This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 0/2] Tunables for glibc
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: Siddhesh Poyarekar <siddhesh at sourceware dot org>
- Cc: GNU C Library <libc-alpha at sourceware dot org>
- Date: Fri, 1 Jul 2016 12:20:11 -0700
- Subject: Re: [PATCH 0/2] Tunables for glibc
- Authentication-results: sourceware.org; auth=none
- References: <1467397878-11514-1-git-send-email-siddhesh at sourceware dot org>
On Fri, Jul 1, 2016 at 11:31 AM, Siddhesh Poyarekar
<siddhesh@sourceware.org> wrote:
> Hi,
>
> Here's another swipe at the tunables. As usual, I'll start with an apology for
> the very slow turnaround time and especially my tendency to come to this around
> freeze time.
>
> This version takes a completely different approach for a couple of reasons and
> ends up making the tunables interface simpler to use. Firstly, the tunables
> are now much richer as Florian requested. A tunable now has a data type
> associated with it, with the ability to attach attributes like type, minimum
> and maximum values to it as well as calling a callback function. Additionally,
> I got rid of the compatibility interface and included that in the tunable
> structure since that is much easier to deal with.
>
> Finally, I've moved the tunables out of the tunables directory and into elf/.
> This is because I had overlooked the fact that if we had to control ifuncs
> using tunables, they would have to be initialized much earlier. They are now
> initialized along with the LD_* environment variables. In fact, it might not
> be a bad idea to move the LD_* variables under the tunables mechanism as well.
I may have missed something. Do you include some testcases in your
patches? Does this frame work support static executables? If yes, how
does it work with static executables?
--
H.J.