Building Etoile on Solaris

[Andreas Höschler]

Hi all,

I just downloaded etoile 0.4.1 and gave it a try on Solaris. Actually,
I am trying to get it built. I got around the usual

        /bin/sh: syntax error at line 1: `end of file' unexpected

problem by doing

        mv /bin/sh /bi n/sh.old
        ln -s /usr/bin/zsh /bin/sh

but now I get

        cd ./Frameworks/EtoileFoundation
        make

Making build-headers for framework EtoileFoundation...
make[1]: *** No rule to make target
`EtoileFoundation.framework/Versions/0/Headers/EtoileFoundation.h',
needed by `internal-framework-build-headers'.  Stop.
make: *** [EtoileFoundation.build-headers.framework.variables] Error 2

and have no idea (yet) how to fix that! Anyone?

Thanks a lot,

   Andreas




_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Chris B. Vetter]

On Thu, Aug 20, 2009 at 8:07 PM, Andreas Höschler<ahoesch@...> wrote:
> I got around the usual problem by doing
>        mv /bin/sh /bin/sh.old
>        ln -s /usr/bin/zsh /bin/sh

Not a good idea.
While Solaris 10 did away with statically linked shells and cron
generally uses /usr/bin/bash nowadays, there are still a ton of
scripts scattered around the system that depend on /bin/sh being a
Bourne shell, not zsh, csh or even bash.
Plus, a dynamically linked root shell will render your system unusable
if and when /usr/lib becomes corrupted.

The mentioned problem is that someone probably simply assumed /bin/sh
to be a renamed bash... Meaning, the script should be fixed, not the
system it is running on be messed with, since that might actually
break the system.

--
Chris

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

Hi Chris,

>> I got around the usual problem by doing
>>        mv /bin/sh /bin/sh.old
>>        ln -s /usr/bin/zsh /bin/sh
>
> Not a good idea.
> While Solaris 10 did away with statically linked shells and cron
> generally uses /usr/bin/bash nowadays, there are still a ton of
> scripts scattered around the system that depend on /bin/sh being a
> Bourne shell, not zsh, csh or even bash.
> Plus, a dynamically linked root shell will render your system unusable
> if and when /usr/lib becomes corrupted.
>
> The mentioned problem is that someone probably simply assumed /bin/sh
> to be a renamed bash... Meaning, the script should be fixed, not the
> system it is running on be messed with, since that might actually
> break the system.

Thanks for the hint. This of course was only a temporary solution to
get etoile built. But since this was only onne of many problems a
reverted to the original state regarding /bin/sh. Anybody successful
with etoile on Solaris recently? Unfortunately I am not such an
experienced scripter.

Thanks!

Regards,

   Andreas


_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[David Chisnall]

Hi Andreas,

Could you try building trunk?  I fixed a few GNUisms in some scripts a  
while ago, but they have not been back-ported to the old release.

David

On 20 Aug 2009, at 19:07, Andreas Höschler wrote:

> Hi all,
>
> I just downloaded etoile 0.4.1 and gave it a try on Solaris. Actually,
> I am trying to get it built. I got around the usual
>
> /bin/sh: syntax error at line 1: `end of file' unexpected
>
> problem by doing
>
> mv /bin/sh /bi n/sh.old
> ln -s /usr/bin/zsh /bin/sh
>
> but now I get
>
> cd ./Frameworks/EtoileFoundation
> make
>
> Making build-headers for framework EtoileFoundation...
> make[1]: *** No rule to make target
> `EtoileFoundation.framework/Versions/0/Headers/EtoileFoundation.h',
> needed by `internal-framework-build-headers'.  Stop.
> make: *** [EtoileFoundation.build-headers.framework.variables] Error 2
>
> and have no idea (yet) how to fix that! Anyone?
>
> Thanks a lot,
>
>   Andreas
>
>
>
>
> _______________________________________________
> Etoile-discuss mailing list
> Etoile-discuss@...
> https://mail.gna.org/listinfo/etoile-discuss

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

Hi David,

thanks for your response:

> Could you try building trunk?  I fixed a few GNUisms in some scripts a
> while ago, but they have not been back-ported to the old release.

