This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: TLS, 2015 edition
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: Roland McGrath <roland at hack dot frob dot com>, Stan Shebs <stanshebs at google dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Sun, 24 May 2015 21:02:08 -0400
- Subject: Re: TLS, 2015 edition
- Authentication-results: sourceware.org; auth=none
- References: <CA+5-Q5K7M3oJ-vvRP24GiCEHqnohWgqPkc7_6K9CLzc6a_MTcA at mail dot gmail dot com> <20150406201048 dot 311F22C3C70 at topped-with-meat dot com>
On 04/06/2015 04:10 PM, Roland McGrath wrote:
> Joseph summarized it pretty well. The backward compatibility picture might
> be somewhat more complex than that, and there is not yet any consensus on
> details of compatibility schemes. But there is consensus for a new
> pre-allocating implementation to be used at least when everything involved
> is newly compiled. IMHO getting the implementation of that started is the
> first thing that needs to be done.
>
+1.
Just reading through old mail and want to add my consensus here.
My interest here is in fixing the user-visible side of the issue,
and making TLS variables safe to use in all appropriate contexts.
While I also gave consensus for the original Google patch, it was
because it solved the immediate problem. It was later reverted
because of the impact on analysis tooling that required malloc
interposition to track allocations (much like the prevent discussion
on removing PLTs entirely via -Wl,-z,now, which I don't yet agree
with).
Cheers,
Carlos.