This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 1/3] Adjust kernel-features.h defaults for recvmsg and sendmsg
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: Joseph Myers <joseph at codesourcery dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Tue, 29 Mar 2016 18:53:41 -0300
- Subject: Re: [PATCH 1/3] Adjust kernel-features.h defaults for recvmsg and sendmsg
- Authentication-results: sourceware.org; auth=none
- References: <1459175641-12520-1-git-send-email-adhemerval dot zanella at linaro dot org> <1459175641-12520-2-git-send-email-adhemerval dot zanella at linaro dot org> <alpine dot DEB dot 2 dot 10 dot 1603292128070 dot 15654 at digraph dot polyomino dot org dot uk>
On 29-03-2016 18:32, Joseph Myers wrote:
> On Mon, 28 Mar 2016, Adhemerval Zanella wrote:
>
>> +/* Support for recvmsg and sendmsg was added in 2.6.12 in most architectures,
>> + with some added later (for instance 2.6.37 for powerpc) and for some only
>> + through syscall (with wire-up syscall added in later versionsi). Define
>> + __ASSUME_RECVMSG_SYSCALL and __ASUME_SENDMSG_SYSCALL if glibc uses direct
>> + syscall, otherwise glibc will use socketcall. */
>
> This comment is a mess.
>
> * I'm doubtful of the "added in 2.6.12". Do you actually mean not in
> 2.6.11 as verified by comparing 2.6.11 and 2.6.12, or do you mean that
> 2.6.12 is the first release *in git* with the syscalls (for the trivial
> reason that it was the first release in git at all)?
Later, since I did not compare against 2.6.11 (only checked the git history).
>
> * "only through syscall" looks wrong in that context; I think you mean
> "only through socketcall".
>
> * "versionsi" typo.
>
> * "__ASUME_SENDMSG_SYSCALL" typo.
>
> I think you mean something more like:
>
> /* On most architectures, most socket syscalls are supported for all
> supported kernel versions, but on some socketcall architectures
> separate syscalls were only added later. */
>
> which does not need to go into details of the particular versions for
> particular architectures.
>
Indeed this is a mess, I will change to your suggestion.