hosc-0.21 breaking changes

[sss-create]

hosc-0.21 apparently breaks tidal-listener since Functions and Types are now differently created in hosc. E.g. data OscProtocol. Maybe Tidal should revert and expect hosc 0.20 for now and either stay there or incorporate the hosc changes into tidal-listener and Tidal itself, which I would volunteer for to do so -- if desired.

[yaxu]

Yes agreed we should stick to hosc-0.20.

There's also a tighter version dependency in hosc-0.21 requiring a more recent ghc than the one in current debian, that I need to get to the bottom of.

It would be great to upgrade to hosc-0.21 at some point though, as we're currently preventing its release on stackage commercialhaskell/stackage#7627

Rohan actually uploaded hosc-0.21 on my request, to get rid of an obsolete dependency that caused tidal to be removed from debian:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1083252

[sss-create]

What is this tighter dependency that prevents hosc to work with the debian ghc? I tried to compile the latest hosc repo code -- which seems to be at version 0.22 already -- with ghc 9.0.2 and got an error resolved by simply specifying bytestring >= 0.12.2.0 in hosc.cabal in quite a naive way.

edit: So implementing hosc-0.21/0.22 in Tidal could free the way for hosc into stackage? Is this, plus maybe the bipolar LFO feature you mentioned in #1101, enough for a new Tidal release?

[yaxu]

Just retracing my steps on my laptop with ghc 8.8.4..

Sound/Osc/Transport/Fd/Udp.hs:44:55: error:
    Not in scope: ‘B.fromStrict’
    Module ‘Data.ByteString’ does not export ‘fromStrict’.
   |
44 | udp_recv_packet_or (Udp fd) = Binary.decodePacketOr . B.fromStrict <$> C.recv fd 8192
   |                                                       ^^^^^^^^^^^^
cabal: Failed to build hosc-0.21 (which is required by tidal-1.10.0). See the
build log above for details.

It looks like this is down to a change in bytestring-0.11.0.0:
haskell/bytestring#279

Importing from Data.ByteString.Lazy seems to fix backward compatibility:

diff --git a/Sound/Osc/Transport/Fd/Udp.hs b/Sound/Osc/Transport/Fd/Udp.hs
index 8076bba..902f456 100644
--- a/Sound/Osc/Transport/Fd/Udp.hs
+++ b/Sound/Osc/Transport/Fd/Udp.hs
@@ -6,6 +6,7 @@ import Control.Monad {- base -}
 import Data.Bifunctor {- base -}
 
 import qualified Data.ByteString as B {- bytestring -}
+import qualified Data.ByteString.Lazy as BL {- bytestring -}
 import qualified Network.Socket as N {- network -}
 import qualified Network.Socket.ByteString as C {- network -}
 
@@ -41,7 +42,7 @@ udp_recv_packet :: Udp -> IO (Packet.PacketOf Packet.Message)
 udp_recv_packet (Udp fd) = fmap Binary.decodePacket_strict (C.recv fd 8192)
 
 udp_recv_packet_or :: Udp -> IO (Either String Packet.Packet)
-udp_recv_packet_or (Udp fd) = Binary.decodePacketOr . B.fromStrict <$> C.recv fd 8192
+udp_recv_packet_or (Udp fd) = Binary.decodePacketOr . BL.fromStrict <$> C.recv fd 8192
 
 -- | Close Udp.
 udp_close :: Udp -> IO ()

I don't think the author of hosc Rohan uses github, so will drop him an email..

[yaxu]

Oops only just read your post properly.. Yep just putting a lower bounds on bytestring on hosc seems to fix it! I'm pretty sure I tried setting the lower bounds in tidal and this didn't work, so I just assumed bytestring was part of core and couldn't be upgraded.. I'll send rohan another email!

[yaxu]

edit: So implementing hosc-0.21/0.22 in Tidal could free the way for hosc into stackage? Is this, plus maybe the bipolar LFO feature you mentioned in #1101, enough for a new Tidal release?

Yes there's quite a lot of other changes waiting for release.

What's Changed

• replace Tempo module by an ableton-link synched clock that comes with tidal-link by @polymorphicengine in replace Tempo module by an ableton-link synched clock that comes with tidal-link #1059
• Add capital-M aliases for major chords by @yaxu in Add capital-M aliases for major chords #1077
• Cherry pick new Boot changes and adjust to incorporate recent changes by @matthewkaney in Cherry pick new Boot changes and adjust to incorporate recent changes #1081
• bump upper bounds of network dependency to 3.3 by @yaxu in bump upper bounds of network dependency to 3.3 #1080
• more efficient sew, plus some auto-reformatting by @yaxu in more efficient sew, plus some auto-reformatting #1078
• Step-oriented functionality by @yaxu in Step-oriented functionality #1072
• Support - for rests/gaps in mininotation by @yaxu in Support - for rests/gaps in mininotation #1087
• tactus tweaks, add s_taperBy by @yaxu in tactus tweaks, add s_taperBy #1088
• add muting with _ for p and d1-16 by @geikha in add muting with _ for p and d1-16 #1082
• fix(slice): support slice 1 _ by @grddavies in fix(slice): support slice 1 _ #1086
• Add getbpm and setbpm to boot functions by @matthewkaney in Add getbpm and setbpm to boot functions #1083
• simplify the clock logic by removing LinkOperations by @polymorphicengine in simplify the clock logic by removing LinkOperations #1090
• Bump actions/download-artifact from 2 to 4.1.7 in /.github/workflows by @dependabot in Bump actions/download-artifact from 2 to 4.1.7 in /.github/workflows #1096
• doTick: use exception-safe modifyMVar_ instead of takeMVar/putMVar by @ahihi in doTick: use exception-safe modifyMVar_ instead of takeMVar/putMVar #1100
• flake.nix: update nixpkgs and remove temporary hosc input by @ahihi in flake.nix: update nixpkgs and remove temporary hosc input #1098
• OSC Playback: Add setcps control message  #1035 added setcps and other controls to OSC Listener by @sss-create in #1035 added setcps and other controls to OSC Listener #1102
• CI pipeline repair by @sss-create in CI pipeline repair #1104

[yaxu]

(oops sorry for pinging everyone!)

[sss-create]

Oops only just read your post properly.. Yep just putting a lower bounds on bytestring on hosc seems to fix it! I'm pretty sure I tried setting the lower bounds in tidal and this didn't work, so I just assumed bytestring was part of core and couldn't be upgraded.. I'll send rohan another email!

@yaxu, your solution was the better alternative, I guess: #1105
My approach was indeed too naive.

[sss-create]

closing with PR #1105