On Apr 24, 2015, at 10:25 PM, Miller Puckette <msp@ucsd.edu> wrote:

Finally but the bullet and tried this...

I made a branch, 0.46, adn committeed that and one other minor bug fix.
The commit is 905695a350bf44d906fbb8cde905fa0b81d12b57

Too early in the game to make a 0.46-7 but I think if you track the 0.46 branch
you'll be OK.

cheers
Miller

On Thu, Apr 23, 2015 at 12:01:49AM -0400, Dan Wilcox wrote:

Actually yeah, looks good. Is there anyway this particular patch can be backported to the 0.46-6 release? Or will there be a 0.46-7 coming up soon? I ask because I’m having libpd checkout pure-data release tags. Not a huge deal if not, since this is still in testing.

Just to confirm:

Current git log for the pure-data sources I’m using:

danomatika:pure-data dano$ git log
commit 35151c473827260c86cc864cc4cd058b24e67139
Author: Miller Puckette <msp@ucsd.edu>
Date:   Wed Apr 22 14:06:36 2015 -0700

   make list of signals local to a pd instance to fix libpd crasher

Performed:

cd libpd
make clobber
make
cd samples/c/pdtest_multi
make clobber
make
./pdtest_multi test.pd ./

Got:

start 80c04a90
stop 80c04a90
continue start 80c04a90
stop 80c04a90
print: 0
1003-frequency: bang
start 80c04a90
stop 80c04a90
continue start 80c04a90
start 80c04db0
stop 80c04db0
continue start 80c04db0
stop 80c04db0
print: 0
1004-frequency: bang
start 80c04db0
stop 80c04db0
continue start 80c04db0
1003-frequency: 1
1004-frequency: 2
1.000000 1.000000 0.999999 0.999999 0.999998 0.999998 0.999997 0.999997 
1.000000 1.000000 0.999998 0.999998 0.999996 0.999996 0.999995 0.999995 
print: 1
0.999944 0.999944 0.999943 0.999943 0.999942 0.999942 0.999941 0.999941 
print: 1
0.999815 0.999815 0.999810 0.999810 0.999804 0.999804 0.999799 0.999799 
print: 2
print: 2

I got the same response with building libpd with & without -03.

--------
Dan Wilcox
@danomatika
danomatika.com <http://danomatika.com/>
robotcowboy.com <http://robotcowboy.com/>

On Apr 22, 2015, at 8:27 PM, Miller Puckette <msp@ucsd.edu> wrote:

Hmm... and (just checking) - you did a make clean and rebuilt everything?
The pd_instance structure changed....

I can't get it to fail over here any more...

cheers
Miller

On Wed, Apr 22, 2015 at 05:30:20PM -0400, Dan Wilcox wrote:

Thanks. Unfortunately still a crash, but now at a different place. I also confirmed that it still runs fine when libpd is compiled with -O3 aka make DEBUG=true.

#0  signal_cleanup [inlined] () at /Users/dano/src/pd/libpd/pure-data/src/d_ugen.c:364
#1  0x0000000100026680 in ugen_stop () at d_ugen.c:562
#2  0x000000010002670e in ugen_start () at d_ugen.c:569
#3  0x000000010003197e in glob_dsp (dummy=0x100102e00, s=0x0, argc=<value temporarily unavailable, due to optimizations>, argv=0x100102e20) at g_canvas.c:1086
#4  0x000000010006d286 in pd_typedmess (x=0x1000a5760, s=0x0, argc=1055232, argv=0x100102e20) at m_class.c:699
#5  0x0000000100096fb8 in libpd_message [inlined] () at /Users/dano/src/pd/libpd/libpd_wrapper/z_libpd.c:236
#6  0x0000000100096fb8 in libpd_finish_message (recv=<value temporarily unavailable, due to optimizations>, msg=0x100000f67 "dsp") at z_libpd.c:562
#7  0x0000000100000b8f in main ()

In detail:

#0  signal_cleanup [inlined] () at /Users/dano/src/pd/libpd/pure-data/src/d_ugen.c:364
364         pd_this->pd_signals = sig->s_nextused;
(gdb) frame 1
#1  0x0000000100026680 in ugen_stop () at d_ugen.c:562
364         pd_this->pd_signals = sig->s_nextused;
(gdb) frame 2
#2  0x000000010002670e in ugen_start () at d_ugen.c:569
569     ugen_stop();
(gdb) frame 3
#3  0x000000010003197e in glob_dsp (dummy=0x100102e00, s=0x0, argc=<value temporarily unavailable, due to optimizations>, argv=0x100102e20) at g_canvas.c:1086
1086     ugen_start();
(gdb) frame 4
#4  0x000000010006d286 in pd_typedmess (x=0x1000a5760, s=0x0, argc=1055232, argv=0x100102e20) at m_class.c:699
699             else (*((t_messgimme)(m->me_fun)))(x, s, argc, argv);
(gdb) frame 5
#5  0x0000000100096fb8 in libpd_message [inlined] () at /Users/dano/src/pd/libpd/libpd_wrapper/z_libpd.c:236
236   pd_typedmess(dest, gensym(msg), n, v);
(gdb) frame 6
#6  0x0000000100096fb8 in libpd_finish_message (recv=<value temporarily unavailable, due to optimizations>, msg=0x100000f67 "dsp") at z_libpd.c:562
236   pd_typedmess(dest, gensym(msg), n, v);
(gdb) frame 7
#7  0x0000000100000b8f in main ()

--------
Dan Wilcox
@danomatika
danomatika.com <http://danomatika.com/> <http://danomatika.com/ <http://danomatika.com/>>
robotcowboy.com <http://robotcowboy.com/> <http://robotcowboy.com/ <http://robotcowboy.com/>>

On Apr 22, 2015, at 5:12 PM, Miller Puckette <msp@ucsd.edu> wrote:

OK ... I think I fixed it... pull newest from git and try.

thanks
M

On Wed, Apr 22, 2015 at 12:35:10PM -0700, Miller Puckette wrote:

Aha - sometimes things crash on OSX because of "minor" buffer overruns that
linux just soldiers past.  I'll break out a mac and try again...

M

On Wed, Apr 22, 2015 at 03:03:09AM -0400, Dan Wilcox wrote:

Weird. It crashes for me on OSX when libpd is built with -O3 and the program is built with or without -O3.

*Note: I just finished reorganizing the libpd samples by language name. If you do a pull, the multi instance test is now located in samples/c/pdtest_multi 

Rebuilding and running it again yields:

Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: 13 at address: 0x0000000000000000
outlet_float (x=0x3f7ffc2a3f7ffc38, f=0) at m_obj.c:388
388     for (oc = x->o_connections; oc; oc = oc->oc_next)

#0  outlet_float (x=0x3f7ffc2a3f7ffc38, f=0) at m_obj.c:388
#1  0x000000010006f0b9 in outlet_bang (x=<value temporarily unavailable, due to optimizations>) at m_obj.c:363
#2  0x00000001000953f9 in metro_tick (x=0x10021ad90) at x_time.c:162
#3  0x0000000100071010 in sched_tick () at m_sched.c:418
#4  0x0000000100096733 in libpd_process_float (ticks=1, inBuffer=<value temporarily unavailable, due to optimizations>, outBuffer=<value temporarily unavailable, due to optimizations>) at z_libpd.c:185
#5  0x0000000100000d67 in main ()

some frame detail:

(gdb) frame 0
#0  outlet_float (x=0x3f7ffc2a3f7ffc38, f=0) at m_obj.c:388
388     for (oc = x->o_connections; oc; oc = oc->oc_next)
(gdb) frame 1
#1  0x000000010006f0b9 in outlet_bang (x=<value temporarily unavailable, due to optimizations>) at m_obj.c:363
363         pd_bang(oc->oc_to);
(gdb) frame 2
#2  0x00000001000953f9 in metro_tick (x=0x10021ad90) at x_time.c:162
162     outlet_bang(x->x_obj.ob_outlet);
(gdb) frame 3
#3  0x0000000100071010 in sched_tick () at m_sched.c:418
418         (*c->c_fn)(c->c_owner);
(gdb) frame 4
#4  0x0000000100096733 in libpd_process_float (ticks=1, inBuffer=<value temporarily unavailable, due to optimizations>, outBuffer=<value temporarily unavailable, due to optimizations>) at z_libpd.c:185
185   PROCESS(,)
(gdb) frame 5
#5  0x0000000100000d67 in main ()

