Re: CoreFoundation

[Luboš Doležel]

On 01/07/2015 02:15 AM, Lundberg, Johannes wrote:
> Hi
> 
> Lately I been finding myself in the need of manipulating the defaults
> register from C code and I remember some discussion about CF a while ago
> where someone was working on it.
> 
> What is the current status? How complete and usable is CF?
> 
> Thanks!
> --
> Johannes Lundberg
> BRILLIANTSERVICE CO., LTD.

Hi,

the most common stuff should be working, but it hasn't seen much use, so
there could be bugs indeed.

I have a feeling, though, that the property-related APIs are incomplete.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: CoreBase from SVN fails to configure due to wrong libdispatch

[Luboš Doležel]

On 01/30/2015 10:38 PM, Stefan Bidigaray wrote:
> I committed a hack to disable the broken check for GCD altogether until
> this can be sorted out.
> 
> I forgot to reply to all on my other message that said:
> 
> So I went hunting for what is happening and found a reference that says
> the configure script is broken
> (http://wiki.inkscape.org/wiki/index.php/CrossCompilingOsX#Building_gnustep-corebase).
>  
> I haven't gone through this particular area in a long time (haven't
> build with libdispatch in over a year).
> 
> It seems we only use GCD in CFRunLoop and it's not entirely clear to me
> what it's doing here (again, need Lubos' input).  For now, my suggestion
> is to pass --without-gcd.  CFStream should still work without GCD
> support, although I'm not sure how CFRunLoop will be impacted.
> 

Hi,

CFRunLoop is supposed to be integrated with libdispatch because that's
how it is on OS X.

dispatch_get_main_queue_eventfd_np() is needed for working integration.
I've checked the package available in Ubuntu and it's either broken or
ancient, because this symbol is not present.

So I'd suggest removing your hack and using --without-gcd if your distro
has old/broken libdispatch.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: CoreBase from SVN fails to configure due to wrong libdispatch

[Luboš Doležel]

On 01/31/2015 05:56 PM, Stefan Bidigaray wrote:
> So I did some searching and found that there are 2 versions of
> libdispatch floating around.  The official version, which is included as
> part of Debian, Ubuntu and most other distros
> (https://libdispatch.macosforge.org/) and a version by nickhutchinson
> which includes dispatch_get_main_queue_eventfd_np()
> (https://github.com/nickhutchinson/libdispatch).  The problem is that
> the official version of the library does not include this function, only
> nickhutchinson's version, and since his version is not what is currently
> distributed with distributions, this function doesn't exist for most
> people, even though they have the latest version of libdispatch.
> 
> Is there any other way we can support this functionality without the use
> of the _np functions?  My search also found we are the only project that
> actually uses it, so it did not gain wide support.
> 

Umm, I can't think of any way :-/

Even Apple uses internal APIs to (ending with "_4CF") to allow for
GCD/CF integration. On OS X, these APIs return a Mach port.

-- 
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: [NSNumber numberWithFloat:] issue

[Luboš Doležel]

Hi,

that's strange - I believe that CFRetain() already worked with small 
NSStrings, because it calls object_getClass() and that returns a special 
class type for these, hence CFRetain() should just forward the call to 
[obj retain], which is a NOP.


Could you please send a backtrace with the crash?

Luboš

On Wed, 18 Feb 2015 20:39:27 -0600, Stefan Bidigaray wrote:

That's because CoreBase does not understand "small objects".  These
are objects that have their contents stored as part of the pointer,
instead of being an actual object.  Because they aren't actually
objects, they don't have any reference counting.  Passing a small
object around is like passing a int or double, and do not have to 
call

malloc or free, making them very fast.

I can implement a quick fix, but it might not until the weekend.  It
would also be very helpful if you created a test case that I could
include in the test suite.  Nothing fancy.  Or, if you're not
familiar with the test suite, a simple program with the expected
output.

Thanks

On Wed, Feb 18, 2015 at 6:16 PM, Amr Aboelela  wrote:


Hello, 

I found that [NSNumber numberWithFloat:] is not toll free bridged
with CoreFoundation
if you do CFRetain on an object result from [NSNumber
numberWithFloat:] it will crash, and if you do a retainCount on it,
it will always give you -1

I feel suspicious about this line:

return boxDouble(aValue, SMALL_FLOAT_MASK);
___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org [1]
https://lists.gnu.org/mailman/listinfo/discuss-gnustep [2]






___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Using GCC

[Luboš Doležel]

On 04/07/2015 07:39 PM, Patryk Laurent wrote:
> 
>> On Apr 6, 2015, at 1:51 PM, Fred Kiefer  wrote:
>>
>> GNUstep should work as good with gcc as it does with clang.
> 
> Hi Fred,
> 
> Just to clarify, does GCC now support features like ARC, blocks and grand 
> central dispatch?
> 
> Thank you,
> Patryk

ARC and Blocks: no

GCD is just an API, so as long as you don't use functions accepting
blocks, you should be fine.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Using GCC

[Luboš Doležel]

On 04/08/2015 05:02 AM, John W Kennedy wrote:

>> Yes and no. There is a kernel GCD and there is a Foundation GCD, and
>> kernel GCD is not necessarily stable when used with Foundation, at
>> least on Apple.
> 
> I should rephase that. On Apple, Foundation is not necessarily stable
> when used with kernel GCD. If you are using Foundation at all, then if
> you wish to use GCD, you must use the Foundation GCD mechanisms
> (NSOperation, NSOperationQueue, etc.), rather than the kernel GCD
> mechanisms (dispatch_get_main_queue, etc.). Naturally, this has no
> direct bearing on whether the code will work on GNUStep, but there is a
> portability danger, and the problem involves race conditions, so it may
> work for months and then suddenly start to fail.
> 

I'd find this VERY strange. dispatch_* functions indeed can be used with
Foundation NSOperation etc. - there is no danger in doing that.
CFRunLoop/NSRunLoop itself is well integrated with GCD and enqueueing
blocks is a risk-free operation.

I'd also refrain from calling it "kernel GCD". It is true that some
parts needed for GCD to function are in the kernel (pthreads and pthread
workqueues), but that's it. It's a pure userspace library implemented on
top of these primitives.

On the contrary, GNUstep's NSRunLoop is not integrated into CFRunLoop
(which would integrate it into GCD). Hence, when using GCD with GNUstep,
you cannot use the main queue, but you can probably use other queues.
-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: NeXT GNU Homage Project Work

[Luboš Doležel]

Dne 16.11.2015 v 07:56 Maxthon Chan napsal(a):

How about this - a bit of endeavour in kernel space, Linux-as-XNU:

1) A kernel module that implement XNU kernel’s syscalls in Linux kernel
2) A Mach-O dynamic loader for Linux that can work with binfmt_misc
3) A port of basic OS X libraries that takes advantage of this interface

This will allow OS X apps and Linux apps work under one single, 
modified Linux kernel.


You have precisely described what Darling is doing, more specifically in 
its current rework ("darling-libsystem" branch).


1) The kernel module is called darling-mach. But it also implements some 
very special BSD syscalls used on OS X (e.g. bsdthread_terminate or 
psynch_mutexwait).

2) libdyld/dyld
3) I'm gradually adding various libs present in OS X to the build. The 
latest additions are gnustep-base and gnustep-corebase.


But that has very little to do with "NeXT" discussed in this thread ;-)

Luboš

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: NeXT GNU Homage Project Work

[Luboš Doležel]

On 11/17/2015 04:34 AM, Maxthon Chan wrote:
> Sorry…
> 
> If this is the case then I would suggest those AppKit and UIKit
> enhances, sort of bringing GNUstep up to speed with OS X:
> 
> 1) Storyboards. This is how Xcode arrange user interfaces now. 2)
> QuartzCore and WIndowServer based on Wayland and EGL. This will
> replace Window Maker with a higher performance interface engine, and
> replace gnustep-back entirely. (“windowmaker-wayland” is a
> Wayland-only compositor that have the OS X look and feel,
> “gnustep-qe” implements QuartzCore API using Wayland and EGL, and
> gnustep-gui rewritten to use gnustep-qe instead of gnustep-back) 3)
> Metal, which can be mapped to Vulcan(“gnustep-metal-vulcan”), Mantle
> (for AMD cards, “gnustep-metal-mantle”) or CUDA (for nVIDIA cards,
> “gnustep-metal-cuda”) 4) SceneKit and SpriteKit, both can be
> implemented on top of Metal. 5) loginwindow (which is necessary in
> building a complete desktop experience)
> 

Well, the biggest problem is the (nearly) complete lack of integration
between CoreBase & Base (NSRunLoop) and opal & quartzcore & gnustep-gui.
And even apps that don't make any CG/CA calls cause crashes, simply
because -gui was never really tested with real OS X apps.

