Arch Linux User Repository

@frakspace, thanks for hunting down the issue! :D

@imrehg Ah, I figured out the problem. Your version-3 pkgrel works for me, but ONLY if I replace "gcc-go" (which is what I originally had installed) with "go". If I tamper with the specified version number so that the PKGBUILD accepts gcc-go, the same failure occurs. So, that was what was wrong, gcc-go is simply broken and not an acceptable substitute. Thanks!

@imrehg I'm not really sure. It spends a considerable amount of time downloading something. Whether it's succeeding, I don't really know, but if the following diagnostics are useful to you, here are some find commands I tried while in the directory where I tried to build it:

[ipget]$ find . -iname *flood*
./src/src/github.com/libp2p/go-floodsub
./src/src/github.com/libp2p/go-floodsub/floodsub_test.go
./src/src/github.com/libp2p/go-floodsub/floodsub.go
[ipget]$ find . -iname *libp2p*
./src/src/github.com/libp2p
./src/src/github.com/libp2p/go-libp2p-protocol
./src/src/github.com/libp2p/go-libp2p-peer
./src/src/github.com/libp2p/go-libp2p-net
./src/src/github.com/libp2p/go-libp2p-host
./src/src/github.com/libp2p/go-libp2p-crypto
./src/src/github.com/libp2p/go-libp2p-transport
./src/src/github.com/libp2p/go-libp2p-interface-connmgr
./src/src/github.com/libp2p/go-libp2p-peerstore
./src/src/github.com/libp2p/go-libp2p-interface-conn
[ipget]$ find . -iname *ultiadd*
./src/src/github.com/multiformats/go-multiaddr-net
./src/src/github.com/multiformats/go-multiaddr-net/multiaddr
./src/src/github.com/multiformats/go-multiaddr-net/multiaddr/multiaddr.go
./src/src/github.com/multiformats/go-multiaddr
./src/src/github.com/multiformats/go-multiaddr/multiaddr_test.go
./src/src/github.com/multiformats/go-multiaddr/multiaddr.go
[ipget]$ find . -iname *ultipar*
[ipget]$ find . -iname tar-utils
[ipget]$ find . -iname tar*
[ipget]$ find . -iname *utils

So I'm not positive what that means, but assuming I'm not a whole lot less competent at using find than I think I am, it looks like the deps from "whyrusleeping" don't exist anywhere. Or they're named something funny. Or I don't understand what's going on even more than I think. I hope it's of some use, anyway.

@frankspace that's weird, seems like go get fails for you? Looks like textflag.h should be used properly, since it's supplied by the go package.... Not sure I have a guess at this point...

@Eschwartz well, how deep people go into a package and change how it is built? Here for example that means patching out part of the Makefile to work differently... Anyway it's done in pkgver 3, let me know if there's any outstanding "damn don't do it this way" remains.

This isn't working for me. Here's what happens:

==> Extracting sources...
  -> Extracting v0.3.0.tar.gz with bsdtar
==> Starting build()...
==> Installing extra dependencies
> github.com/libp2p/go-floodsub
# github.com/minio/blake2b-simd
src/github.com/minio/blake2b-simd/compress_amd64.go:21:5: error: reference to undefined name ‘avx2’
  if avx2 {
     ^
src/github.com/minio/blake2b-simd/compress_amd64.go:23:12: error: reference to undefined name ‘avx’
  } else if avx {
            ^
src/github.com/minio/blake2b-simd/compress_amd64.go:25:12: error: reference to undefined name ‘ssse3’
  } else if ssse3 {
            ^
# github.com/minio/sha256-simd
src/github.com/minio/sha256-simd/sha256blockAvx2_amd64.s:35:10: fatal error: textflag.h: No such file or directory
 #include "textflag.h"
          ^~~~~~~~~~~~
compilation terminated.

Thanks, but we've got perfectly serviceable packages for gx, and gx-go, which you can use instead of downloading prebuilt versions in make build that don't even do checksum verification.

Also the previous maintainer was an upstream dev who uploaded three non -bin packages (gx and gx-go as well) which all distributed his prebuilt binaries. -bin packages are not automatically bad though, don' worry. :)

Done, built from source.

Revising my last statement: now I see that it can be built without host IPFS client, as it will bootstrap itself. On the other hand, currently the version seems to be unbuildable (because of dependency errors). Thus I think there's a good path to make this package source-buildable - but for that likely upstream changes are needed, and in the meantime the binary distribution has to do. Will be looking into workarounds.