OpenSolaris Port

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

OpenSolaris Port

dtbartle
Hello,

I've been working on porting GNU libc to OpenSolaris for the past few
months. I've put a web page detailing my work [0] (the actual diff is
in glibc-2.7-kopensolaris-gnu.diff.gz). NPTL has also been ported; the
mutex/rwlock/semaphore parts and thread parts have been rewritten to
use OpenSolaris-specifc locking/threading syscalls. Most of the patch
could theoretically be added to glibc-ports, although there are a
number of changes that would need to be made to base glibc,
particularly in NPTL. Any patches to non-sysdeps stuff has been done
in a way that would not affect building on the supported platforms;
i.e. I've added #define's and used #if/#else blocks to add the
OpenSolaris-specific functionality. A good example is the use of
PTHREAD_T_IS_TID, which causes pthread_t's to be thread/lwp id's
instead of pointers to the actual thread descriptor, which is what
sun's libpthread/libthread does.

My current goal is to end up with something that would completely
replace sun's libc. Many of sun's extensions (f.e. the zone_* bits in
libc) are simply wrappers around syscalls, so I've reimplemented them
in the port. Since glibc is licensed under LGPL and includes a linking
exception, it should be legal to link OpenSolaris CDDL components
(e.g. libzonecfg.so.1 or /usr/bin/zonecfg) against glibc. I've been
focusing on i386, but it should be little work to make an x86_64 and
sparc port.

There's not much point in merging any changes yet, since it's still a
work in progress. This email is just a notification. Comments and
ideas are appreciated.

-- David

[0] http://csclub.uwaterloo.ca/~dtbartle/opensolaris/
Reply | Threaded
Open this post in threaded view
|

Re: OpenSolaris Port

Carlos O'Donell-2
On Sat, Aug 30, 2008 at 10:17 PM, David Bartley
<[hidden email]> wrote:
> There's not much point in merging any changes yet, since it's still a
> work in progress. This email is just a notification. Comments and
> ideas are appreciated.

You should try to centralize the various NO_* macros into the
kernel-features.h file, the same goes for all the __ASSUME_* macros
you define or undefine.

Have you gotten to the point where you can run the testsuite? :-)

Cheers,
Carlos.
Reply | Threaded
Open this post in threaded view
|

Re: OpenSolaris Port

dtbartle
On Sun, Aug 31, 2008 at 5:13 PM, Carlos O'Donell
<[hidden email]> wrote:

> On Sat, Aug 30, 2008 at 10:17 PM, David Bartley
> <[hidden email]> wrote:
>> There's not much point in merging any changes yet, since it's still a
>> work in progress. This email is just a notification. Comments and
>> ideas are appreciated.
>
> You should try to centralize the various NO_* macros into the
> kernel-features.h file, the same goes for all the __ASSUME_* macros
> you define or undefine.
>

Good idea, thanks.

> Have you gotten to the point where you can run the testsuite? :-)

Yes. Most of the failures are expected, mostly do to unimplemented stuff.

>
> Cheers,
> Carlos.
>
>

-- David