-value
stub as for posix_fadvise (SYSCALL_ERRVAL set for syscall-template.S).
INTERNAL_SYSCALL cases just implementing those patterns are covered.

> On that basis, the following could have such a comment after the 2.2 
> patch:
[...]
> And the following after this patch:

I'll take your word for which files are in which categories.  Of course,
the "Consider" comment should be applied somewhat liberally, and the actual
consideration will decide whether to replace the file with a syscalls.list
entry or to change the comment to explain why that's not an option.


Thanks,
Roland