Force SOCK_STREAM in getAddrInfo hints to fix macOS/BSD connect bug

[luntain]

Socket.defaultHints leaves addrSocketType as NoSocketType, so getaddrinfo() is free to return entries in any order for the requested host/port. On macOS this can return a SOCK_DGRAM (UDP) entry before the SOCK_STREAM (TCP) one for the same loopback host/port; internalConnectOrCancel takes the first result unconditionally, so hpgsql ends up opening a UDP socket to Postgres.

connect()/send() on that UDP socket "succeed" (UDP has no handshake), so the startup message appears to be sent successfully. The failure only surfaces on the first recv(), as ECONNREFUSED from an ICMP port-unreachable reply, since nothing is listening on that UDP port. Postgres never sees any connection attempt when this happens.

Explicitly requesting addrSocketType = Stream removes the UDP entries from getAddrInfo's results.

This fixes issue #28

[mzabani]

Thank you very much for creating the ticket and even a PR. The PR's summary is well written, and the change looks good and passes CI and in my Linux computer.

Are you happy with the PR in its current state? I can approve and merge if you are.

Also, can you use hpgsql from master, or does it help if I release a new version in hackage? I can try to make a release tomorrow or next week if one is needed.

[luntain]

Hi Marcelo, Thanks. You can go ahead and merge it. I'm building with a local copy of the repo, but I think that a release is a good idea and could potentially save some time for other users. Regards, Kamil

…

On Thu, Jul 2, 2026 at 9:57 PM Marcelo Zabani ***@***.***> wrote: *mzabani* left a comment (mzabani/hpgsql#29) <#29 (comment)> Thank you very much for creating the ticket and even a PR. The PR's summary is well written, and the change looks good and passes CI and in my Linux computer. Are you happy with the PR in its current state? I can approve and merge if you are. Also, can you use hpgsql from master, or does it help if I release a new version in hackage? I can try to make a release tomorrow or next week if one is needed. — Reply to this email directly, view it on GitHub <#29?email_source=notifications&email_token=AAAFW5U2XMOHNCBN73PFACD5C25CVA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBWHE4TKMZZGI22M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-4869953925>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAFW5S6X4IL6NBXXGL6PYL5C25CVAVCNFSNUABGKJSXA33TNF2G64TZHMYTCNZZGE2TEOJXGQ5US43TOVSTWNBXHE3TKNRTGY2DNILWAI> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/AAAFW5SOYLPIQN6VNDUCCZ35C25CVA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBWHE4TKMZZGI22M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/AAAFW5QWNCUFJRUPEOLXJLT5C25CVA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBWHE4TKMZZGI22M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[mzabani]

Good point, I will try to release a new version as soon as I can.

Thank you very much for your contribution, and feel free to keep them coming.

[mzabani]

New version with this fix just published to Hackage.