With trunk and with mmy

mv /bin/sh /bin/sh.old
ln -s /usr/bin/zsh /bin/sh

I at least get a reasonable error message! :-)

making all for framework ObjectiveC2...
  Compiling file runtime.c ...
runtime.c: In function `class_addIvar':
runtime.c:149: warning: implicit declaration of function `strdup'
runtime.c:149: warning: assignment makes pointer from integer without a
cast
runtime.c:150: warning: assignment makes pointer from integer without a
cast
runtime.c: In function `class_addMethod':
runtime.c:180: warning: assignment makes pointer from integer without a
cast
runtime.c: In function `objc_allocateClassPair':
runtime.c:743: warning: assignment makes pointer from integer without a
cast
  Compiling file NSBlocks.m ...
  Compiling file blocks_runtime.m ...
blocks_runtime.m:33: error: parse error before "id"
blocks_runtime.m:34: fatal error: method definition not in class context
compilation terminated.
make[4]: *** [obj/blocks_runtime.m.o] Error 1
make[3]: *** [internal-framework-run-compile-submake] Error 2
make[2]: *** [ObjectiveC2.all.framework.variables] Error 2
make[1]: *** [internal-all] Error 2
make: *** [internal-all] Error 2

Trunk seems no good to start with. Exactly which scripts have you
modified? I would like to copy them over to the 0.4.1 tree and give
that a try!

Thanks,

   Andreas

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

Hi all,

> I at least get a reasonable error message! :-)
>
> making all for framework ObjectiveC2...
>  Compiling file runtime.c ...
> runtime.c: In function `class_addIvar':
> runtime.c:149: warning: implicit declaration of function `strdup'
> runtime.c:149: warning: assignment makes pointer from integer without
> a cast
> runtime.c:150: warning: assignment makes pointer from integer without
> a cast
> runtime.c: In function `class_addMethod':
> runtime.c:180: warning: assignment makes pointer from integer without
> a cast
> runtime.c: In function `objc_allocateClassPair':
> runtime.c:743: warning: assignment makes pointer from integer without
> a cast
>  Compiling file NSBlocks.m ...
>  Compiling file blocks_runtime.m ...
> blocks_runtime.m:33: error: parse error before "id"
> blocks_runtime.m:34: fatal error: method definition not in class
> context
> compilation terminated.
> make[4]: *** [obj/blocks_runtime.m.o] Error 1
> make[3]: *** [internal-framework-run-compile-submake] Error 2
> make[2]: *** [ObjectiveC2.all.framework.variables] Error 2
> make[1]: *** [internal-all] Error 2
> make: *** [internal-all] Error 2

I looked into this problem and am not sure why the compiler does not
like

/* Makes the compiler happy even without Foundation */
@interface Dummy
- (id)retain;
- (void)release;
@end

Where is the general object pointer id supposed to be defined? I have

gcc -v

Reading specs from /usr/local/lib/gcc/i386-pc-solaris2.10/3.4.4/specs
Configured with: ../configure --with-as=/usr/ccs/bin/as
--with-ld=/usr/ccs/bin/ld --disable-nls --disable-libgcj
--enable-shared --enable-languages=c,c++,objc
Thread model: posix
gcc version 3.4.4

Thanks,

   Andreas


_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Chris B. Vetter]

On Fri, Aug 21, 2009 at 12:50 PM, Andreas Höschler<ahoesch@...> wrote:
> Where is the general object pointer id supposed to be defined? I have
> gcc -v
> Reading specs from /usr/local/lib/gcc/i386-pc-solaris2.10/3.4.4/specs

Looks like you're using Sunfreeware's compiler, which IMHO is borked
beyond repair.
Give Blastwave a try instead: http://www.blastwave.org/

--
Chris

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

Hi Chris,

>> Where is the general object pointer id supposed to be defined? I have
>> gcc -v
>> Reading specs from /usr/local/lib/gcc/i386-pc-solaris2.10/3.4.4/specs
>
> Looks like you're using Sunfreeware's compiler, which IMHO is borked
> beyond repair.
> Give Blastwave a try instead: http://www.blastwave.org/

