Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Nov 29 23:37

    jcelerier on master

    [audio] Allow to disable portau… (compare)

  • Nov 29 23:35

    jcelerier on master

    [musical] Fix off by one error … (compare)

  • Nov 29 09:46
    thibaudk synchronize #746
  • Nov 26 12:46
    vincentgoudard commented #755
  • Nov 26 12:46
    vincentgoudard commented #755
  • Nov 26 12:45
    vincentgoudard commented #755
  • Nov 26 12:40
    vincentgoudard commented #755
  • Nov 26 12:40
    vincentgoudard commented #755
  • Nov 26 12:10
    avilleret commented #755
  • Nov 26 11:50
    vincentgoudard labeled #755
  • Nov 26 11:50
    vincentgoudard labeled #755
  • Nov 26 11:50
    vincentgoudard opened #755
  • Nov 26 11:50
    vincentgoudard labeled #755
  • Nov 25 09:56

    jcelerier on master

    [audio] ALSA: fix period count … (compare)

  • Nov 24 17:52

    jcelerier on master

    [mapper] Use floats even in the… (compare)

  • Nov 24 17:36

    jcelerier on master

    [websocket] Make sure that we s… [asan] Workaround for an ASAN f… [units] Initialize make_unit_he… (compare)

  • Nov 22 16:23

    jcelerier on master

    [portaudio] Use the new target … (compare)

  • Nov 22 14:48

    jcelerier on master

    [win32] Update wiiuse for win7 [protocols] Fix a weird codegen… (compare)

  • Nov 21 08:59

    avilleret on github-action

    try to fix deken upload trying to understand why QML ma… trying to understand what is go… (compare)

  • Nov 20 21:14

    avilleret on github-action

    Remove RtMidi17 thing which had… [git] Remove asio submodule whi… [execution] Fix a graph executi… and 22 more (compare)

Jean-Michaël Celerier
@jcelerier
@avilleret do we still care about 32-bit ?
it seems that MSVC is having an internal compiler error but only on 32-bit :|
max 8 is 64-bit only no ?
and we don't support max 7 anymore, or do we ?
Antoine Villeret
@avilleret
hi @jcelerier, imho we can ditch 32bit builds. Puredata is still available as 32bit on Windows, but who need it ?
Jean-Michaël Celerier
@jcelerier
it's just the max build which is failing
but if you think that 32 bit windows pd is unused then yeah
Antoine Villeret
@avilleret
if it builds fine, then keep it until it fail, concerning Max, I don't care supporting Max7, so we can disable Max 32bit builds
I'll try to port more build to github-action today, without any guarantee though
Alex Norman
@x37v
@jcelerier okay cool, i'll try out this osc factory. Thanks!
Alex Norman
@x37v
@jcelerier not sure if all of these protocols should support multiplex, but i updated my test to use the factory setup: ossia/libossia#735
Alex Norman
@x37v
does the new protocol setup for OSC have an output only UDP configuration?
Alex Norman
@x37v
looks like no, and i'm not entirely sure how i'd add it with the new setup
Jean-Michaël Celerier
@jcelerier
fixing that quickly @x37v
in terms of API i'm just changing the socket_configuration to optionals
so that it's easy to choose between only sender, only receiver, or both
will push as soon as the tests pass
:)
Jean-Michaël Celerier
@jcelerier
thanks for the extensive test, reviewing it
Alex Norman
@x37v
fantastic @jcelerier ! thanks, I was thinking optionals would be nice, though, the container is called bidir or something so i wasn't quite sure :)
will test out this approach in the example in the next few hours
Alex Norman
@x37v
I updated my PR, rebased against the async_protocols branch and fixed up the test, still failing on that SET getting sent out though
but as is, this should be usable for my purposes, it'll just have extra messages out that i consider noise.. but not that big of a deal
Alex Norman
@x37v
okay, another thing. considering the new complexity of osc protocol configurations, seems like it would be nice to have a visitor based to get ip/port so, for instance, i can find and remove OSC senders from the muliplex protocols.. maybe something already exists?
i think I can figure out the exact type i'm using and do a dynamic cast like i used to do, but i imagine its something that would be generally useful in the lib
Alex Norman
@x37v
okay, the SET output isn't just noise now, it is causing crashes because websockets can handle a lot bigger OSC encoded messages on the input than UDP osc can send out (I believe)
when i add a UDP OSC output to my multiplex protocol that also has a OSCQuery protocol, i get crashes when I send these large WS messages.
it might make sense for me to chunk these smaller either way
Jean-Michaël Celerier
@jcelerier

okay, the SET output isn't just noise now, it is causing crashes because websockets can handle a lot bigger OSC encoded messages on the input than UDP osc can send out (I believe)

agh indeed, I'll add a filter + log

hmm
although a crash shouldn't happen
do you have a repro ? I have some tests that send large messages, and if it's > the max UDP size, it just gives an error during send
well I can easily do a check on my end given what you said, trying that

okay, another thing. considering the new complexity of osc protocol configurations, seems like it would be nice to have a visitor based to get ip/port so, for instance, i can find and remove OSC senders from the muliplex protocols.. maybe something already exists?

It does not exist, what would your dream API for that look like ?

Alex Norman
@x37v

do you have a repro ? I have some tests that send large messages, and if it's > the max UDP size, it just gives an error during send

I'll try to create a test case for it.

Alex Norman
@x37v

okay, another thing. considering the new complexity of osc protocol configurations, seems like it would be nice to have a visitor based to get ip/port so, for instance, i can find and remove OSC senders from the muliplex protocols.. maybe something already exists?

It does not exist, what would your dream API for that look like ?

Hmm. I'm not sure, I kinda liked just being able to dynamic cast to the osc protocol before and get the port and ip.. so something simple ideally without having to make my own visitor.. in the mean time I have a std::set of pairs of ip/port that i keep and to remove a specific OSC outbound I clear everything from the multiplex protocols except an oscquery protocol, then i remove the ip/port from the std::set and add the OSC outbounds back.
not ideal, but works.. but really only because the oscquery protocol is pretty simple to get the type of

Jean-Michaël Celerier
@jcelerier
thanks @x37v, checked, fixed that large size thing and commented on the PR
btw @/all I'm not sure this is entirely 100% strictly necessary but maybe we should have at some point a small chat session on improving the contribution process ? so far it seems to work well but I'm sure we can find some small quality-of-life improvements, also maybe do a pass on whatever automated tool we can use, e.g. was thinking of commit hooks for various sanitizations
Alex Norman
@x37v
@jcelerier updates (re OSC out crash issue) are working well, thanks!
Alex Norman
@x37v
ahh cool, the async-protocols stuff is in master now! great
Thibaud Keller
@thibaudk
Hi @jcelerier , sorry, I am looking for the script to regenerate dataspace enums
is it a .sh file ?
Jean-Michaël Celerier
@jcelerier
It's genvariant
But it needs fixing
Alex Norman
@x37v
been a while..
I'm using multiplex_protocol to enable the addition/deletion of UDP OSC listeners.. generally it works but I'm occasionally getting an exception Bad Connection when I'm updating a value from another thread.. so, it seems like adding/removing protocols isn't thread safe
though, I should upgrade my libossia build as it has been a while
Alex Norman
@x37v
I'm using the context approach, with one thread polling the context.. there are 2 or 3 other threads that might push values, but only one polls the context.. the exception happens outside the poll, so maybe the context isn't actually fully encapsulating all network i/o?
Alex Norman
@x37v
interestingly, the exception is coming from the websocket library via the oscquery protocol, which is the one protocol that i don't remove
Jean-Michaël Celerier
@jcelerier
hi @x37v :)