(I will work on CFRunLoop-based NSRunLoop in near future.)

Right now, my impression is that if I spend a few months hacking on a
new CG/CA/AppKit implementation layered on top of Qt's accelerated
rendering engine, it will suck big time, but still have a better
compatibility and performance.

2) My point is not to clone OS X, so I'd leave this to X11/Qt/...

3) and 4) The question is how many apps need this.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: NeXT GNU Homage Project Work

[Luboš Doležel]

On 11/18/2015 09:46 AM, Maxthon Chan wrote:
> Not just that, maybe we need a new TFB mechanism that is ARC- and
> Swift-proof, and maybe rewrite opal and quartzcore. The reason I suggest
> building QC on top of EGL and Wayland is because how technology layers
> from OS X maps here.

ARC is a compile-time stuff, so it's not important to Darling.

> 2) X11 is on its way out so I think we need to put our eyes on the
> future-proof things.

Which in "my approach" is Qt, because it abstracts away the graphics
server/windowing server stuff.

> 3 & 4) Metal, SpriteKit and SceneKit are for games, preferably
> distributed through Steam. If we have clones of OS X game frameworks
> then Steam (and maybe App Store) games for OS X will become one
> recompile away from being released under Linux. Game Center and other OS
> X-specific game features can be implemented by calling into Steam
> platform’s similar features.

I doubt many games for OS X use these. Most of the time, with games
ported from Windows, I see just OpenGL, OpenAL or CoreAudio, IOKit for
input and some very simple code to create a window (often in Carbon!).

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

OS X emulation layering

[Luboš Doležel]

On 11/18/2015 10:02 AM, Maxthon Chan wrote:
> We probably need to rearrange libraries a bit since the current
> layering is just too messy.
> 
> Foundation technology stack: libc -> libobjc (implement NSObject
> here) -> gnustep-corebase (with TFB classes implemented here) ->
> gnustep-base

I've done that slightly differently:

* libsystem depends on system_c, ..., and libobjc2 (needed because of
Blocks runtime embedded in libobjc2).

* corebase and base compiled together into a single .so. Required due to
an unpleasant dependency loop looming between the two. (NSCF* classes
require base and NSRunLoop will require corebase).

> Graphics technology stack: EGL (my suggestion) or
> gnustep-metal -> gnustep-qe -> gnustep-opal -> gnustep-gui
> Audio
> technology stack: ALSA -> gnustep-coreaudio (with Audio Units)

There is no gnustep-coreaudio, but there is Darling CoreAudio which can
already support a very simple app such as mpg123. It is built on top of
ALSA and ffmpeg APIs (audio decompression and processing), but can be
easily changed to support OSS for example.

There is still a lot of work left though (e.g. Audio Queues).

> Using Qt will result in Qt interfacing code being embedded deep into
> gnustep-gui (NSWindow at least, maybe even GUI elements) which may
> break some OS X apps that uses QC or CG to draw their own UI
> elements. EGL is a few layers lower and actually maps to the layer of
> OpenGLES that QC on iOS works on so we can just build the graphics
> stack iOS style.

I've already thought this through (the case you're mentioning) and it
can work. I'd specifically build on top of the Qt Quick engine.

Starting on the EGL layer is not realistic. Not without a large team of
people hacking on this, which is simply not there.
-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Swift is now open source (Apache 2 License)

[Luboš Doležel]

On 12/03/2015 08:22 PM, Maxthon Chan wrote:
> On their side, I’m gonna go there and propose the option of using an
> existing Foundation reimplementation in place of the version they
> packed.

Over the time that I've been analyzing OS X, how things work deep down,
I came to the conclusion that developers at Apple must be *really* very
bored.

While on Linux, one can see the developers are lazy, strive for a clean
and simple design, Apple's developers probably dream about what crazy
complicated project of reimplementing something that has already been
done to start next.

A nice example are workqueues as used by GCD (libdispatch). They
implemented on top of a portable pthread pool. Then they thought, hey,
why don't we save us making maybe 1 system call, get the kernel
needlessly involved in managing the thread pool and feel absolutely
fantastic about it.

Another example. Pthread locking primitives. On Linux, there is a single
simple system call named futex. All the mutexes, conditions, r/w locks
etc. are built on top of that.
In Apple, they thought this would be way too simple. So instead, they
introduced a dozen of system calls (psynch_*) with gazillions of
parameters and very strange semantics. When I read through the
implementation in XNU, I wonder why did someone torture himself so much.

So when I see Apple redoing something, I'm just not surprised anymore.
These folks don't want easy and simple. Because then they would have
nothing to do and would get fired.

> On our side, maybe it is time to scrap our CoreBase and reimplement
> Base on top Apple’s CoreFoundation.

a) That's a lot of work someone would have to do. And to little benefit,
I'm afraid.
b) Makes you depend on APSL'd code.

I'd rather suggest that GNUstep finally moves to Git so that I can
contribute back more easily. Right now, it's easier for me to commit
into my own gnustep-corebase fork on GitHub then to commit into SVN,
wait for it to bubble to the Git mirror and merge it into the Darling fork.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Swift is now open source (Apache 2 License)

[Luboš Doležel]

On 12/03/2015 10:59 PM, Svetlana A. Tkachenko wrote:
>> Maybe abandon the Savannah ship entirely and migrate to Github?
> 
> I oppose GitHub.
> It is proprietary software.
> I do not have an account there and would be unable to contribute.
> 
> I think Savannah supports git.
> 

...!

GitHub still uses Git, which is open source.

The rest is just a website. Do you require the backend source code for
every website you visit?

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Savannah vs. Gitlab

[Luboš Doležel]

Dne 4.12.2015 16:24, David Chisnall napsal:

On 4 Dec 2015, at 09:55, Ivan Vučica  wrote:


Primary hosting on Github? *shrug* there are no strong benefits, I 
think, and sufficient number of people would be happier if we hosted 
elsewhere.


I strongly disagree with this.  The benefits for collaboration from
being able to easily see who has forked your repo, to post comments on
their code asking about upstreaming, and so on are huge.

GitHub is far more than just a hosting service (though one that has an
issue tracker that integrates tightly with the revision control
system, and a decent code review system), it’s also a social platform
and that’s where the majority of its value lies.

David


I concur. We're talking about a "number of people" who would prefer 
Savannah, but the amount of people actually submitting code is very low. 
Just going to Savannah's website makes me feel like this is a dead 
project.


If - for whatever reason - you later come to the conclusion that GitHub 
was not the right choice, migrating elsewhere is a matter of hours.


If I were Greg, I'd also consider moving the whole project away from 
FSF. While I recognize the importance FSF has (had) for open source, 
just the need to do a snail mail round-trip with FSF to be allowed to 
contribute is a great obstacle for any FSF project. And I can tell you, 
a legally worthless one in my case - the Czech law explicitly forbids 
reassigning copyright except in well-defined cases (employee to employer 
and inheritance after author's death).


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Savannah vs. Gitlab

[Luboš Doležel]

On 12/07/2015 08:08 PM, Gregory Casamento wrote:
> German,
> 
> I will have to think on it for a bit to be honest.   I got a personal
> email from RMS asking to please NOT move to GitHub and that they are
> considering issuing a statement which says that no GNU project should
> be on GitHub.

Respecting all of hist great work, RMS is an extremist. I wonder how
he's going to function in the world of 3D printers. Is he not going to
buy a toothbrush, to which there is no open source physible?

