Re: mips(el) n32 and n64 host triplet

2006-03-10 Thread Daniel Jacobowitz
On Fri, Mar 03, 2006 at 04:46:01PM +0100, Aurelien Jarno wrote:
 There are already triplets for this ;-)  Take a look at the glibc
 configuration; I believe you'd want mips64-linux-gnuabi64 et al.
 
 I can't find such thing in the glibc? In which file?

Sorry, I missed this message (went in the wrong folder).

Right now it's in ports/sysdeps/mips/preconfigure.  See the switch on
config_os.

-- 
Daniel Jacobowitz
CodeSourcery


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-03 Thread Aurelien Jarno

Daniel Jacobowitz a écrit :

On Wed, Mar 01, 2006 at 12:43:50PM +, Thiemo Seufer wrote:


I gave it some thought. Currently we only have mips64{,el}-linux-gnu,
which, to make matters worse, defaults to a n32 compiler with
multilibed o32/n32/n64 libraries. IMHO the best thing to do is to
introduce a completely separate set of triplets for the more advanced
single-ABI mips configurations. E.g:

mipsn32-linux-gnu
mipsn32el-linux-gnu
mipsn64-linux-gnu
mipsn64el-linux-gnu

The distinction between mips64 and mips_n_64 is a bit subtle, but
I don't see a better way while keeping the scheme relatively simple.

Comments?



There are already triplets for this ;-)  Take a look at the glibc
configuration; I believe you'd want mips64-linux-gnuabi64 et al.


I can't find such thing in the glibc? In which file?

IMHO, this triplets have the advantage to be compatible with the current 
config.{guess,sub}, and to be coherent with the new ARM ABI triplet, 
which is arm-linux-gnueabi.


On the other side they look a bit strange...

Bye,
Aurelien

--
  .''`.  Aurelien Jarno | GPG: 1024D/F1BCDB73
 : :' :  Debian developer   | Electrical Engineer
 `. `'   [EMAIL PROTECTED] | [EMAIL PROTECTED]
   `-people.debian.org/~aurel32 | www.aurel32.net


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-02 Thread Thiemo Seufer
On Wed, Mar 01, 2006 at 03:50:15PM -0500, Daniel Jacobowitz wrote:
 On Wed, Mar 01, 2006 at 04:14:44PM +, Thiemo Seufer wrote:
  Um. Actually, I'd like not to have mips64-linux-gnuabi32, but rather
  a convention which is acceptable over the whole toolchain.
 
 Why is this not acceptable over the whole toolchain?

ISTR there was a arch-vendor-os-libc-abi quintuplet in use. Was that
abandoned in favour of the approach you used?


Thiemo


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-01 Thread Thiemo Seufer
On Wed, Mar 01, 2006 at 12:23:35AM +0100, Aurelien Jarno wrote:
 Hi all!
 
 While thinking of multiarch, I remarked that there is no host triplet
 for mips(el) using the n32 or the n64 ABIs. Both of them use 
 mips64-linux-gnu, at least it is what is done in the glibc, the
 difference being done on the -mabi=n32 or -mabi=n64 option.
 
 Multiarch is based on the fact there is no namespace collision of the
 libraries are installed into (/usr)/lib/${host-triplet}.
 
 Does host triplets have already been planned for these two possible
 future ports?

I gave it some thought. Currently we only have mips64{,el}-linux-gnu,
which, to make matters worse, defaults to a n32 compiler with
multilibed o32/n32/n64 libraries. IMHO the best thing to do is to
introduce a completely separate set of triplets for the more advanced
single-ABI mips configurations. E.g:

mipsn32-linux-gnu
mipsn32el-linux-gnu
mipsn64-linux-gnu
mipsn64el-linux-gnu

The distinction between mips64 and mips_n_64 is a bit subtle, but
I don't see a better way while keeping the scheme relatively simple.

Comments?


Thiemo


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-01 Thread Aurelien Jarno

Thiemo Seufer a écrit :

On Wed, Mar 01, 2006 at 12:23:35AM +0100, Aurelien Jarno wrote:


Hi all!

While thinking of multiarch, I remarked that there is no host triplet
for mips(el) using the n32 or the n64 ABIs. Both of them use 
mips64-linux-gnu, at least it is what is done in the glibc, the

difference being done on the -mabi=n32 or -mabi=n64 option.

