|
From: Bernard B. <ber...@gm...> - 2020-07-23 21:25:59
|
Dear Anthony, I also found out that we get an error message with an empty Trace file when there is a mistake in the command, which is not the case here. I got a mistake trying to run a grammar whose name contains a space… This is now fixed by putting spaced names or paths between quotes. I also completed the reformat_grammar() procedure after checking all reserved words on BP2.9.8, many of which I had forgotten as they date back to modelling drum improvisation! The files I modified are attached. They are also in https://leti.lt/bolprocessor/ Indeed the crash log tells something but I am unable to make sense of it! It might be difficult to trace the process on such a complicated grammar. I use it only because it contains a great diversity of instructions. So the error may just be stored on the to-do list. Scripts will be useful once we have real-time MIDI in/out, but we'll need to ignore the ones having effects on the interface, such as loading a file, changing work folder, sounding a beep, waiting for the space key to be pressed etc. I will prepare a list of the ones that should be ignored. Many of them were reproducing menu commands, as a script could be constructed in the "learning mode": every action on the interface was recorded as a script command. But this all has to do with real-time operations. Probably a few script commands will be used by the PHP interface, e.g. load a file, produce x items etc. and executed via command lines. At some point we may need to find a method for waiting for a mouseclick or any human action triggering the performance of an item that has been prepared. Once interface-connected script commands have been neutralized, the relevant script operations should work flawlessly. There is nothing complicated in the process. "BP2 script" is a header on top of a script or glossary file. It does nothing in particular. When I was developing the core program in Delhi, BP2 was connected to my D50 synth. Every evening (once my colleagues had left the office) I used to run "+sc.checkall" designed for trying all features. This took about 30 minutes. I was listening while doing other things, then checking errors on the Trace. By ear I would immediately identify odd behaviour. ;-) I am now tempted to work on the import of MIDI files into sound object prototypes, and converting MIDI sequences to Csound. Looking at the C code should make it easy. Bernard Anthony Kozar wrote on 23/07/2020 22:19: > I have reproduced the problem here. When I attempt to create a Csound > score on the command line, BP console gets as far as time setting the > last sequence (?) and then crashes: > > 3566 objects. Compression rate = 51 Sequence 11/11 > Sizing objects > Setting limits > Placing objects > Solving constraints > Calculating global drift > Calculating legato/staccato > Segmentation fault: 11 > > When you don't get any output running a command from the PHP interface > it may help to run it directly on the command line. > > The crash log shows that this occurred in MakeSound() and the > exception that caused the crash was > > Exception Type: EXC_BAD_ACCESS (SIGSEGV) > Exception Codes: KERN_INVALID_ADDRESS at 0x0000000000000000 > > I'm guessing that that is an attempt to access a null pointer. I'm > guessing that it is related to the use of script commands but I'll > have to trace the problem in the debugger to be sure. > > Interestingly, the segmentation fault does not occur if you run this > grammar with --rtmidi instead of --csoundout. Then the output > continues after time setting: > > 3566 objects. Compression rate = 51 Sequence 11/11 > Sizing objects > Setting limits > Placing objects > Solving constraints > Calculating global drift > Calculating legato/staccato > Ignoring script command: MIDI controller #98 = 1 channel 1 > (Scripts do not work yet in the console version). > Ignoring script command: MIDI program 63 > (Scripts do not work yet in the console version). > Ignoring script command: Beep > (Scripts do not work yet in the console version). > Ignoring script command: Wait for space > (Scripts do not work yet in the console version). > Ignoring script command: Tick cycle OFF > (Scripts do not work yet in the console version). > > The fact that the very next message after where it seg faults for > --csoundout is "Ignoring script command: MIDI controller #98" is > evidence that my hunch is correct. > > Nothing related to script commands currently works in BP console. And > the only other crash I have experienced with it so far had to do with > loading a glossary file that I think began with "BP2 SCRIPT". IIRC, I > took some shortcuts to get the console version to compile in 2013 and > I did not really finish substituting appropriate code everywhere for > the non-functional script commands. > > I will add fixing these crashes to my to do list. Making script > commands work in the console version is a large task though and I was > not sure how important it is to do. Certainly some of the script > commands are no longer relevant (e.g. Freeze windows). And I have the > impression (which may not be correct) that many of the others > duplicate functionality that can be achieved in other ways(??). > > I'd like to hear what you think. > > Anthony > > > On 7/22/20, 11:14 AM, Bernard Bel wrote: >> Dear Anthony, >> >> When I compile grammar “-gr.ShapesInRhythm” in the PHP interface, it >> says "no error". >> >> When producing a "BP data file" I get the (presumably correct) data on >> the Data window. >> >> However, attempting to produce a Csoud file ends up in an error. >> However, no messages have been produced and the Trace window remained >> empty. >> >> Note that this grammar does produce Csound (and MIDI) files on BP2.9.8 >> although it asks the user to press the space key before doing it. This >> is due to instruction "_script(Wait for space)" which should be ignored >> in the context of creating a file. >> >> I guess that BP is not able to create Csound events reproducing >> performance parameters such as "_script(MIDI controller #98 = 1 channel >> 1)", but at least when it fails it should send a message to Trace. >> >> Maybe this will be solved by loading “-se.ShapesInRhythm.QTM” ? >> >> Bernard > > > _______________________________________________ > bolprocessor-devel mailing list > bol...@li... > https://lists.sourceforge.net/lists/listinfo/bolprocessor-devel |