In my view, due to the decentralized nature of Git, using GitHub does
not restrict anyone's freedom. If RMS thinks otherwise, then FSF should
provide a comparable alternative.

> I believe what's best is to move to a git platform which is free
> software like gogs and maintain the mirror on GitHub.  Savannah is
> just not doing it for us as they STILL have not responded to my
> request to add repositories.   At this point it is just too much.

Having just a mirror on GitHub, yet accepting pull requests is not very
realistic. Git is good for decentralized development, but having
multiple "push targets" brings synchronization problems, because then
you can have potentially diverging histories.

I will go even further and say that when you accept pull requests on
GitHub, you de facto *are* hosted on GitHub (and not just mirrored),
even if you declare that server XYZ is the official host.

Either way, I'll add another alternative:
http://repo.or.cz/ is used by many open source projects. It is hosted by
the Charles University in Prague. It is historically the first public
Git hosting.

But it pales in comparison with GitHub. Not pull requests, no issue
tracking, no code review.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Savannah vs. Gitlab

[Luboš Doležel]

On 12/10/2015 02:26 AM, Svetlana A. Tkachenko wrote:
> 1) as far as I can see GitHub things are separate from decision being
> discussed - this thought is reasonable, right? it is so because github
> can't be the official place

I haven't yet seen a final word from Greg about GitHub. So I don't know
why you say it cannot.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Please, no GitHub

[Luboš Doležel]

Dne 10.12.2015 06:27, Richard Stallman napsal:

GitHub does a number of things that are wrong.  It requires users to
run nonfree programs (Javascript code in web pages) to do certain
operations (see http://gnu.org/philosophy/javascript-trap.html).


So instead of giving people a choice of not running these scripts, you 
rather have them use another site that possibly doesn't provide such 
functionality at all?



It
encourages bad licensing practices, such as posting code with no
license (which makes it nonfree), and failing to state the license
clearly.  Among free licenses, it encourages weak licenses over
copyleft.


GitHub does not encourage posting code without a license. In its very 
simple wizard, it gives you a choice of adding a license file on your 
behalf. The default is NONE simply because adding a file means *creating 
a commit*, which would cause many people to run into trouble if they had 
run 'git init' locally first (git push would fail).


And again, it does not encourage weak licenses. It lists a dozen of 
possible licenses with three of the most popular being emphasized. It 
gives people a choice instead of pushing a single philosophy. And I say 
that as a proponent of GPL, who has never released any of his code under 
a BSD, MIT or similar license.


But the licence of GNUstep is clear, so this is no concern.


Those are the flaws I remember, and they are grave.  There may be more
I don't recall now.


They are not grave, this is simply nitpicking. Your vision of freedom is 
rather absolutistic, since forcing freedom upon others goes against the 
very nature of freedom.


GitHub has been tremendously beneficial to free software, especially in 
the light of gradual demise of SourceForge, which used to be the hosting 
facility of choice for free software.


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Please, no GitHub

[Luboš Doležel]

On 12/12/2015 06:01 AM, Richard Stallman wrote:
>   > They have created a website to suggest licenses for newcomers
>   > (http://choosealicense.com/ <http://choosealicense.com/>) and all
>   > license listed there are free licenses, with GPLv2 and GPLv3
>   > featured at the same time under one of their three major
>   > categories.
> 
> It fails to mention GPL v3-or-later at all.  It tends to leads people
> to use GPLv3-only or GPLv2-only, both of which we deprecate.
> 

"GPLv3 or later" is not a license. GPLv2 is a license, so is GPLv3. They
can hardly promote a license that doesn't even exist yet.

Since choosealicense.com hosts a full copy of GPLv3
(http://choosealicense.com/licenses/gpl-3.0/), there is indeed a section
recommending declaring "either version 3 of the License, or (at your
option) any later version" in the source files.

As a personal opinion, I don't like stating "GPLv3 or later", because it
makes me release source code under a license I never could have read and
agreed to. I am, technically, giving FSF the right to release my source
code under any terms they come up with for the next GPL.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Questions

[Luboš Doležel]

On 02/29/2016 07:17 PM, Liam Proven wrote:

Quote:*run existing Cocoa apps*

Do you mean to try to run Mac binaries on Linux?

If so, that is the domain of an emulator. One that is trying to
emulate a proprietary platform which is the property of a very large,
wealthy, litigious company. That is unreasonable and unrealistic IMHO.
Outlandish is your term, not mine.


When this is brought up, I always interject :-)

http://www.darlinghq.org. Already runs many binaries, also Apple's 
toolchain (with the notable exception of xcodebuild, which depends on 
half of the operating system), soon should be able to fully host a 
Jenkins/TC build slave. It can mount DMG images, install PKGs and it 
provides a more or less complete OS X shell environment.


As far as the relationship between Darling and GNUstep is concerned, 
Darling builds on GNUstep's Foundation (with tighter CF integration), 
extends on GNUstep's CoreFoundation and develops its own AppKit (for 
various technical reasons, over which I don't want to do flamewars here).


Given how far I got with my one man show, it is for sure not 
unrealistic. In a way, GNUstep is currently a ~3 man show, given that I 
see only ca. 3 truly active committers.


And I don't know what could Apple sue me over.

P.S.: Darling's AppKit is already more advanced than the nothing that 
can be seen in Git (in some NIB compatibility areas, it is already more 
advanced than GNUstep). But I won't do any commits for a long time, 
because people are very impatient and I don't want to receive 5 mails 
per day saying "hey, your AppKit sucks, hey this commit breaks build" etc.


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Questions

[Luboš Doležel]

I don't use VMware, though -- only VirtualBox. So, sadly, useless to
me and anyone who strongly prefers FOSS tools.



OS X does run under VirtualBox, but the performance of OS X in any VM is 
*very* poor.

--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: mulle-objc #MakeObjCGreatAgain

[Luboš Doležel]

Dne 1.12.2016 v 17:30 David Chisnall napsal(a):

On 1 Dec 2016, at 16:26, Matt Butch  wrote:

I’m hoping to be able to use Objective-C on servers, so Mulle-Objc might be a 
good option for that. If it can get a good community behind it, even better.

I’m not sure why Mulle-Objc brings there.  I know of three different Foundation 
implementations (GNUstep, one for FreeBSD that makes heavy use of kqueue, and 
WinObjC) that use all use the GNUstep Objective-C runtime, which provides a 
superset of the features that Apple provides.

David


I second this. I fail to see the benefits of yet another runtime...

Luboš




smime.p7s
Description: Elektronicky podpis S/MIME
___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

ABI compatibility with Cocoa

[Luboš Doležel]

Hello,

as part of my diploma thesis I'm working on a userspace Darwin 
compatibility layer for Linux. The primary task is to get unmodified 
console applications running on top of a Wine-like layer. Currently, I'm 
working on a compatible libSystem.dylib reimplementation - i.e. 
resolving ABI issues and implementing missing Mach and BSD kernel calls, 
albeit in many cases it's just a dummy function.


http://darling.dolezel.info/

As a long-term aim, I would also like to get other applications running 
- personally especially 3D games, but ordinary GUI apps as well. One of 
the important pieces is Cocoa and GNUstep seems to be a good choice for 
dealing with AppKit implementation on Linux.


Towards this end, I'd like to ask what you see as the biggest hurdles in 
ABI (not just API) compatibility when replacing Apple's AppKit with GNUstep?


I'd like to note that I realize what I'm trying to achieve is a 
*tremendous* task, but well, I believe it can be done!


Regards,
--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: ABI compatibility with Cocoa

[Luboš Doležel]

On Mon, 6 Aug 2012 15:41:20 +0100, David Chisnall wrote:

You might be interested in a NetBSD project that did the same thing
about 8 years ago.  They eventually gave up because all of the
interesting OS X software (e.g. the window server) has hooks so 
deeply

into the Mach layer, and everything on top of it has close
dependencies on it, that you effectively need 99% of OS X to have
anything interesting.


Oh yes, I've read all about that NetBSD project. But their approach was 
quite different from mine. They actually tried to make *the system* be 
like Darwin/OS X, while I'm doing everything in the user space. Just 
like Wine doesn't emulate any sort of very low level windowing 
functionality, they simply reimplemented the Windows APIs that are used 
to create GUIs.


