Composers Desktop Project Forum

Access Violation has been fixed in the latest version 5.07, now available for download.

I'll look into the aspects of Create Envelope File. (Pity that was not a separate thread and I would have caught it before releasing 5.07. Too bad.)

A new version of Soundshaper (5.07, 9 May 2020) is now available for download. This version fixes an Access Violation affecting TEXTURE processes and revises the Initial Setup and reading of the Settings file, which will be renamed automatically.

Version 5.06 (3 May) fixed a significant bug affecting processes with more than 10 parameters, including TEXTURE and GRAIN.

The Access Violation affecting Texture Grouped (and Decorate) has been pinned down and fixed. I'll need to do many more tests to confirm this and will issue a replacement .exe as soon a possible.   

In the end, it wasn't anything to do with Texture Grouped. I assumed the data was being read in wrongly or parameter-related objects (like slider ranges) were being re-defined at the wrong time. No, it was re-writing the page title. So I put this later and all was well.  Just the sort of geeky Windoze-type of problem I do programming for. Or not.   

And why did it affect only Texture?
-- No idea. Computers are magic. Sometimes they behave like magic too.

I also found the RESET possibility, but I found that it tended to give an A.V. every time (instead of just every other time). And I experimented a lot with snapshots; it is useful with this problem to save parameter sets via the snapshot buttons, but again I found that if recalled one when there was an A.V. I got a worse "crash".

However, I can confirm that your Reset + recall snapshot method works. (4th May) Thanks for the suggestion, rotagrad.

For now, I'm continuing to work with the code and explore other ways of doing things, which might hopefully avoid the problem altogether. 4th May update: I've now isolated where the problem occurs and roughly why -- also why it would not give an A.V. every other time.

I've tracked down where the A.V. occurs, but not why. It's something to do with the initial display of the parameter page when it's re-visited to edit the parameters. The weird thing is it doesn't do it every time -- but every OTHER time, consistently.

However, this gives a workaround: click CANCEL to leave the parameter page and try again (double-click the cell again). This seems to work. A third edit will give another A.V. but the 4th should be OK and so on. You will also get one when you exit the program.

In my searches through the code, I found a different serious bug affecting all processes with >10 parameters, mainly TEXTURE and GRAIN. I've issued an interim .exe correcting this (5.06).

Meanwhile I'll continue to look for the cause of the A.V. You're correct, rotagrad, that the bug also affects TEXTURE DECORATE.  But the same workaround seems to work. Maybe whatever these two have in common will give a clue...

Thanks for posting this bug. I've verified the problem you mention. I ran Grouped with a single input and the default values. Like you I got the access violation when I went to edit the parameters. I then cancelled and tried again; this time I got the edit boxes but the values were all 0. I also tried recalling a preset and again I got the violation.

Access violations can be quite tricky to track down, but usually occur from impossible values, like accessing a slider value beyond its declared range, for instance. I'll take a close look at the code here and see if I can track down anything sloppy in that regard. The re-setting of values to zero may give a clue. Grouped used to work OK, so I can also compare old code with the current version.

Robert - Author, Soundshaper.

Sonora, Thanks for getting in touch by email and thanks for your patience.

Unfortunately, the problems are still unresolved (at 8 Apr.). The program was not reading the CDP programs folder correctly from the configuration file, despite the fact that the file was correct and the folder was there. It was also not shutting down correctly. Both of these issues need further investigation, though neither has arisen in other installations, as far as I know. While we were exchanging emails, you reported that you had installed Soundshaper successfully on another Win10 computer. I have also tested it recently on 3 Win10 64-bit machines, one Win10 32-bit, one Win7 64-bit and one Win7 32-bit.

The problem may have something to do with the fact that Soundshaper is a 32-bit program running, in this case, on 64-bit Windows. Having said that, Windows automatically runs such programs using its built-in emulator WOW64, so in theory issues should not arise. You also tried running it in Win7 mode, without success. Re-building Soundshaper in native 64-bit would be possible, but could require a lot of work -- I'll look into this.

We will continue to communicate and experiment with other solutions and report back to the forum if we solve the problems.

