4 ms·

> pkg is an amazing system that beats dpkg or rpm What's your reasoning for this statement? I find poudriere[1] handy for compiling one's own packages from por

by voytec 4y ago

> pkg is an amazing system that beats dpkg or rpm What's your reasoning for this statement? I find poudriere[1] handy for compiling one's own packages from ports configured to fit one's goals, but I see nothing outstanding in plain pkg. It has issues when there's an IPv6 address up (even link-local) but no egress v6 routing. It would hang and wait for a timeout when an IPv6 address is selected from resolved addresses. After a timeout - pkg connects to v4 endpoint but if there's several packages to be downloaded, it can fall back to trying to connect to v6 with the next package. On the other hand, I've recently had Ubuntu 22.04 register Python-related packages which were not successfully installed, resulting in all dpkg/apt/apt-get commands failing due to py3clean script throwing tracebacks, until /var/lib/dpkg/status was edited and these packages were in fact installed and then removed. [1] https://github.com/freebsd/poudriere/wiki https://github.com/freebsd/poudriere/wiki

bluGill 4y ago
That you can mix ports and pkg is a big deal. 99% of the time the default binary pkg is good enough, but that last 1% I can get the exact options I want. I haven't used IPv6 with pkg though. (stupid ISPs).
asveikau 4y ago
Not op here, but: I think the philosophy of ports and packages (shared with other BSDs) has an upside over dpkg and rpm in that you can customize options when you build from source. The downside is that building from source takes a long time. If you follow Debian testing or sid, you will experience a lot of dependency breakage that I've never seen on a BSD, but that may not be dpkg or apt's fault so much as ongoing package maintenance causing those issues.
- bluGill 4y ago
  A lot of that is BSD has learned when to rebuild downstream ports, and they have a good ci system to rebuild them. I don't know whar Debian does for the above.
ArchOversight 4y ago
> It has issues when there's an IPv6 address up (even link-local) but no egress v6 routing. I have multiple FreeBSD machines, all of them with IPv6 up on link-local, and pkg doesn't hang at all. Even without valid globally unique IPv6 addressing, but valid ULA addressing (unique local address) and valid routing for IPv6 on an internal network, no issues. It sounds like you've got some globally unique IPv6 addresses assigned to an interface somewhere and address selection (and thus the addresses returned from getaddrinfo) will return IPv6 first and try that before falling back to IPv4. You can configure this address selection by using the ip6addrctl tool which will allow you to validate and change that address selection. It is based upon https://www.rfc-editor.org/rfc/rfc3484 https://www.rfc-editor.org/rfc/rfc3484 FreeBSD also has predefined policies you can set in /etc/rc.conf: ip6addrctl_enable="YES" # Set to YES to enable default address selection ip6addrctl_verbose="NO" # Set to YES to enable verbose configuration messages ip6addrctl_policy="AUTO" # A pre-defined address selection policy # (ipv4_prefer, ipv6_prefer, or AUTO)
- voytec 4y ago
  > It sounds like you've got some globally unique IPv6 addresses assigned to an interface somewhere Single inet6 address up is fe80::xxx. My outdated ISP does not provide any v6 connectivity. But my cellular (other) ISP provides v6 and I use tethering while traveling, so would like to avoid disabling INET6. > addresses returned from getaddrinfo) will return IPv6 first and try that before falling back to IPv4. It's random with no v6 preference. ping6 fails immediately upon attempt to reach outside world: ping6: UDP connect: No route to host ping6 2606:4700:4700::1111 0.00s user 0.00s system 50% cpu 0.005 total And here's an example[1] of how how pkg hangs for about a minute when it selects unreachable v6 address. [1] https://bsd.to/NuYu/raw https://bsd.to/NuYu/raw

by voytec 4y ago

bluGill 4y ago

That you can mix ports and pkg is a big deal. 99% of the time the default binary pkg is good enough, but that last 1% I can get the exact options I want. I haven't used IPv6 with pkg though. (stupid ISPs).

asveikau 4y ago

Not op here, but: I think the philosophy of ports and packages (shared with other BSDs) has an upside over dpkg and rpm in that you can customize options when you build from source. The downside is that building from source takes a long time. If you follow Debian testing or sid, you will experience a lot of dependency breakage that I've never seen on a BSD, but that may not be dpkg or apt's fault so much as ongoing package maintenance causing those issues.

bluGill 4y ago
A lot of that is BSD has learned when to rebuild downstream ports, and they have a good ci system to rebuild them. I don't know whar Debian does for the above.

ArchOversight 4y ago

> It has issues when there's an IPv6 address up (even link-local) but no egress v6 routing. I have multiple FreeBSD machines, all of them with IPv6 up on link-local, and pkg doesn't hang at all. Even without valid globally unique IPv6 addressing, but valid ULA addressing (unique local address) and valid routing for IPv6 on an internal network, no issues. It sounds like you've got some globally unique IPv6 addresses assigned to an interface somewhere and address selection (and thus the addresses returned from getaddrinfo) will return IPv6 first and try that before falling back to IPv4. You can configure this address selection by using the ip6addrctl tool which will allow you to validate and change that address selection. It is based upon https://www.rfc-editor.org/rfc/rfc3484 https://www.rfc-editor.org/rfc/rfc3484 FreeBSD also has predefined policies you can set in /etc/rc.conf: ip6addrctl_enable="YES" # Set to YES to enable default address selection ip6addrctl_verbose="NO" # Set to YES to enable verbose configuration messages ip6addrctl_policy="AUTO" # A pre-defined address selection policy # (ipv4_prefer, ipv6_prefer, or AUTO)

voytec 4y ago
> It sounds like you've got some globally unique IPv6 addresses assigned to an interface somewhere Single inet6 address up is fe80::xxx. My outdated ISP does not provide any v6 connectivity. But my cellular (other) ISP provides v6 and I use tethering while traveling, so would like to avoid disabling INET6. > addresses returned from getaddrinfo) will return IPv6 first and try that before falling back to IPv4. It's random with no v6 preference. ping6 fails immediately upon attempt to reach outside world: ping6: UDP connect: No route to host ping6 2606:4700:4700::1111 0.00s user 0.00s system 50% cpu 0.005 total And here's an example[1] of how how pkg hangs for about a minute when it selects unreachable v6 address. [1] https://bsd.to/NuYu/raw https://bsd.to/NuYu/raw