That means I will certainly not be dealing with the window server the 
level the NetBSD guys did, I'll only be providing a 
semi-reimplementation of the upper-level libraries the applications use.


I'm also not making any modifications to the Linux kernel at all.


Replacing Cocoa's AppKit with GNUstep is difficult.  In theory, you
should be able to just add Apple runtime parts for the
runtime-specific things in, for example, NSObject.  This isn't
trivial, because OS X does layering in different places to us.  
You'll

also find problems from any applications that rely on implementation
details of the OS X frameworks.

David



Right, I know there will be a lot of missing libs that will have to be 
rewritten. But as long as the function calls remain the same API/ABI 
wise (e.g. same arguments...) between GNUstep and AppKit, it could get 
many apps running.


Well, I'll let you know of my progress! If things work out, I think it 
could get more people interested in improving GNUstep.


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: ABI compatibility with Cocoa

[Luboš Doležel]

Hello Ivan,

thanks for the link! I mailed Christine and we'll see.

Luboš

On Mon, 6 Aug 2012 15:47:55 +0200, Ivan Vučica wrote:

Hi Luboš,

Just on a side note, did you see a similar project called "Magenta"?
Author is targeting iOS compatibility instead of OS X, but she
primarily seems to be working on lower-level portions.
  http://crna.cc
I haven't tried it out yet, but you two seem to be working on similar
things, so it might be interesting for you to get in touch.

On 5. 8. 2012., at 15:32, Luboš Doležel  wrote:


Hello,

as part of my diploma thesis I'm working on a userspace Darwin 
compatibility layer for Linux. The primary task is to get unmodified 
console applications running on top of a Wine-like layer. Currently, 
I'm working on a compatible libSystem.dylib reimplementation - i.e. 
resolving ABI issues and implementing missing Mach and BSD kernel 
calls, albeit in many cases it's just a dummy function.


http://darling.dolezel.info/

As a long-term aim, I would also like to get other applications 
running - personally especially 3D games, but ordinary GUI apps as 
well. One of the important pieces is Cocoa and GNUstep seems to be a 
good choice for dealing with AppKit implementation on Linux.


Towards this end, I'd like to ask what you see as the biggest 
hurdles in ABI (not just API) compatibility when replacing Apple's 
AppKit with GNUstep?


I'd like to note that I realize what I'm trying to achieve is a 
*tremendous* task, but well, I believe it can be done!


Regards,
--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

NSString retain crash

[Luboš Doležel]

Hi,

I have a problem where setting NSCell's value to a string crashes. I've 
tracked to problem down to a simple test that reproduces the crash:


NSString* str = @"test";
[str retain];

Backtrace:

#0  objc_msg_lookup_internal (receiver=, 
selector=, receiver=, selector=out>, sender=) at sendmsg2.c:66
#1  objc_msg_lookup (receiver=, selector=out>) at sendmsg2.c:398

#2  0x00400796 in main () at str-retain.m:6

Why does the retain call cause a crash? I've learned that such constant 
strings don't use refcounting internally, but I don't think it should 
crash anyway.


Thanks!
--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: NSThreading

[Luboš Doležel]

On Windows 7, the application IMO needs to actually request admin 
privileges in order to have them.


Try running the app under admin privileges (right click - Run as 
administrator) and it should be OK.


Lubos

On 14.11.2012 10:11, Andor Kocsis wrote:

Thanks to everyone for fast and accurate help. Selector fixed to
"do_it:", and it works now. However still have a warning, and have no
idea, how to avoid it. Programs output now:


2012-11-14 09:13:32.020 helloworld[3096] obj1#tick
2012-11-14 09:13:31.980 helloworld[3096] WARNING - unable to create
shared user
defaults!
2012-11-14 09:13:32.030 helloworld[3096] main#tick


It complains only when NSThread starts - no warning about user defaults
without NSThread. I am with admin rights on my Windows 7 system, and
should not be a problem to create a file, if something library function
really want to do that in depths. Also searched for a user defaults
file, what it should to be on Windows system, thought no problem with
creation, if system already have it. Not so much found.

Can i download one somewhere? Or any other way to fix it?


*From:* Tom Davie 
*To:* Jamison Hope 
*Cc:* Andor Kocsis ; discuss-gnustep@gnu.org
*Sent:* Tuesday, November 13, 2012 11:24 PM
*Subject:* Re: NSThreading


On 13 Nov 2012, at 22:08, Jamison Hope mailto:j...@theptrgroup.com>> wrote:

 > That should be 'NSSelectorFromString(@"do_it:")' with the colon at
the end.
 > At least, that fixes it compiling/running on Cocoa, so I assume it's the
 > same issue for GNUstep.. and it is, look at your error message:

Or, alternatively, @selector(do_it:), which will allow the compiler to
verify that the selector actually exists somewhere.

Bob





___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep




--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Pixen 0.1 Compiles OK (lot of warnings of course) Starts Up and Immediately Crashes

[Luboš Doležel]

On 01/24/2013 08:29 AM, Citizen Jimserac wrote:

On Jan 22, 12:14 pm, Riccardo Mottola 
wrote:

Hi,

which version are you using? The website says it is now a paid appstore app.
I see there is an old port here:

http://home.gna.org/gsimageapps/

I don't know how well it worked

Riccardo


Version 0.1

I see some downloads for older versions here and there but they are
just the ./Pixen.app without the source code.

Since it is now an appstore app, is it OK to still fool with the old
source code??

I don't have a Mac, just linux and and old xp box which serves
perfectly to hold a flowerpot.

J.


The original source code license is still valid... so it should be OK.

I tried to make the current binary work with GNUstep, but it needs the 
WebKit.framework:


$ dyld Contents/MacOS/Pixen

Cannot load 
@loader_path/../Frameworks/Sparkle.framework/Versions/A/Sparkle: Cannot 
load /System/Library/Frameworks/WebKit.framework/Versions/A/WebKit: File 
not found: /System/Library/Frameworks/WebKit.framework/Versions/A/WebKit


WebKit.framework seems to be a hell of an API to reimplement...

--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Pixen 0.1 Compiles OK (lot of warnings of course) Starts Up and Immediately Crashes

[Luboš Doležel]

On 01/24/2013 01:16 PM, Luboš Doležel wrote:

The original source code license is still valid... so it should be OK.

I tried to make the current binary work with GNUstep, but it needs the
WebKit.framework:

$ dyld Contents/MacOS/Pixen

Cannot load
@loader_path/../Frameworks/Sparkle.framework/Versions/A/Sparkle: Cannot
load /System/Library/Frameworks/WebKit.framework/Versions/A/WebKit: File
not found: /System/Library/Frameworks/WebKit.framework/Versions/A/WebKit

WebKit.framework seems to be a hell of an API to reimplement...



Oh and QTKit and CoreAnimation too.

--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: compiling gui with clang/libobjc2 failure

[Luboš Doležel]

On Fri, 1 Feb 2013 08:34:03 +, David Chisnall wrote:

On 1 Feb 2013, at 07:36, Sebastian Reitenbach wrote:


#  ifdef __BLOCKS__
#include 
#  endif /* __BLOCKS__ */


This is definitely a cups problem.  They are assuming that blocks
support (a compiler feature) implies the presence of libdispatch (a
library).  I don't know if libdispatch has been ported to OpenBSD.

I think the -base configure checks for the presence of libdispatch
because we use it for parallel array operations, so the best thing to
do is use that macro to conditionally #undef __BLOCKS__ before
including cups.h

David



Yep, I hit the same problem on Linux. I simply edit the CUPS header 
file and remove the include... Apple, which runs the CUPS development 
now, apparently assumes that blocks support is OS X-only.


Reminds me of a similar issue with unistd.h from glibc that uses 
"__block" for an argument name. And they reject changing the 
(irrelevant) name to something else.


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: make check failing for corebase CFRunLoop

[Luboš Doležel]

Hi,

before you give up on CoreBase, try out my branch at github that 
already has TFB for most types.

https://github.com/LubosD/gnustep-corebase

Should you hit any trouble, patches would indeed by welcome. The aim is 
indeed to merge the work back into subversion.