No, I don't. I always build one for myself with

        cd /usr/src
        gunzip gcc-3.4.4.tar.gz
        gtar xvf gcc-3.4.4.tar
        cd gcc-3.4.4
        mkdir objdir
        cd objdir
        ksh
        export CONFIG_SHELL=/bin/ksh
        export PATH=/usr/sfw/bin:/usr/xpg4/bin:/usr/ccs/bin:/usr/sbin:/usr/bin
        ../configure --with-as=/usr/ccs/bin/as --with-ld=/usr/ccs/bin/ld
--disable-nls  --disable-libgcj --enable-shared
--enable-languages=c,c++,objc
        gmake bootstrap
        gmake install

Anything wrong with that?

Thanks,

    Andreas


_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Niels Grewe]

Hi,

On Fri, Aug 21, 2009 at 12:11:55PM +0200, Andreas Höschler wrote:

> making all for framework ObjectiveC2...
>  Compiling file runtime.c ...
> runtime.c: In function `class_addIvar':
> runtime.c:149: warning: implicit declaration of function `strdup'
> runtime.c:149: warning: assignment makes pointer from integer without a cast
> runtime.c:150: warning: assignment makes pointer from integer without a cast
> runtime.c: In function `class_addMethod':
> runtime.c:180: warning: assignment makes pointer from integer without a cast
> runtime.c: In function `objc_allocateClassPair':
> runtime.c:743: warning: assignment makes pointer from integer without a cast

If I recall correctly strdup() is not mandatory according to POSIX. So
with some C libraries you have to explicitly turn it on (glibc behaves
that way, for example). Maybe you could check your system's strings.h
whether strdup() is wrapped in some #ifdef that needs be set.

Cheers,


Niels

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

Hi all

> On Fri, Aug 21, 2009 at 12:11:55PM +0200, Andreas Höschler wrote:
>> making all for framework ObjectiveC2...
>>  Compiling file runtime.c ...
>> runtime.c: In function `class_addIvar':
>> runtime.c:149: warning: implicit declaration of function `strdup'
>> runtime.c:149: warning: assignment makes pointer from integer without
>> a cast
>> runtime.c:150: warning: assignment makes pointer from integer without
>> a cast
>> runtime.c: In function `class_addMethod':
>> runtime.c:180: warning: assignment makes pointer from integer without
>> a cast
>> runtime.c: In function `objc_allocateClassPair':
>> runtime.c:743: warning: assignment makes pointer from integer without
>> a cast
>
> If I recall correctly strdup() is not mandatory according to POSIX. So
> with some C libraries you have to explicitly turn it on (glibc behaves
> that way, for example). Maybe you could check your system's strings.h
> whether strdup() is wrapped in some #ifdef that needs be set.

Thanks for the hint. I found

/usr/include/strings.h
/usr/share/src/gcc-4.2.3/fixincludes/tests/base/strings.h
/usr/share/src/gcc-3.4.4/gcc/fixinc/tests/base/strings.h

on the Solaris box but none of these files contains any occurance of
strdup! :-(

Is Etoile really so system (very special linux) specific in the
meanwhile? I got Etoile 0.2 built on Solaris with great efforts. Now
with 0.4.x this seems to be even more difficult. :-( Any more ideas?

Am I (still) the only one that is using GNUstep/Etoile on Solaris? In
conjunction with Sun Ray Server and Secure Global Desktop, Solaris is
by far the most interesting target platform for Etoile IMHO. So it's
definitely worth supporting this platform. All these software products
(including the OS) can be downloaded for free from www.sun.com and run
on a wide variety of x86 hardware. I really recommend anyone interested
to give it a try. I can also provide a VBox image of the latest Solaris
10 x86 (with GNUstep already preinstalled) if anyone in interested.

Thanks,

    Andreas


_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[David Chisnall]

Hi Andreas,

We'd definitely like to support Solaris.  I'm surprised that it  
doesn't support strdup() - it's pretty widely-supported (it comes from  
4.4BSD, and I thought Solaris got all of that stuff via SunOS).

