Re: CVS commit: src/sys

[christos%astron.com@localhost (Christos Zoulas)]

In article <20121014191032.GC7211%britannica.bec.de@localhost>,
Joerg Sonnenberger  <joerg%britannica.bec.de@localhost> wrote:
>On Sun, Oct 14, 2012 at 06:57:58PM +0000, Christos Zoulas wrote:
>> In article <20121014184816.GA7211%britannica.bec.de@localhost>,
>> Joerg Sonnenberger  <joerg%britannica.bec.de@localhost> wrote:
>> >
>> >This is wrong in multiple ways:
>> >(1) Why do we care about FreeBSD compatibility? The existing code is
>> >pretty much useless and should just die already. There is no support for
>> >anything modernish anyway.
>> 
>> Because we want to share implementations of 3rd party programs.
>
>Doesn't sound like a good reason to repeat cruft and more importantly,
>to extend the list of hard-coded sysctl values. All new code should use
>sysctl by name.
>
>> >(3) The information is already available via libpthread and AT_STACKBASE
>> >when processing the aux vector directly.
>> 
>> Is there a way to get information from AUXV after ld.so is done, from libc?
>> I know about the stack functions in libpthread, but what happens if you
>> are not threaded?
>
>_dlauxinfo().
>
>> 
>> >As such, I ask for reverting this.
>> 
>> I don't mind. I was just doing some tests with boehm-gc. It would be nice
>> if there was a way to get to the stack base and size portably.
>
>We could provide the stack-related attributes in libc itself for
>non-threaded applications, but I am not sure how useful that is.
>boehm-gc pretty much has to deal with multi-threaded code anyway.

Fine, removed.

christos