Lubos

On Thu, 30 May 2013 12:14:24 +1000, Mark Aufflick wrote:

Thanks for the update. My codebase does use some core foundation and
toll free bridging internally, so I'll keep an eye out, but it will 
be

a while before I get that deep into it I suspect!

On 2013-05-30 02:09:53 +, Stefan Bidi said:

Anything that fails in CFRunLoop is a known issue.  Essentially I
wrote the tests, but Ivan to run them and haven't had time to work on
corebase for many months now.  CFRunLoop isn't there except for
stubs.  It's one of those that I'll get to one day when I have more
time.

Also, keep in mind that despite being able to do some basic toll-free
bridging and interaction with base, corebase is still not well
integrated with the rest of GNUstep.  I'd be interested to know if
you actually build any software against it and what bugs you run 
into.



Stef

On Wed, May 29, 2013 at 8:48 PM, Mark Aufflick  wrote:

Hi all,

Still trying to make my way through a full compile of bleeding edge
GNUstep (clang, objc2, gnustep runtime, all from trunk/master) on
Ubuntu 12.04.

I have base and libdispatch all nicely installed (see my version of
the Ubuntu recipe on my blog here: http://mark.aufflick.com/o/1743457
[2] )

Next I configure/make corebase which goes well, but make check yields
these bugs:

CFRunLoop/source.m:

Failed file:     source.m aborted without running all tests!

Failed test:     source.m:35 ... Got main run loop.

Are these known issues, or something I'm doing wrong in my building
process? I can build and run ProjectCenter though, but I guess
NSRunLoop doesn't necessarliy rely on CFRunLoop in GNUstep?

Thanks for any help.

Mark.

___

Discuss-gnustep mailing list

Discuss-gnustep@gnu.org [3]

https://lists.gnu.org/mailman/listinfo/discuss-gnustep [4]



Links:
--
[1] mailto:m...@htb.io
[2] http://mark.aufflick.com/o/1743457
[3] mailto:Discuss-gnustep@gnu.org
[4] https://lists.gnu.org/mailman/listinfo/discuss-gnustep


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Copying native library into app bundle and linking from there

[Luboš Doležel]

On Thu, 4 Jul 2013 14:13:51 +0200, Ivan Vučica wrote:

I think no -- there is probably no built in support as ATM almost all
GNUstep apps are shipped as source code.

Use @rpath on non-OS X and use @executable_path on OS X as part of
the path to the library (in Xcode that's the INSTALL_PATH).

Please google around for further reference; this is somewhat complex
to get right, but is not all THAT difficult.

Even with OS X where this is a common scenario you have to think
carefully about what you are doing -- the defaults for producing
frameworks are not optimized for your scenario, but for producing
frameworks to be placed in /Library/Frameworks (and even then...)

If I'm wrong, someone can correct me

Regards,

Ivan Vučica


I think @rpath is also specific to OS X. In ELF files, I believe you 
don't specify full paths to libraries (just "libobjc.so.4" for example), 
so you can't even put "@rpath" into a library reference.


That being said, ELF does have RPATH/RUNPATH attributes which simply 
influence the search paths. RPATH takes precedence over defaults 
(globals), RUNPATH is checked after defaults.


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Copying native library into app bundle and linking from there

[Luboš Doležel]

On Thu, 4 Jul 2013 14:35:15 +0200, Ivan Vučica wrote:

...what David said.

And for further information (yes, I went and unpacked the laptop in
the bus ;)) see "man 8 ld.so".

Regards,

Ivan Vučica


Oh yes, I didn't know about these extra variables. They seem to be a 
perfect match!
But "@rpath" in referenced lib names are definitely OS X specific. I 
don't really understand why they chose to put full/relative paths into 
library references themselves.


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Darling

[Luboš Doležel]

Yes. Unlike most of the other articles floating around the web - which
are just blindly copied from Phoronix - this one actually goes into
depth and adds some value.

Currently, I'm looking into getting some other stuff working, for now at
a very elementary level: AudioUnit, AudioConverter, CGL, CoreVideo, just
to get (first) mpg123 working and later on Angry Birds for OS X.

Angry Birds seem to be perfect for several reasons:

1) They're cool and everybody knows them. Good for "marketing".

2) Unlike most of applications I was working on, they're not available
on Linux. So no stupid questions why I'm working on Midnight Commander
for OS X.

3) They mostly rely on OpenGL and OpenAL for the real stuff, so they
shouldn't be all that hard to get working.

That being said, the most worrying factor remains gnustep-gui (other
gnustep libs are fine). The screenshots seen at Phoronix[1] come from
*the only* GUI app that ever actually worked with Darling/gnustep-gui. I
could, indeed, and would love to pitch in and help fix the issues in
gnustep-gui, but so far it has proved too complicated.

It made me so desperate that I started considering writing a Cocoa
reimplementation of my own on top of something well proven, such as Qt,
which I know very well and would, at the same time, allow me to avoid
doing a lot of boring low-level work.

[1] http://www.phoronix.com/scan.php?page=news_item&px=MTI0Njc

Luboš

On 08/01/2013 06:57 PM, Gregory Casamento wrote:
> Overall the article was a positive one. :)  I responded to it in the
> comments.
> 
> 
> On Thu, Aug 1, 2013 at 12:33 PM, Liam Proven  <mailto:lpro...@gmail.com>> wrote:
> 
> Darling: Wine for OS X Applications
> 
> by Jon Buys - Jul. 30, 2013
> 
> The latest in the "interesting, but perhaps misguided" department,
> Darling is a project that aims to build a WINE-like environment for
> running OS X applications on Linux.
> 
> http://ostatic.com/blog/darling-wine-for-os-x-applications
> 
> --
> Liam Proven • Profile: http://lproven.livejournal.com/profile
> Email: lpro...@cix.co.uk <mailto:lpro...@cix.co.uk> •
> GMail/G+/Twitter/Flickr/Facebook: lproven
> MSN: lpro...@hotmail.com <mailto:lpro...@hotmail.com> •
> Skype/AIM/Yahoo/LinkedIn: liamproven
> Tel: +44 20-8685-0498  • Cell: +44
> 7939-087884 
> 
> ___
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org <mailto:Discuss-gnustep@gnu.org>
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep
> 
> 
> 
> 
> -- 
> Gregory Casamento
> Open Logic Corporation, Principal Consultant
> yahoo/skype: greg_casamento, aol: gjcasa
> (240)274-9630 (Cell)
> http://www.gnustep.org
> http://heronsperch.blogspot.com
> 
> 
> ___
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep
> 


-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Darling

[Luboš Doležel]

On Fri, 02 Aug 2013 18:20:35 +0200, Fred Kiefer wrote:

Hi Lubos,

sad to hear that GNustep gui is giving you so much trouble. Could you
please expand a bit on what it is that goes wrong for you? There is 
one
big difference between gui and base that is that base already 
supports

the hiding of the instance variables. Would it help you to have that
implemented in gui as well?


Hi,

if you mean adding an extra level of indirection like this:

@interface SomeClass {
   void* allMyIvarsAreHere;
}

then this is definitely needed for 32-bit OS X apps. 64-bit ones are OK 
thanks to the non-fragile ABI.
I've been mostly dealing with 64-bit apps so far, but I'll surely speak 
up when this becomes a problem. I think NSView will be the first class 
that won't "fit".


There are other reasons why base should be easier, for example most 
base

classes are actually class clusters and user code has no expectations
about the actual object layout. Subclassing base classes is far less
common than sunclassing NSView. We cannot do anything about these and
some other differences, but I am willing to support you with things 
that

need changing in gui. But be aware that binary compatibility to Apple
has never been a goal for GNUstep.


The ABI compatibility is actually not such a big deal and GNUstep is 
doing surprisingly well in this regard, more so after the 
NSInteger/NSUInteger size correction.


The problem is the NIB loader (the NSClassSwapper problem and other 
random issues), missing support for short language names ("en.lproj"), 
missing classes (would you accept dummy, empty classes for autolayout 
support?) and other crashes I haven't had time to debug properly. 
Basically, it's all things that you would encounter even when using 
gnustep-gui for native apps.