Mario, can you email me with this problem to:  mail [at] ensemble-software.net ? The problem is not likely to be Windows 64-bit or robocopy, so I'll need to get more detail from you and look into this further. I'll be happy to post a reply when we get it fixed.  Robert

The CDP documentation has been updated today (3rd Apr 2020). The online html docs were already the latest, as were the individual printable pdf versions, but the zip files of both were out-of-date (Aug 2018). Now corrected, with apologies for any inconvenience. The docs are housed at www.ensemble-software.net/CDPDocs

Two changes I would appreciate some help with: SFEDIT ISOLATE in cgroedit.htm has apparently been updated but to my eye it seems the same as in the unchanged sfedit.pdf. Similarly PAPLAY in cmcrefmn.htm / mctoolkit.pdf. Let me know if you can spot any difference(s) .

When it works, the MIDI-Sequence facility in Soundshaper is a two-pass process: it uses the public-domain program mf2t (as updated by Richard Dobson) to create a text file version of the MIDI file and then attempts to convert that into a .seq file for use with the CDP sequencer (extend sequence 2).

There are some problems with the present version which will need detailed investigation. It appears to work OK with a single track. In tests I have got it to convert a 4-track MIDI file, but I have also had access violations which have required a re-boot. I've found no problems with the first-pass mf2t, which can convert either type 0 or type 1 MIDI files.

If the .seq file is created successfully, you have to look at its first-line header before using it. It might look something like this - 60 60 60 60  - which indirectly tells you that 4 input files are needed. The first will be that in the current Soundshaper cell, the others must be selected via the OPTIONAL INPUTS panel, which will pop up when you select Sequencer from the process menus. These can be the same or different sounds. So in the above example, you might click the current cell three times to select the same source for inputs 2-4. An odd quirk of Sequencer is that two inputs are the minimum, so if there is only one source, you must click the source cell once as a dummy 2nd input. (I've no idea why this is.)

In conclusion, I think MIDI-Sequence can be made to work in a limited way, but there are clearly problems with it at present. I'll look into these further, but don't expect a quick fix. If the problems persist, the facility wil be removed from Soundshaper and perhaps offered as an external program to be used "as-is".

Thanks for your post, sonora. I'll look into it.

I don't know if it's been tried with the latest version.

Soundshaper already works on Parallels and similar emulations (or so I've been reliably informed). This is surely an easier and cheaper solution than a cheap PC?

I didn't mention financial constraints that are also hindering an OSX port at the moment. I do however have an old PC running Linux and as time permits, I plan to get CDP running on that and then start gradually re-building Soundshaper for Linux. The Free Pascal/Lazarus route is multi-platform, so code adapted for Linux might well cover most of what's needed for OSX. Crucially, the unit which assembles the command-line batch files that run CDP will need to be converted from the Windows CMD language to BASH, which can be tested on Linux and would also be used for OSX. The new player component, using the 3rd-party BASS library, has already been written in Free Pascal/Lazarus. Much of the user interface stuff  - menus, patchgrid and parameter setting - is pretty standard and should port without any trouble. The graph-edit page I'm less sure about. Other supporting pages - CDP Mix page and Data editor - could probably wait until after a first release.

I'm afraid not at the moment. I don't have a Mac to develop it on. In principle, the plan would be to adapt the present code (Delphi) to run in the Free Pascal equivalent (Lazarus), then adapt that to run in Linux and OSX. In practice, there are bound to be Mac and Linux-specific things that will take a lot of sorting out. Secondly, is there much demand for it? I'm not keen to spend months of work to create something that might be used by a handful of people.

You're quite correct: the version of specnu released in 2016 does not seem to have the options rand or squeeze, and specnu slice mode 5 has also disappeared.  These were all present in Release 7.0 in Feb. 2014, but have somehow got lost in the 7.1 upgrade. Rand randomises the order of windows, giving a flutter effect, squeeze reduces the frequency range, altering the timbre, while slice 5 inverts the spectrum, creating dramatically different (usually inharmonic) frequency content. All useful functions which would enhance the CDP suite if they could be restored -- please?