You can add an implementation like this:

char *strdup(const char *in)
{
        size_t len = strlen(in);
        char *out = malloc(len+1);
        strcpy(out, in);
        return out;
}

If you declare that as static and pop it in runtime.c with #ifdef  
__solaris__ around it, then it should work fine.  Let me know about  
any GNUisms in the makefiles.  We test regularly on GNU/Linux and  
FreeBSD, which use different /bin/sh (bash and pdksh, I believe) so we  
catch most of them, but there may be some missing.

David

On 21 Aug 2009, at 15:03, Andreas Höschler wrote:

> Hi all
>
>> On Fri, Aug 21, 2009 at 12:11:55PM +0200, Andreas Höschler wrote:
>>> making all for framework ObjectiveC2...
>>> Compiling file runtime.c ...
>>> runtime.c: In function `class_addIvar':
>>> runtime.c:149: warning: implicit declaration of function `strdup'
>>> runtime.c:149: warning: assignment makes pointer from integer  
>>> without
>>> a cast
>>> runtime.c:150: warning: assignment makes pointer from integer  
>>> without
>>> a cast
>>> runtime.c: In function `class_addMethod':
>>> runtime.c:180: warning: assignment makes pointer from integer  
>>> without
>>> a cast
>>> runtime.c: In function `objc_allocateClassPair':
>>> runtime.c:743: warning: assignment makes pointer from integer  
>>> without
>>> a cast
>>
>> If I recall correctly strdup() is not mandatory according to POSIX.  
>> So
>> with some C libraries you have to explicitly turn it on (glibc  
>> behaves
>> that way, for example). Maybe you could check your system's strings.h
>> whether strdup() is wrapped in some #ifdef that needs be set.
>
> Thanks for the hint. I found
>
> /usr/include/strings.h
> /usr/share/src/gcc-4.2.3/fixincludes/tests/base/strings.h
> /usr/share/src/gcc-3.4.4/gcc/fixinc/tests/base/strings.h
>
> on the Solaris box but none of these files contains any occurance of
> strdup! :-(
>
> Is Etoile really so system (very special linux) specific in the
> meanwhile? I got Etoile 0.2 built on Solaris with great efforts. Now
> with 0.4.x this seems to be even more difficult. :-( Any more ideas?
>
> Am I (still) the only one that is using GNUstep/Etoile on Solaris? In
> conjunction with Sun Ray Server and Secure Global Desktop, Solaris is
> by far the most interesting target platform for Etoile IMHO. So it's
> definitely worth supporting this platform. All these software products
> (including the OS) can be downloaded for free from www.sun.com and run
> on a wide variety of x86 hardware. I really recommend anyone  
> interested
> to give it a try. I can also provide a VBox image of the latest  
> Solaris
> 10 x86 (with GNUstep already preinstalled) if anyone in interested.
>
> Thanks,
>
>    Andreas
>
>
> _______________________________________________
> Etoile-discuss mailing list
> Etoile-discuss@...
> https://mail.gna.org/listinfo/etoile-discuss

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Felix Holmgren]

> Solaris is
> by far the most interesting target platform for Etoile IMHO. So it's
> definitely worth supporting this platform.

I'd be interested to hear in more detail about why you feel like this.

>I really recommend anyone interested
> to give it a try. I can also provide a VBox image of the latest Solaris
> 10 x86 (with GNUstep already preinstalled) if anyone in interested.
I'd be interested! I don't have a strong commitment to any system at
the present.

/Felix

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[David Chisnall]

On 21 Aug 2009, at 15:22, Felix Holmgren wrote:

>> Solaris is
>> by far the most interesting target platform for Etoile IMHO. So it's
>> definitely worth supporting this platform.
>
> I'd be interested to hear in more detail about why you feel like this.

At the kernel level, Solaris has:

- ZFS (so does FreeBSD, although the Solaris version is a bit better  
tested)

- DTrace (we don't use this much yet, but it is possibly the most  
shiny thing ever invented for debugging and I want to integrate it  
with LanguageKit at some point).

- Probably the best threading implementation of any Free *NIX at the  
moment (including kernel scalability).