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Comments on the website from people on twitter...

[Luboš Doležel]

Hi,

if you need marketing, just make OS X, Cocoa, Foundation and Core
Foundation (and later Core Graphics, I hope) implementations in GNUstep
the main topic of the home page.

Honestly, with all the Linux/Unix desktop environments in the world, few
people outside of GNUstep care about GWorkspace or Etoilé, despite how
good they may be.

The same applies to writing GNUstep-based apps *for Linux/Unix*. No
reason to do that when you have Qt and GTK+. Think about Mono - apart
from people under de Icaza's influence, most people wouldn't choose Mono
for native development on Linux/Unix.

"Bring your OS X applications to Linux or Windows" - is the sort of
motto that catches random people's attention.

That's passive marketing. Active marketing would be pushing news like
"Our AppKit is now compatible with OS X 10.x" to sites like Phoronix.

...just my 2 cents. I certainly don't want to force any vision upon you.

Luboš


On 08/09/2013 01:28 AM, Gregory Casamento wrote:
> I just read the article in English.  I believe he removed the claim that
> GNUstep is not maintained because I don't see it in that version of the
> article.  Still, we must get our website redesigned and updated so it
> gives a better impression.
> 
> On Thursday, August 8, 2013, Lars Sonchocky-Helldorf wrote:
> 
> "GNUstep is abandoned" seems to be a common misconception. I read
> this today here:
> 
> http://www.golem.de/news/darling-mac-os-x-anwendungen-unter-linux-1308-100858.html
> 
> The author of that piece (which is mostly a translation of
> 
> http://arstechnica.com/information-technology/2013/08/os-x-apps-run-on-linux-with-wine-like-emulator-for-mac-software/
> ) added only this bit of "information": "Allerdings wird GNUStep
> kaum noch aktiv gepflegt." which translates to: "however, nowadays
> GNUstep is barely maintained". I wrote a mail to the author but
> didn't get a response up to now.
> 
> We definitely need to do something about our public reception.
> 
> 
> cheers,
> 
> Lars
> 
> Am 08.08.2013 um 22:58 schrieb Gregory Casamento:
> 
> > The comment is typical of those I see on a daily basis on twitter...
> >
> > He began by saying the site makes the project look abandoned.  I
> asked what he thought could be improved... his responses are below.
> > ---
> > Keith Smiley @bheron I’d love to help with this if there’s
> something that needs it.
> >
> > @bheron I also think a nice getting started would help. Jumping
> into wiki pages can be a bit daunting. Just a 0 to hello world.
> >
> > @bheron I’m glad you’re still working hard. I think the website
> needs an update. It just looks a little dated the screens as well
> > ---
> > One of our biggest issues is that people just don't know what
> we're all about or that we're still around.
> >
> > We have the best framework out there, but if we don't get the word
> out people just won't use it.

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Comments on the website from people on twitter...

[Luboš Doležel]

Dne 9. srpna 2013 20:36:34 Riccardo Mottola  
napsal:

Nikolaus!

Dr. H. Nikolaus Schaller wrote:
> Am 09.08.2013 um 10:55 schrieb Riccardo Mottola:
>> I put in our developer and people list a link to their respective blog. 
If I missed somebody, please tell!

>>
>> http://www.gnustep.org/developers/whoiswho.html
>>
>> (David, you were completely missing, if you want a better description 
and want a link to one of your blogs, please just send me a private email).

> Developer of SWK and GNUstep Software Index is missing...
>
Heck, you are right, a big omission. Fixed it.

Riccardo


What about Stefan for his corebase work?



___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: GNUstep Kickstarter Project

[Luboš Doležel]

Hi,

cool. I would be pleasantly surprised if the goal is reached, though, 
without some publicity. Try Phoronix or Slashdot.


What would be nice is to show some apps that were ported from OS X to 
Linux/Windows, if there are any.


I have had some limited success with The Unarchiver, I guess the phase 
of moon prevented it from crashing yesterday, but it's crashing again 
today, so I can't take any screenshots. I must create some minimal test 
case to reproduce the error and help get it fixed.


Luboš

On 12.8.2013 23:40, Gregory Casamento wrote:

Goals added.   They are tentative.

Greg


On Mon, Aug 12, 2013 at 5:26 PM, Gregory Casamento
mailto:greg.casame...@gmail.com>> wrote:

Thanks for your input, I'll add some goals in the project
description as soon as they become a little more clear.
Unfortunately I can't say that WebKit will be completed at $100,000
because, to be honest, until the project is actually funded, it's
impossible to tell what can be done.  I've tried to keep the goals
realistic.  At this point, the goal, as stated, is to get GNUstep
fully up to 10.6 and if possible beyond.  Given that Apple is a
multi-billion dollar company and we are not I don't think this goal
is out of the question.  Over promising can get us into a huge
amount of trouble if we don't deliver.

Gregory Casamento
greg.casame...@gmail.com 
Open Logic Corporation, President/CEO
GNUstep Lead Developer


On Aug 12, 2013, at 5:20 PM, James Carthew mailto:jcart...@gmail.com>> wrote:

 > I'm no expert, but I have donated over $1000 to the Star Citizn
kickstarter. They had very clear goals, outlined what they'd do with
more money, and they had video/images of work they had already done
which proved they were serious and capable of delivering goals.
 >
 >
 > On 13 August 2013 07:18, James Carthew mailto:jcart...@gmail.com>> wrote:
 > Good idea, but I think some more planning should be added to
this. Usually when kickstarters succeed they have stretch goals and
very clear presentation of what's currently available, and what will
be available after the project succeeds.. You're going to need a lot
more media on that page. My advice would be to throw up screenshots
of gnustep applications running on gnustep and OSX, show a lot of
comparisons between the apis, show people what you're planning to
create/work on. Saying fund my idea, here's the idea isn't enough
with Kickstarter. You have to really sell the idea you want made.
Sell a vision. Outline clear goals. Say at $100,000 Vespucci/Webkit
will be completed, at $150,000 System Preferences.app will support
networking, X, and audio configuration etc. Clear goals that show
people what they will get out of it. At least put some videos of
gnustep apps on linux and gnustep apps under osx on there.
 >
 >
 > On 13 August 2013 06:31, Gregory Casamento
mailto:greg.casame...@gmail.com>> wrote:
 > Updated, I added a $2 fee to cover shipping outside the US only
because I suspect it will be a little more expensive.
 >
 >
 > On Mon, Aug 12, 2013 at 4:29 PM, Gregory Casamento
mailto:greg.casame...@gmail.com>> wrote:
 > Does it say that?  I'll change it if I can.  Thanks.
 >
 > GC
 >
 >
 > On Mon, Aug 12, 2013 at 4:23 PM, Dr. H. Nikolaus Schaller
mailto:h...@goldelico.com>> wrote:
 >
 > Am 12.08.2013 um 22:13 schrieb Gregory Casamento:
 >
 >> Hey guys, I've launched a kickstarter project in the hopes of
getting some time to work exclusively on GNUstep in order to try to
address some of the issues we have surrounding API completeness,
documentation issues, etc..., here it is:
 >>
 >> http://www.kickstarter.com/projects/203272607/gnustep-project
 >
 > Cool.
 >
 > But why ship only within US?
 >
 > I think it is not a problem to get someone to copy CDs inside the
EU or e.g. Australia and ship them out at low cost as a letter to
save postage.
 >
 > BR,
 > Nikolaus
 >


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: GNUstep Kickstarter Project

[Luboš Doležel]

On Tue, 13 Aug 2013 16:26:46 +1000, Abhi Beckert wrote:

On 13 Aug 2013, at 6:13 am, Gregory Casamento  wrote:


Hey guys, I've launched a kickstarter project in the hopes of
getting some time to work exclusively on GNUstep


I've just linked to this on Sequel Pro's twitter account (open source
mac app with ~50,000 active users). Porting to Windows is our second
most common feature request.

I'm also going to talk to the team about pledging some of the
donations we've received over the years... it's mostly sitting in a
bank account doing nothing right now.