--------
Dan Wilcox
@danomatika
danomatika.com <http://danomatika.com/>
robotcowboy.com <http://robotcowboy.com/>

On Apr 22, 2015, at 12:15 AM, Miller Puckette <msp@ucsd.edu> wrote:

I think it was optimized since I had already made libpd, not from the
sampes/.../multi directory.  But anyhow I re-did it as you suggest with
the same result....  can't make it fail...

cheers
M

On Wed, Apr 22, 2015 at 12:03:28AM -0400, Dan Wilcox wrote:

You ran it without the optimizations since I added the debug option. Remove DEBUG=true from line 33 in the Makefile: https://github.com/libpd/libpd/blob/master/samples/c_samples/multi/Makefile#L33 <https://github.com/libpd/libpd/blob/master/samples/c_samples/multi/Makefile#L33> and do a full clean before rebuilding:

cd ../../../ && make clobber && cd - && make
--------
Dan Wilcox
@danomatika
danomatika.com <http://danomatika.com/>
robotcowboy.com <http://robotcowboy.com/>

On Apr 21, 2015, at 11:57 PM, Miller Puckette <msp@ucsd.edu> wrote:

Hi Dan et al -

I gave this a try:

git clone https://github.com/libpd/libpd.git

[copies pd sources into libpd/pure-data]

cd libpd
make

cd samples/c_samples/multi/
make
./multi_pdtest multi_test.pd `pwd`

and got output:

print: 0
1003-frequency: bang
print: 0
1004-frequency: bang
1003-frequency: 1
1004-frequency: 2
1.000000 1.000000 0.999999 0.999999 0.999998 0.999998 0.999997 0.999997 
1.000000 1.000000 0.999998 0.999998 0.999996 0.999996 0.999995 0.999995 
print: 1
0.999944 0.999944 0.999943 0.999943 0.999942 0.999942 0.999941 0.999941 
print: 1
0.999815 0.999815 0.999810 0.999810 0.999804 0.999804 0.999799 0.999799 
print: 2
print: 2

This on Fedora 21, 64 bits, Intel hardware.  

I guess something subtle is happening, maybe in Pd and unrelated to libpd?

cheers
Miller

On Tue, Apr 21, 2015 at 06:00:15PM -0400, Dan Wilcox wrote:

Howdy Miller,

Following up from the dev list last year, I added your multi instance test to the c samples included with libpd: https://github.com/libpd/libpd/tree/master/samples/c_samples/multi <https://github.com/libpd/libpd/tree/master/samples/c_samples/multi>

The one thing I want to double check is the changes to z_libpd.c you mentioned in http://lists.puredata.info/pipermail/pd-dev/2014-05/019832.html: <http://lists.puredata.info/pipermail/pd-dev/2014-05/019832.html:>

Here's how I modified libpd_wrapper/z_libpd.c:

55d54
<   sys_time = 0;
110c109
<   sched_tick(sys_time + sys_time_per_dsp_tick);
---

sched_tick();

130c129
<     sched_tick(sys_time + sys_time_per_dsp_tick); \
---

sched_tick(); \

Currently, that line is https://github.com/libpd/libpd/blob/master/libpd_wrapper/z_libpd.c#L171 <https://github.com/libpd/libpd/blob/master/libpd_wrapper/z_libpd.c#L171> and I’m getting a segfault if I replace it with sched_tick(); AND set gcc optimization to -O3

Here’s a gdb backtrace:

#0  0x0000000100091f2f in outlet_float (x=0x3f7ffc2a3f7ffc38, f=0.999937057) at m_obj.c:388
#1  0x00000001000ac572 in pdfloat_bang (x=0x10021ac20) at x_connective.c:89
#2  0x0000000100093d58 in pd_bang (x=0x10021ac20) at m_pd.c:267
#3  0x0000000100091ddd in outlet_bang (x=0x10021ae20) at m_obj.c:363
#4  0x00000001000c45e4 in metro_tick (x=0x10021ada0) at x_time.c:162
#5  0x0000000100095021 in sched_tick () at m_sched.c:418
#6  0x00000001000c5d4d in libpd_process_float (ticks=1, inBuffer=0x7fff5fbffa50, outBuffer=0x7fff5fbff750) at z_libpd.c:173
#7  0x0000000100000d18 in main ()

If I don’t optimize, it works fine:

print: 0
1003-frequency: bang
print: 0
1004-frequency: bang
1003-frequency: 1
1004-frequency: 2
1.000000 1.000000 0.999999 0.999999 0.999998 0.999998 0.999997 0.999997 
1.000000 1.000000 0.999998 0.999998 0.999996 0.999996 0.999995 0.999995 
print: 1
0.999944 0.999944 0.999943 0.999943 0.999942 0.999942 0.999941 0.999941 
print: 1
0.999815 0.999815 0.999810 0.999810 0.999804 0.999804 0.999799 0.999799 
print: 2
print: 2

--------
Dan Wilcox
@danomatika
danomatika.com <http://danomatika.com/>
robotcowboy.com <http://robotcowboy.com/>

On Apr 21, 2015, at 10:56 AM, Kjetil Matheussen <k.s.matheussen@gmail.com> wrote:

But for libpd, are you sure you need to add anything? Can't just the user
call the pdinstance_new and pd_setinstance functions directly?

http://lists.puredata.info/pipermail/pd-dev/2014-05/019832.html <http://lists.puredata.info/pipermail/pd-dev/2014-05/019832.html>


(BTW. When I wrote about libpds, I hadn't forgotten about the support for pd instances,
but since I didn't have all details in my head then, I didn't mention it. I should have though.)


On Tue, Apr 21, 2015 at 4:42 PM, Dan Wilcox <danomatika@gmail.com <mailto:danomatika@gmail.com>> wrote:
This should be possible with the current version of libpd which includes Miller’s multiple instance updates, see http://lists.puredata.info/pipermail/pd-dev/2014-05/019839.html <http://lists.puredata.info/pipermail/pd-dev/2014-05/019839.html>

I just haven’t gotten around to adding libpd-specific wrapper functions for this yet, but Miller provides code in that dev list exchange.

--------
Dan Wilcox
@danomatika
danomatika.com <http://danomatika.com/>
robotcowboy.com <http://robotcowboy.com/>

On Apr 21, 2015, at 6:00 AM, pd-dev-request@lists.iem.at <mailto:pd-dev-request@lists.iem.at> wrote:

From: Oliver Greschke <info@o-g-sus.de <mailto:info@o-g-sus.de>>
Subject: [PD-dev] Can somebody help to create a desktop / VST / AU version of a PD / libPD / app ?
Date: April 21, 2015 at 3:15:44 AM EDT
To: pd-dev@lists.iem.at <mailto:pd-dev@lists.iem.at>


Hi,
I am the creator of the Elastic Drums iOS app (with great PD help from Matt Davey).
It’s made with PureData, libPD and Objective-C.
I got asked a couple of times now, if there will be ever a standalone desktop version or even better Plugin (VST, AU) version of the app.

As far as I know, there are not ready to use workarounds to do so. Which is sad, because I can imagine a lot of fantastic plugins emerging from PD

Has somebody here some experience with doing such ports?
Then please contact me.

Cheers,
Oliver

info@o-g-sus.de <mailto:info@o-g-sus.de>

_______________________________________________
Pd-dev mailing list
Pd-dev@lists.iem.at <mailto:Pd-dev@lists.iem.at>
http://lists.puredata.info/listinfo/pd-dev <http://lists.puredata.info/listinfo/pd-dev>

_______________________________________________
Pd-dev mailing list
Pd-dev@lists.iem.at
http://lists.puredata.info/listinfo/pd-dev

_______________________________________________
Pd-dev mailing list
Pd-dev@lists.iem.at
http://lists.puredata.info/listinfo/pd-dev