- Full and tested support for things like SCTP in the network stack,  
which will probably be very useful for CoreObject.

- Zones - think FreeBSD jails only more so (although FreeBSD 8 jails  
can now do most of what Zones can).

- A full, complete, and working OSS 4 implementation.  FreeBSD  
supports most of OSS 4, Linux is completely worthless when it comes to  
sound, with every release of every distro having a new, broken, and  
incompatible way of doing it.  If it has drivers for your hardware,  
OpenSolaris will give better

In the userland, there is a modern init system, based on a similar  
model to Launchd.  There is also a huge pile of legacy crap designed  
to run code from anything from SunOS 1.0 onwards.  Hopefully Solaris  
11 will move a load of this into legacy-compat zones where it  
belongs.  The new package manager is also shiny.

I generally prefer a more BSD-style userland, but Solaris is  
definitely high up the list of operating systems I'd recommend people  
consider.  If you read some of the early mailing list discussions, we  
were considering making OpenSolaris our standard platform when we put  
together a demo Étoilé system.

I'd definitely welcome more extensive testing on Solaris, but at the  
moment we only get regular testing on platforms that the developers  
use, which generally means Ubuntu, Fedora, or FreeBSD.

>> I really recommend anyone interested
>> to give it a try. I can also provide a VBox image of the latest  
>> Solaris
>> 10 x86 (with GNUstep already preinstalled) if anyone in interested.

Yes please!  I gave up trying to get Nexenta to install all of the  
dependencies for Étoilé.

David
_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Niels Grewe]

On Fri, Aug 21, 2009 at 03:20:05PM +0100, David Chisnall wrote:
> We'd definitely like to support Solaris.  I'm surprised that it  
> doesn't support strdup() - it's pretty widely-supported (it comes from  
> 4.4BSD, and I thought Solaris got all of that stuff via SunOS).

Actually, it still might. It's just that I shouldn't be allowed to give
advice when I'm in a hurry. strdup() is usually defined in string.h, not
strings.h as I claimed earlier. Checking back with Open Group reveals
that it is in fact a X/Open extension [0] and since I just dug up some
random document claiming that Solaris 2.5 (!) is X/Open compliant [1], I
highly suspect that it's somewhere there.

Sorry for the confusion,


Niels

[0] http://www.opengroup.org/onlinepubs/009695399/functions/strdup.html 
[1] http://docs.sun.com/app/docs/doc/802-1953/6i5uv2sj5?a=view 

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

>>> Solaris is
>>> by far the most interesting target platform for Etoile IMHO. So it's
>>> definitely worth supporting this platform.
>>
>> I'd be interested to hear in more detail about why you feel like this.
>
> At the kernel level, Solaris has:
>
> - ZFS (so does FreeBSD, although the Solaris version is a bit better
> tested)
>
> - DTrace (we don't use this much yet, but it is possibly the most
> shiny thing ever invented for debugging and I want to integrate it
> with LanguageKit at some point).
>
> - Probably the best threading implementation of any Free *NIX at the
> moment (including kernel scalability).
>
> - Full and tested support for things like SCTP in the network stack,
> which will probably be very useful for CoreObject.
>
> - Zones - think FreeBSD jails only more so (although FreeBSD 8 jails
> can now do most of what Zones can).
>
> - A full, complete, and working OSS 4 implementation.  FreeBSD
> supports most of OSS 4, Linux is completely worthless when it comes to
> sound, with every release of every distro having a new, broken, and
> incompatible way of doing it.  If it has drivers for your hardware,
> OpenSolaris will give better
>
> In the userland, there is a modern init system, based on a similar
> model to Launchd.  There is also a huge pile of legacy crap designed
> to run code from anything from SunOS 1.0 onwards.  Hopefully Solaris
> 11 will move a load of this into legacy-compat zones where it
> belongs.  The new package manager is also shiny.
>
> I generally prefer a more BSD-style userland, but Solaris is
> definitely high up the list of operating systems I'd recommend people
> consider.  If you read some of the early mailing list discussions, we
> were considering making OpenSolaris our standard platform when we put
> together a demo Étoilé system.
>
> I'd definitely welcome more extensive testing on Solaris, but at the
> moment we only get regular testing on platforms that the developers
> use, which generally means Ubuntu, Fedora, or FreeBSD.