Sounds awesome!
- Abhi



I like the app, so I'll try my best to add missing APIs/frameworks into 
Darling (http://pastebin.com/M6kSMJmR).
There will be some missing APIs on GNUstep's side, though - at least 
something from WebKit and Core Data (what shape is it in?).


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: error building gui

[Luboš Doležel]

On 08/14/2013 10:10 PM, Riccardo Mottola wrote:
> Hi,
> 
> with the latest gui, when building, I get this error:
> 
>  Compiling file NSAlert.m ...
> In file included from ../Headers/Additions/GNUstepGUI/GMAppKit.h:54:0,
>  from NSAlert.m:65:
> ../Headers/AppKit/NSSavePanel.h:85:1: error: expected declaration
> specifiers or '...' before ')' token
> ../Headers/AppKit/NSSavePanel.h:85:1: warning: no semicolon at end of
> struct or union [enabled by default]
> 
> GCC is choking on:
> DEFINE_BLOCK_TYPE(GSSavePanelCompletionHandler, void, NSInteger);
> 
> I suppose this is the macro defined in GSBlocks.h
> 
> perhaps this macro is not working properly? since which version GCC has
> block support?
> 
> Riccardo

Hi,

GCC doesn't have blocks support, that's why these macros exist.

I've fixed GSBlocks.h, both the macro itself and the version checking
that lead to the use of that legacy macro.

Please let me know if it works for you now.

-- 
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: error building gui

[Luboš Doležel]

On Thu, 15 Aug 2013 08:42:14 +0200, Riccardo Mottola wrote:

Hi Luboš

Luboš Doležel wrote:
Hi, GCC doesn't have blocks support, that's why these macros exist. 
I've fixed GSBlocks.h, both the macro itself and the version checking 
that lead to the use of that legacy macro. Please let me know if it 
works for you now. actually, right now, base doesn't compile on gcc at 
all:


Making all for library libgnustep-base...
 Compiling file NSArray.m ...
NSArray.m: In function '-[NSArray 
enumerateObjectsWithOptions:usingBlock:]':
NSArray.m:1743: error: 'BLOCK_SCOPE' undeclared (first use in this 
function)
NSArray.m:1743: error: (Each undeclared identifier is reported only 
once

NSArray.m:1743: error: for each function it appears in.)
NSArray.m:1743: error: expected ';' before 'BOOL'
NSArray.m:1744: warning: ISO C90 forbids mixed declarations and code
NSArray.m:1758: error: 'shouldStop' undeclared (first use in this 
function)



Riccardo


Aaah, it uncovered another problem. Should be fixed.

--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: error building gui

[Luboš Doležel]

On Thu, 15 Aug 2013 12:07:34 +0200, Riccardo Mottola wrote:

Making all for library libgnustep-base...
 Compiling file NSInvocationOperation.m ...
In file included from 
../Headers/Foundation/NSInvocationOperation.h:29,

 from NSInvocationOperation.m:30:
../Headers/Foundation/NSOperation.h:39: error: expected ')' before 
'void'

../Headers/Foundation/NSOperation.h:89: error: expected ')' before
'GSOperationCompletionBlock'
../Headers/Foundation/NSOperation.h:154: error: expected ')' before
'GSOperationCompletionBlock'
gmake[4]: *** [obj/libgnustep-base.obj/NSInvocationOperation.m.o] 
Error 1


perhaps this isagain an error, this time in
DEFINE_BLOCK_TYPE_NO_ARGS(GSOperationCompletionBlock, void);

but it looks consistent to me with the other change you made. Perhaps
you cannot pass "void" as an argument?

Riccardo


Looks like all hell broke loose when I fixed the GCC version detection.
Please retry.

Thanks!
--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: FreeBSD ports

[Luboš Doležel]

For me, GNUstep apps don't show up on the screen half of the time, they're 
hidden behind the previously focused window. I use KDE.


Even apart from startup, the focus behaves weird at times.

Luboš


Dne 25. září 2013 11:52:01 Ivan Vučica  napsal:

@David:

GL transparency was sort-of fixed at the Cambridge meeting. Look at commits 
I made to gnustep-back during that time.


I don't know anything about the z-ordering issues; WMs do a good job with that.

Regards,

Ivan Vučica
via phone

> On 25 Sep 2013, at 11:20, "Lundberg, Johannes" 
 wrote:
> The window level seems to something wrong with as well. It works in 
relation to for example xterm, but not with other gnustep windows, 
z-ordering doesn't do what it is suppose to do... Maybe I can provide some 
more info if I examine some more. I tried both the port and from svn. 
Perhaps the problem is elsewhere?...

> --
> Johannes Lundberg
> Project leader and lead developer of Mirama OS (previously Viking OS)
> BRILLIANTSERVICE CO., LTD.
> My blog
> Mirama homepage blog
> Company homepage
>
>> On Wed, Sep 25, 2013 at 4:51 PM, David Chisnall  
wrote:
>> The port is the latest release.  I'm not sure when the fix happened, but 
if it was after the release then it won't have made it in.  If someone 
(Fred?) can point me to the svn revision, I can add the fix as a patch in 
the port.

>> David
>> On 25 Sep 2013, at 07:01, "Lundberg, Johannes" 
 wrote:

>> > Hi
>> >
>> > Just installed FreeBSD 10.0 alpha 2 and GNUstep from ports.
>> >
>> > Installing worked like a charm. So easy and painful .Thank you!! :)
>> >
>> > So far almost everything seems good except one thing. Transparency on 
OpenGL views won't work.

>> >
>> > I downloaded gnustep-back from svn, compiled and installed over the 
one from ports and it works fine now.

>> >
>> > Is the port version an older version which doesn't contain this 
improvement?

>> >
>> > --
>> > Johannes Lundberg
>> > Project leader and lead developer of Mirama OS (previously Viking OS)
>> > BRILLIANTSERVICE CO., LTD.
>> >
>> > My blog
>> > Mirama homepage blog
>> > Company homepage
>> > ___
>> > Discuss-gnustep mailing list
>> > Discuss-gnustep@gnu.org
>> > https://lists.gnu.org/mailman/listinfo/discuss-gnustep
>> > > ___
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep
___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Vertical refresh and tearing...

[Luboš Doležel]

I believe that vsync is typically controlled by X driver's settings. It 
is not application's responsibility to make such decisions.


Luboš

On Thu, 9 Jan 2014 21:07:46 +0900, Lundberg, Johannes wrote:

Hi

Yeah it is something I made myself and something I can not share. If
I have some time I could write a short program that would demonstrate
the tearing but I'm rather busy now so that is not so likely I 
think..

However, I think the person who wrote the backend should be able to
answer the question easy..

Actually it is almost oyasumi (bed time) :)

--Johannes Lundberg
BRILLIANTSERVICE CO., LTD.

On Thu, Jan 9, 2014 at 9:02 PM, a b  wrote:


I am new to this group so I will not attempt to answer your
question. I have a question
for you:

Is the application something you downloaded from the GNUstep
website? If so,
could you say which application it is so that others could look at
at in order to see
the effect you are observing?

If it is something you wrote, would you be able to share a short
example demo
which would allow others to determine what is causing the problem?

ohayo or konichiwa as the case may be (please excuse the spelling,
if incorrect)

-
From: johan...@brilliantservice.co.jp [1]
Date: Thu, 9 Jan 2014 20:25:04 +0900
Subject: Vertical refresh and tearing...
To: discuss-gnustep@gnu.org [2]

Hi

Currently it does not seem like rendering is synchronized with
vertical refresh. Both my OpenGL (NSOpenGLView) app and a app using
animated NSView has a lot of tearing (on FreeBSD with cairo
backend).

Has anyone similar experiences or insight in what would be required
to get rid of the tearing? Is it simply a matter of making

additions

to the code in the backend?

