headers
Christopher Frauenberger | 1 Mar 2007 01:38
Picon

Re: compiling PySClang

Hi,

this should be fixed now (along with the relative library path).  
Please checkout the new version of the Xcode project. I also added a  
section about compilation to the Readme file. Any additions /  
experience reports are welcome (thanks Jonathan, I added parts of  
your little howto)

Chris

On 28 Feb 2007, at 14:04, sonify wrote:

> Hello gentle folks,
>
> scsynth and sclang build ok.
>
> I'm trying to compile supercollider/PySCLang using   
> PySCLang.xcodeproj with Xcode 2.4.1 on an intel OSX10.4.8
> (SVN revision number 5882, though the problem below hasn;t changed  
> since SVN 5871 of last week)
>
> The code compiles without error, When attempting to link
> __Z20initSerialPrimitivesv is thrown up as an undefined symbol .  
> Can't find a reference to it.
>
> Xcode output at that point is:
>
> ld: Undefined symbols:
> __Z20initSerialPrimitivesv
> /scbuild/supercollider/Psycollider/PySCLang/build/PySCLang.build/
(Continue reading)

Permalink | Reply | 3 comments |
headers
sonify | 1 Mar 2007 01:57
Picon

Re: compiling PySClang

Hi Chris - thanks for this. I assumed you were busy and someone else  
may have been able to help
I'm getting an Xcode open error:
 > Project /scbuild/supercollider/Psycollider/PySCLang/ 
PySCLang.xcodeproj cannot be opened because the project file cannot  
be parsed.
would you please check that the svn version works for you - or  
possibly an upload error?

thanks a million,

David

On 01/03/2007, at 11:38 AM, Christopher Frauenberger wrote:

> Hi,
>
> this should be fixed now (along with the relative library path).  
> Please checkout the new version of the Xcode project. I also added  
> a section about compilation to the Readme file. Any additions /  
> experience reports are welcome (thanks Jonathan, I added parts of  
> your little howto)
>
> Chris
>
>
> On 28 Feb 2007, at 14:04, sonify wrote:
>
>> Hello gentle folks,
>>
(Continue reading)

Permalink | Reply | 2 comments |
headers
Rohan Drape | 1 Mar 2007 06:01

inf/nan and g++ (was: demand ugens and inf on x86)

Dear Stefan & List,

On Wed Jan 18 12:19:46 PST 2006, stefan kersten wrote:

> On Wed, Jan 18, 2006 at 10:06:31PM +1100, Rohan Drape wrote:
> > Also, does anyone else have this issue with inf on x86?  I used to
> > think I had broken it with some local changes but it's not working on
> > what I am pretty sure is a clean tree?

> this is a known issue, haven't had the time to have a look
> yet ... any ideas?

The transcript below suggests we ought to drop -ffast-math!

It also suggests sc_isnan can be written more simply as x!=x.

Deleting -ffast-math from SConstruct fixes the problem I had with
doneActions in tDuty not being run, but doesn't seem to fix the inf
problem in sclang, though perhaps scons didn't catch everything.

Regards,
Rohan

PS. The g++ manual here says that -ffast-math sets -fno-math-errno,
-funsafe-math-optimizations, -fno-trapping-math, -ffinite-math-only,
-fno-rounding-math, -fno-signaling-nans and fcx- limited-range.  I'd
guess it is -fno-signaling-nans and -ffinite-math-only that are
causing headaches, so a keen makefile writer might like to keep some
of the others, but I've not checked this.
(Continue reading)

Permalink | Reply | 1 comment |
headers
Christopher Frauenberger | 1 Mar 2007 10:39
Picon

Re: compiling PySClang

Thats weird, working fine here (Xcode 2.4)

Try to delete the project before your update, maybe subversion merges  
files and this is what corrupts it....

Has anyone else that problem too??

Chris

On 1 Mar 2007, at 00:57, sonify wrote:

> Hi Chris - thanks for this. I assumed you were busy and someone  
> else may have been able to help
> I'm getting an Xcode open error:
> > Project /scbuild/supercollider/Psycollider/PySCLang/ 
> PySCLang.xcodeproj cannot be opened because the project file cannot  
> be parsed.
> would you please check that the svn version works for you - or  
> possibly an upload error?
>
> thanks a million,
>
> David
>
> On 01/03/2007, at 11:38 AM, Christopher Frauenberger wrote:
>
>> Hi,
>>
>> this should be fixed now (along with the relative library path).  
>> Please checkout the new version of the Xcode project. I also added
(Continue reading)

Permalink | Reply | 1 comment |
headers
sonify | 1 Mar 2007 15:26
Picon

Re: compiling PySClang

Hi Chris, all,

Yes, that did the trick  - stupid me!
I should have guessed that, because PySCLang.xcodeproj is a directory  
so SVN is may just looking for file changes within that directory,
Anyway, deleting PySCLang.xcodeproj before SVN is the way to go. I  
wonder whether this might be a good approach for  other . xcodeproj  
as well?