There is more than that. Solaris 10 is the home of SSRS (Sun Ray Server
Software) and SSGD (Sun Secure Global Desktop), two very important
legacy packages. All our servers in the field run Solaris (RDBMS, NFS,
NIS, postfix,...). On the client side we have either Macs (MacOSX) or
Sun Rays. A Sun Ray is an Ultra Thin Client (no fan, no disk, no noise,
no heat), actually a graphics card with a network connector, a key
board and a mouse. If you log into a Sun Ray you start a session on the
server, the Sun Ray simply displays the buffer (rendered on the
server). Imagine a usual Sun machine (64GB RAM, 16 x 146GB ZFS, 8
processor cores) and about 30-40 Sun Rays connected to it serving
GNUstep/Etoile sessions. Actually that's exactly what we do. The Sun
Rays do not need to be in the LAN. The Sun Rays can also be used in a
WAN.

        Sun Ray -- DSL-Router --- {Internet} --- Sun Server (Solaris 10 with
SRSS installed)

Now, imagine you have one of those at home. You pull your smart card
from your home ray, drive to your office and insert your smart card
into any Sun Ray there. There's your session again with the cursor
still blinking at the same position in your word process or
presentation software where you left it (at home). This is the well
know card trick, that already convinced a lot of customers. :-) A Sun
Ray costs 250 EUR - 300 EUR (10 EUR on Ebay) and is absolutely
maintenance free (stateless). You might want to purchase a bigger
server every 3-5 years but you don't have to touch the desktops (Sun
Rays). They can easily do their job for 10 or more years. With SRSS
V5.0 there will also be a Sun Ray soft client emulating a Sun Ray on a
Windows Desktop (could be handy in some cases).

Then there is Sun Secure Global Desktop which is just another beast
that allows you to bring a Solaris/GNUstep/Etoile (of course GNOME as
well) based desktop onto any computer (e.g. a Windows machine) without
the need to install any software on the client.

OK, we are getting off-topic here. But please check out
hhttp://www.sun.com, get a copy of Solaris 10 and Sun Ray Server
Software (for free) and see for yourself.

Regards,

   Andreas



_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

On Freitag, Aug 21, 2009, at 16:39 Europe/Berlin, David Chisnall wrote:

>>> I really recommend anyone interested
>>> to give it a try. I can also provide a VBox image of the latest
>>> Solaris
>>> 10 x86 (with GNUstep already preinstalled) if anyone in interested.
>
> Yes please!  I gave up trying to get Nexenta to install all of the
> dependencies for Étoilé.

  I will provide an image. Give me a few hours ...

Regards,

   Andreas



_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

> n Fri, Aug 21, 2009 at 03:20:05PM +0100, David Chisnall wrote:
>> We'd definitely like to support Solaris.  I'm surprised that it
>> doesn't support strdup() - it's pretty widely-supported (it comes from
>> 4.4BSD, and I thought Solaris got all of that stuff via SunOS).
>
> Actually, it still might. It's just that I shouldn't be allowed to give
> advice when I'm in a hurry. strdup() is usually defined in string.h,
> not
> strings.h as I claimed earlier. Checking back with Open Group reveals
> that it is in fact a X/Open extension [0] and since I just dug up some
> random document claiming that Solaris 2.5 (!) is X/Open compliant [1],
> I
> highly suspect that it's somewhere there.
>
> Sorry for the confusion,

There it is:

pico /usr/include/string.h

#if defined(__EXTENSIONS__) || \
         (!defined(_STRICT_STDC) && !defined(__XOPEN_OR_POSIX)) || \
         defined(_XPG4_2)
extern char *strdup(const char *);
#endif

OK, here we go

cd /usr/src/etoileTrunk/Etoile
make OBJCFLAGS="-D__EXTENSIONS__=1"