Multiarch is based on the fact there is no namespace collision of the
libraries are installed into (/usr)/lib/${host-triplet}.

Does host triplets have already been planned for these two possible
future ports?



I gave it some thought. Currently we only have mips64{,el}-linux-gnu,
which, to make matters worse, defaults to a n32 compiler with
multilibed o32/n32/n64 libraries. IMHO the best thing to do is to
introduce a completely separate set of triplets for the more advanced
single-ABI mips configurations. E.g:

mipsn32-linux-gnu
mipsn32el-linux-gnu
mipsn64-linux-gnu
mipsn64el-linux-gnu

The distinction between mips64 and mips_n_64 is a bit subtle, but
I don't see a better way while keeping the scheme relatively simple.

Comments?


That looks sane to me. If those names are accepted by everybody, I think 
it would be good to add them in config.{guess,sub}, both in Debian and 
in upstream. My experience in the GNU/kFreeBSD ports shows that it may a 
long time to propagate to the packages, so it may be a good idea to do 
that part now.


Then it would be possible to add them to dpkg, so that for example 
'dpkg-architecture -qDEB_HOST_GNU_TYPE' returns 'mipsn64-linux-gnu'.


--
  .''`.  Aurelien Jarno | GPG: 1024D/F1BCDB73
 : :' :  Debian developer   | Electrical Engineer
 `. `'   [EMAIL PROTECTED] | [EMAIL PROTECTED]
   `-people.debian.org/~aurel32 | www.aurel32.net


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-01 Thread Daniel Jacobowitz
On Wed, Mar 01, 2006 at 12:43:50PM +, Thiemo Seufer wrote:
 I gave it some thought. Currently we only have mips64{,el}-linux-gnu,
 which, to make matters worse, defaults to a n32 compiler with
 multilibed o32/n32/n64 libraries. IMHO the best thing to do is to
 introduce a completely separate set of triplets for the more advanced
 single-ABI mips configurations. E.g:
 
 mipsn32-linux-gnu
 mipsn32el-linux-gnu
 mipsn64-linux-gnu
 mipsn64el-linux-gnu
 
 The distinction between mips64 and mips_n_64 is a bit subtle, but
 I don't see a better way while keeping the scheme relatively simple.
 
 Comments?

There are already triplets for this ;-)  Take a look at the glibc
configuration; I believe you'd want mips64-linux-gnuabi64 et al.

-- 
Daniel Jacobowitz
CodeSourcery


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-01 Thread Thiemo Seufer
On Wed, Mar 01, 2006 at 09:45:13AM -0500, Daniel Jacobowitz wrote:
 On Wed, Mar 01, 2006 at 12:43:50PM +, Thiemo Seufer wrote:
  I gave it some thought. Currently we only have mips64{,el}-linux-gnu,
  which, to make matters worse, defaults to a n32 compiler with
  multilibed o32/n32/n64 libraries. IMHO the best thing to do is to
  introduce a completely separate set of triplets for the more advanced
  single-ABI mips configurations. E.g:
  
  mipsn32-linux-gnu
  mipsn32el-linux-gnu
  mipsn64-linux-gnu
  mipsn64el-linux-gnu
  
  The distinction between mips64 and mips_n_64 is a bit subtle, but
  I don't see a better way while keeping the scheme relatively simple.
  
  Comments?
 
 There are already triplets for this ;-)  Take a look at the glibc
 configuration; I believe you'd want mips64-linux-gnuabi64 et al.

Um. Actually, I'd like not to have mips64-linux-gnuabi32, but rather
a convention which is acceptable over the whole toolchain.

Btw, the 'kern64' in that preconfigure looks like a bug, but appears
to be unused.


Thiemo


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: mips(el) n32 and n64 host triplet

2006-03-01 Thread Daniel Jacobowitz
On Wed, Mar 01, 2006 at 04:14:44PM +, Thiemo Seufer wrote:
 Um. Actually, I'd like not to have mips64-linux-gnuabi32, but rather
 a convention which is acceptable over the whole toolchain.

Why is this not acceptable over the whole toolchain?  Alternatively,
if you're only talking about the o32 case, then use mips-linux-gnu for
that.

-- 
Daniel Jacobowitz
CodeSourcery


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]