The .so imports and "looks" OK. I'll do some testing over the weekend  
and report bugs, if any.

thanks for your help,

David
(also using Xcode 2.4.1)

On 01/03/2007, at 8:39 PM, Christopher Frauenberger wrote:

> Thats weird, working fine here (Xcode 2.4)
>
> Try to delete the project before your update, maybe subversion  
> merges files and this is what corrupts it....
>
> Has anyone else that problem too??
>
> Chris
>
> On 1 Mar 2007, at 00:57, sonify wrote:
>
(Continue reading)

Permalink | Reply | 0 comments |
headers
Julian Rohrhuber | 1 Mar 2007 14:57
Picon
Favicon

flop for methods

The possibility to flop functions has become very useful in many cases.

// eg.

a = (0..12).scramble;

g = { |x, y| a.indexOf((a[x] + a[y] % a.size)) }.flop;

g.(a, a);

Would it be possible to add this functionality to methods?

This could encourage to make sc even more expanding than it already is.
--

-- 

.
Permalink | Reply | 0 comments |
headers
James McCartney | 1 Mar 2007 20:37
Picon

Re: flop for methods

That would be interesting, but it isn't easily possible automatically.
SC doesn't currently allow you to compose functions for methods. e.g.:

myClass.addMethod('methodName', {|args| expression }.flop);

SC classes are constructed statically for speed reasons. Now that
computers are faster, this could possibly be relaxed to be more
dynamic, but it would be a low level reworking.

On 3/1/07, Julian Rohrhuber <rohrhuber@...> wrote:
> The possibility to flop functions has become very useful in many cases.
>
> // eg.
>
> a = (0..12).scramble;
>
> g = { |x, y| a.indexOf((a[x] + a[y] % a.size)) }.flop;
>
> g.(a, a);
>
> Would it be possible to add this functionality to methods?
>
> This could encourage to make sc even more expanding than it already is.
> --
>
>
>
>
>
> .
(Continue reading)

Permalink | Reply | 3 comments |
headers
Stefan Kersten | 1 Mar 2007 23:00
Picon

Re: inf/nan and g++ (was: demand ugens and inf on x86)

On 01.03.2007, at 06:01, Rohan Drape wrote:
> Dear Stefan & List,
>
> On Wed Jan 18 12:19:46 PST 2006, stefan kersten wrote:
>
>> On Wed, Jan 18, 2006 at 10:06:31PM +1100, Rohan Drape wrote:
>>> Also, does anyone else have this issue with inf on x86?  I used to
>>> think I had broken it with some local changes but it's not  
>>> working on
>>> what I am pretty sure is a clean tree?
>
>> this is a known issue, haven't had the time to have a look
>> yet ... any ideas?
>
> The transcript below suggests we ought to drop -ffast-math!
>
> It also suggests sc_isnan can be written more simply as x!=x.
>
> Deleting -ffast-math from SConstruct fixes the problem I had with
> doneActions in tDuty not being run, but doesn't seem to fix the inf
> problem in sclang, though perhaps scons didn't catch everything.

thanks rohan, i'll fix it this weekend. but i guess it doesn't  
explain the inf problem as a repeat count in Demand ugens? this  
problem also appears on intel/osx. unless the xcode project uses - 
ffast-math, too ...

<sk>
Permalink | Reply | 0 comments |
headers
Rohan Drape | 2 Mar 2007 03:51

inf/nan and g++ (was: demand ugens and inf on x86)

On Thu Mar 1 23:00:19 PST 2007 Stefan Kersten wrote:
> thanks rohan, i'll fix it this weekend. but i guess it doesn't  
> explain the inf problem as a repeat count in Demand ugens? this  
> problem also appears on intel/osx. unless the xcode project uses - 
> ffast-math, too ...

Yes, the 'inf' as repeat count still fails.

I've been using forever=1e+9.

1e+10 doesn't work, I assume because the repeat counts are stored as
int32 values, and 1e+10 > 2^31.

DSeq_next and Dwhite_next etc. read:

  float x = DEMANDINPUT(0);
  unit->m_repeats = sc_isnan(x) ? 0 : (int32)floor(x + 0.5f);

I don't really understand how this works on ppc, perhaps
(int32)floor(inf+0.5) is 2^31?  It doesn't seem to be on gcc/intel,
the results I receive from the test file below are:

  (int32_t)floor(inf + 0.5) = -2147483648
  (int32_t)floor(1e+9 + 0.5) = 1000000000
  (int32_t)floor(1e+10 + 0.5) = -2147483648

Regards,
Rohan

++
(Continue reading)

Permalink | Reply | 2 comments |
headers
Till Bovermann | 2 Mar 2007 21:39

Re: propose: additional Color support

they are broken in terms of the alpha values...
it is not preserved... (but intends to do so...)

I'll check in the additions on Monday.

On 27.02.2007, at 19:33, Julian Rohrhuber wrote:

>>> No one objects, which means I'll check this in in near future...
>>>
>>> regards
>>> Till
>>
>> please do, they look useful to me.
>
> yes, agreed.
> But you mentioned that Color - add / subtract were broken - they  
> seem not to be broken here.
> -- 
>
>
>
>
>
> .
> _______________________________________________
> sc-dev mailing list
> sc-dev@...
> http://www.create.ucsb.edu/mailman/listinfo/sc-dev
Permalink | Reply | 0 comments |

Gmane