Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 11 Feb 2013 12:22:37 +0100
From: Szabolcs Nagy <nsz@...t70.net>
To: musl@...ts.openwall.com
Subject: Re: [PATCH 2/3] Have different definitions of
 __pthread_tsd_main agree in size

* Jens Gustedt <jens.gustedt@...ia.fr> [2013-02-11 08:40:20 +0100]:
> Am Sonntag, den 10.02.2013, 19:31 -0500 schrieb Rich Felker:
> > Nope, that defeats the whole purpose, which is to avoid wasting space
> > when it's not needed. This warning is bogus. There's no reason a weak
> > symbol can't have different size than a strong one that optionally
> > replaces it.
> 
> Hm, about wasting space I am not very convinced, but maybe I didn't
> understand well enough why you need all these aliases after all, and
> why you can't refer to the real symbol directly.
> 

aliases are there for a reason

> In any case, this is perhaps better done with the tool chain. I have
> good experience by having
> 
>   -fdata-sections -ffunction-sections
> 
> for the compiler options and then
> 
>    -Wl, --gc-sections
> 
> for the link. In the case of musl, this removes exactly all the dummy
> sections :) plus two others (pad and sccp), see below.
> 

are you sure about the correctness of these?

they seem to be broken to me: weak aliases are not respected
and functions are dropped even if there are weak references to
them which is bad..

eg if sccp is dropped then in any code that does not
use pthread, the cancellable syscalls will be broken
(if i understand these right)

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.