Making all for framework UnitKit...
make[5]: Nothing to be done for `internal-framework-compile'.
  Copying localized resources into the framework wrapper...
Making all in Source/ToolSource...
Making all for tool ukrun...
make[5]: Nothing to be done for `internal-tool-compile'.
Making all in ObjectiveC2...
Making build-headers for framework ObjectiveC2...
zsh: parse error near `\n'
Making all for framework ObjectiveC2...
  Compiling file blocks_runtime.m ...
blocks_runtime.m:33: error: parse error before "id"
blocks_runtime.m:34: fatal error: method definition not in class context
compilation terminated.
make[4]: *** [obj/blocks_runtime.m.o] Error 1
make[3]: *** [internal-framework-run-compile-submake] Error 2
make[2]: *** [ObjectiveC2.all.framework.variables] Error 2
make[1]: *** [internal-all] Error 2
make: *** [internal-all] Error 2

pico ./Frameworks/ObjectiveC2/blocks_runtime.m

...
/* Makes the compiler happy even without Foundation */
@interface Dummy
- (id)retain;
- (void)release;
@end
...

The problem with the missing id definition remains!? Anything wrong
with my compiler buid?

gcc -v
Reading specs from /usr/local/lib/gcc/i386-pc-solaris2.10/3.4.4/specs
Configured with: ../configure --with-as=/usr/ccs/bin/as
--with-ld=/usr/ccs/bin/ld --disable-nls --disable-libgcj
--enable-shared --enable-languages=c,c++,objc
Thread model: posix
gcc version 3.4.4

Thanks,

    Andreas


_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[David Chisnall]

> The problem with the missing id definition remains!? Anything wrong
> with my compiler buid?

Please try with the latest svn.

David

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[Andreas Höschler]

Hi David,

>> The problem with the missing id definition remains!? Anything wrong
>> with my compiler buid?
>
> Please try with the latest svn.

Thanks a lot! One step further! :-)

make OBJCFLAGS="-D__EXTENSIONS__=1"
...
Making all in ObjectiveC2...
Making build-headers for framework ObjectiveC2...
zsh: parse error near `\n'
Making all for framework ObjectiveC2...
  Compiling file blocks_runtime.m ...
  Compiling file properties.m ...
properties.m: In function `objc_getProperty':
properties.m:15: error: storage size of '_stackExceptionData' isn't
known
properties.m:15: confused by earlier errors, bailing out
make[4]: *** [obj/properties.m.o] Error 1
make[3]: *** [internal-framework-run-compile-submake] Error 2
make[2]: *** [ObjectiveC2.all.framework.variables] Error 2
make[1]: *** [internal-all] Error 2

Any idea for that one?

Thanks,

    Andreas

_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss

[David Chisnall]

On 21 Aug 2009, at 17:32, Andreas Höschler wrote:

> Hi David,
>
>>> The problem with the missing id definition remains!? Anything wrong
>>> with my compiler buid?
>>
>> Please try with the latest svn.
>
> Thanks a lot! One step further! :-)
>
> make OBJCFLAGS="-D__EXTENSIONS__=1"
> ...
> Making all in ObjectiveC2...
> Making build-headers for framework ObjectiveC2...
> zsh: parse error near `\n'
> Making all for framework ObjectiveC2...
> Compiling file blocks_runtime.m ...
> Compiling file properties.m ...
> properties.m: In function `objc_getProperty':
> properties.m:15: error: storage size of '_stackExceptionData' isn't  
> known
> properties.m:15: confused by earlier errors, bailing out
> make[4]: *** [obj/properties.m.o] Error 1
> make[3]: *** [internal-framework-run-compile-submake] Error 2
> make[2]: *** [ObjectiveC2.all.framework.variables] Error 2
> make[1]: *** [internal-all] Error 2
>
> Any idea for that one?

It looks like your compiler doesn't support Objective-C exception  
handling constructs.  Did you enable exceptions when you built it?  
The ObjectiveC 2 framework provides the Objective-C 2 runtime support  
- it assumes that the compiler already knows how to handle all of  
Objective-C 1...

David
_______________________________________________
Etoile-discuss mailing list
Etoile-discuss@...
https://mail.gna.org/listinfo/etoile-discuss