[PD] Inlet - Unexpected Behaviour

[Jonathan Wilkes]

    On Saturday, August 29, 2020, 03:04:25 PM EDT, Alexandre Torres Porres <porres at gmail.com> wrote:  
 
 Em qui., 6 de ago. de 2020 às 16:08, Jonathan Wilkes <jancsika at yahoo.com> escreveu:

 > On Thursday, August 6, 2020, 2:07:09 PM EDT, matthew brandi <mfbrandi at outlook.com> wrote: 
 > Dear people

> In my role as village idiot, I am asking whether the string "fwd" in a message has a
special meaning to inlet.

> Naively, I was expecting inlet to pass the string to the subpatch, but it seems not
to. See example patch attached.

AFAICT that's a regression due to the way Pd Vanilla implemented message forwarding for 
 [inlet~ fwd]. That's a feature that allows a signal inlet of a subpatch/abstraction to forward 
non-signal messages to the right outlet of [inlet~ fwd]. (The right outlet sprouts when the 
"fwd" argument is present.)
Another regression-- there is no longer an error if you try to send a non-signal message to 
[inlet~].
Another regression-- [inlet~ fwd] unconditionally allocates space on the stack to copy the 
entire incoming message. If you generate a long enough message this will blow the stack 
and cause Pd to crash. Esp. important given that Windows stack is much smaller than the RAM 
available for heap allocation on most machines.
Also-- I *think* Pd Vanilla doesn't forward pointer messages through [inlet~ fwd]. It appeared to be an oversight-- at least I didn't see any comment about it.

A GSoC student spent some time reimplementing this in Purr Data, so none of thiese should be 
issues there.


> I think it's a good idea if you're changing and fixing stuff to also send a PR to vanilla as a proposal. Would you consider doing that as well?
Before you ask that, have a look at the PRs. You can view the list of open PRs for Pure Data Vanilla here:
https://github.com/pure-data/pure-data/pulls

It appears Christof (Spacechild1) submitted a patch for this almost a month ago. So it wouldn't make sense to send another 
PR for this same fix. Christof's patch should work just fine to solve these issues.
If you're asking in general, that's unfortunately just not practical. We'd have to manually create a new patch set and test it for Vanilla, for (nearly) every single change we make. It's much easier to just report here when I find crashers-- Miller and Christof are quite quick to fix them. They know the Vanilla build/test/development process much better than I do so that seems the much preferable route for everyone involved.
Best,Jonathan  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20200829/8c18e174/attachment-0001.html>