Best regards

--Johannes Lundberg

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-



秘密保持について：この電子メールは、名宛人に送信したものであり、秘匿特権の対象となる情報を含んでいます。




もし、名宛人以外の方が受信された場合、このメールの破棄、およびこのメールに関する一切の開示、




複写、配布、その他の利用、または記載内容に基づくいかなる行動もされないようお願い申し上げます。

---
CONFIDENTIALITY NOTE: The information in this email is confidential
and intended solely for the addressee.
Disclosure, copying, distribution or any other action of use of
this
email by person other than intended recipient, is prohibited.
If you are not the intended recipient and have received this email
in
error, please destroy the original message.
___ Discuss-gnustep
mailing list Discuss-gnustep@gnu.org [3]
https://lists.gnu.org/mailman/listinfo/discuss-gnustep [4]
___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org [5]
https://lists.gnu.org/mailman/listinfo/discuss-gnustep [6]


 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
秘密保持について：この電子メールは、名宛人に送信したものであり、秘匿特権の対象となる情報を含んでいます。
もし、名宛人以外の方が受信された場合、このメールの破棄、およびこのメールに関する一切の開示、
複写、配布、その他の利用、または記載内容に基づくいかなる行動もされないようお願い申し上げます。
---
CONFIDENTIALITY NOTE: The information in this email is confidential
and intended solely for the addressee.
Disclosure, copying, distribution or any other action of use of this
email by person other than intended recipient, is prohibited.
If you are not the intended recipient and have received this email in
error, please destroy the original message.

Links:
--
[1] mailto:johan...@brilliantservice.co.jp
[2] mailto:discuss-gnustep@gnu.org
[3] mailto:Discuss-gnustep@gnu.org
[4] https://lists.gnu.org/mailman/listinfo/discuss-gnustep
[5] mailto:Discuss-gnustep@gnu.org
[6] https://lists.gnu.org/mailman/listinfo/discuss-gnustep
[7] mailto:a_brig...@hotmail.com


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Problem with catching exceptions on ARM

[Luboš Doležel]

On Mon, 03 Feb 2014 15:55:00 +0100, Mathias Bauer wrote:



Unfortunately it turned out that neither the 3.4 nor the trunk
version of clang are usable on arm at all:

http://llvm.org/bugs/show_bug.cgi?id=18622

Regards,
Mathias


Hi,

David Chisnall gave me a piece of advice that adding "-integrated-as" 
into CFLAGS helps. And it does (although it still is a bug).


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Is a move to github in order??

[Luboš Doležel]

On 12.5.2014 21:22, Gregory Casamento wrote:

Hey guys,

With the recent move of Etoile to github I've been thinking about a
similar move for gnustep.

There are a number of reasons for this:

0)
​It is possible to check out and commit to a github repository using SVN
and HG.  This is #0 since it will be a primary objection to such a move.​
1) GitHub is a very well known platform
2) it would allow code review of patches to be done more easily
3) we could host both the wiki and the website on GitHub.
4) GitHub users could easily fork our repository, reuse and contribute
back more easily (we would of course still need FSF papers)

Please give me your thoughts on this. I think it's a good idea for the
technical reasons I've given above.



Hi,

yes, please, do that. The "social" aspect of GitHub is very beneficial 
for open source projects.


--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Cider: Native Execution of iOS Apps on Android

[Luboš Doležel]

On Wed, 14 May 2014 14:48:35 -0400, David Chisnall wrote:


They do some nice tricks to graft the XNU code into Linux, but it's
conceptually quite similar to the foreign ABI layers most *NIX 
systems

have, although they have per-thread kernel-ABI contexts to allow a
single program to make both XNU and Linux system calls, depending on
where it is.  The other quite neat thing is their library that allows
calling between libraries using slightly different ABIs.  This is not
a new technique either (Transitive / Manchester pioneered a lot of
this work over a decade ago), but it means that you can use native
Android libGL (for example) from an iOS app.

It's a research project not a serious product though, because it
relies on using a load of Apple libraries that they don't have a
license to use on Android.


Exactly. To sum it up: to use their work, your phone's kernel must be 
modified and you must illegally use Apple's libraries, because Cider 
only provides kernel interfaces (e.g. surfaces for drawing, so the UIKit 
for instance needs to be copied from iOS).


With this in mind, I don't see why they consider the Wine/Darling 
approach so unworkable.


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Copyright assignment requirement

[Luboš Doležel]

On Tue, 3 Jun 2014 09:10:24 +0100, Richard Frith-Macdonald wrote:

On the issue of copyright my feeling is similar but different;
I know people hate the admin process of doing the copyright
assignment (much more than people disliking the principle, though
there are a few of those too), but we need the copyright to be
assigned to the FSF (or some similar arrangement to ensure that all
code in the project is legitimately released under the LGPL license 
we

are using) because otherwise the companies I deal with really might
have a problem using GNUstep.


In my opinion, FSF should seriously reconsider the assignment process, 
because it is invalid for most of Europe probably.
Around here, I can grant you an exclusive license with the permission 
to (re)license the code, but I can never assign the copyright to someone 
else. This is only possible if I were an employee of FSF or if I died 
and assigned the rights in my last will.


--
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: What would be the most complete GNUStep system?

[Luboš Doležel]

> As for DMGs, I know that you feel they are convenient and how they very
> easily can be virtually monuted, burned onto optical media or (in old
> times) to floppies. I don't think we have support for that though and
> how it could be implemented in a portable way.

darling-dmg offers a FUSE module.

https://github.com/LubosD/darling-dmg

-- 
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: What would be the most complete GNUStep system?

[Luboš Doležel]

On 10/27/2014 07:44 PM, Gregory Casamento wrote:
> On Mon, Oct 27, 2014 at 2:10 PM, Luboš Doležel  wrote:
>>> As for DMGs, I know that you feel they are convenient and how they very
>>> easily can be virtually monuted, burned onto optical media or (in old
>>> times) to floppies. I don't think we have support for that though and
>>> how it could be implemented in a portable way.
>>
>> darling-dmg offers a FUSE module.
>>
>> https://github.com/LubosD/darling-dmg
> 
> Here's another link for DMG reading...
> 
> http://sourceforge.net/projects/dmgmount/
> 

Well, this projects seems to convert DMG into an ordinary disk image,
which means a lot of useless copying.

darling-dmg behaves more like OS X - DMG is not extracted, but used
directly.

-- 
Luboš Doležel


___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: What would be the most complete GNUStep system?

[Luboš Doležel]

On 10/27/2014 10:04 PM, Asiga Nael wrote:
> Btw, nobody suggested Darwin. It's open-source, and AFAIK it supports
> fat binaries natively (I don't think fat binaries are an OSX
> addition, they must be in Darwin already). Maybe it can even mount
> DMGs natively (I don't know if DMG are a Darwin or an OSX thing).
> Regarding app bundles, that's a desktop thing, not an OS thing, so
> that can be done from GNUstep. So, I tend to believe Darwin with
> GNUstep would have every feature I wish to have in my everyday OS.
> 
> And, btw, PureDarwin, which seems to be an easy-to-install Darwin
> distribution, supports MacPorts natively. Given that GNUstep is
> supported in MacPorts, this means maybe you can have Darwin+GNUstep
> working today (there're GNUstep screenshots at the PureDarwin
> website).
> 
> Is there any reason for being cautious about Darwin?
> 
> asiga

Darwin = XNU + userspace (partially taken from GNU)

XNU = Mach kernel with BSD stuff added, twisted into a mixture, where
some of the Mach stuff is broken and some of the BSD stuff is broken as
well.

You can't even boot a functional Darwin system without closed source
drivers from Apple.

So as long as you're looking for a "desktop", there is no reason to
bother about the underlying kernel. And Darwin is NOT a good choice of a
kernel at all.

-- 
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep

Re: Window stacking order

[Luboš Doležel]

On Tue, 11 Nov 2014 10:21:13 -0500, Charles Philip Chan wrote:

On 11 Nov 2014, johan...@brilliantservice.co.jp wrote:

I tried several different window managers on Linux and on all of 
them

I can not get GNUstep apps to play nicely with other apps when it
comes to windows stacking order.

Currently I am using openbox and trying to configure an always on 
top

X11 app but the GNUstep app gets on top when given focus even though
it shouldn't. Other apps (stock X11 apps) do not get on top when
focused.


Don't know about other window manager, but it works fine for me in
Window Maker.

Charles


I can confirm some strange focus issues with KWin.

--
Luboš Doležel

___
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnustep