kea-dhcp FTBFS with boost1.87

[thomas]

When compiling kea-dhcp on a system with boost-1.87, it breaks with

  ...
  CXX      libkea_asiolink_la-common_tls.lo
  CXX      libkea_asiolink_la-interval_timer.lo
  CXX      libkea_asiolink_la-io_address.lo
io_address.cc: In constructor 'isc::asiolink::IOAddress::IOAddress(const std::string&)':
io_address.cc:40:34: error: 'from_string' is not a member of 'boost::asio::ip::address'
   40 |     asio_address_ = ip::address::from_string(address_str, err);
      |                                  ^~~~~~~~~~~
io_address.cc: In member function 'uint32_t isc::asiolink::IOAddress::toUint32() const':
io_address.cc:119:39: error: 'class boost::asio::ip::address_v4' has no member named 'to_ulong'
  119 |         return (asio_address_.to_v4().to_ulong());
      |                                       ^~~~~~~~
  CXX      libkea_asiolink_la-io_endpoint.lo
make[5]: *** [Makefile:766: libkea_asiolink_la-io_address.lo] Fehler 1
make[5]: *** Es wird auf noch nicht beendete Prozesse gewartet …
In file included from io_endpoint.cc:12:
../../../src/lib/asiolink/tcp_endpoint.h: In constructor 'isc::asiolink::TCPEndpoint::TCPEndpoint(const isc::asiolink::IOAddress&, short unsigned int)':
../../../src/lib/asiolink/tcp_endpoint.h:45:74: error: 'from_string' is not a member of 'boost::asio::ip::address'
   45 |             new boost::asio::ip::tcp::endpoint(boost::asio::ip::address::from_string(address.toText()),
      |                                                                          ^~~~~~~~~~~
In file included from io_endpoint.cc:13:
../../../src/lib/asiolink/udp_endpoint.h: In constructor 'isc::asiolink::UDPEndpoint::UDPEndpoint(const isc::asiolink::IOAddress&, short unsigned int)':
../../../src/lib/asiolink/udp_endpoint.h:45:74: error: 'from_string' is not a member of 'boost::asio::ip::address'
   45 |             new boost::asio::ip::udp::endpoint(boost::asio::ip::address::from_string(address.toText()),
      |                                                                          ^~~~~~~~~~~
make[5]: *** [Makefile:773: libkea_asiolink_la-io_endpoint.lo] Fehler 1
make[5]: Verzeichnis „/mnt/tmp/kea/source/kea-2.6.1/src/lib/asiolink“ wird verlassen
make[4]: *** [Makefile:883: all-recursive] Fehler 1
make[4]: Verzeichnis „/mnt/tmp/kea/source/kea-2.6.1/src/lib/asiolink“ wird verlassen
make[3]: *** [Makefile:475: all-recursive] Fehler 1
make[3]: Verzeichnis „/mnt/tmp/kea/source/kea-2.6.1/src/lib“ wird verlassen
make[2]: *** [Makefile:468: all-recursive] Fehler 1
make[2]: Verzeichnis „/mnt/tmp/kea/source/kea-2.6.1/src“ wird verlassen
make[1]: *** [Makefile:655: all-recursive] Fehler 1
make[1]: Verzeichnis „/mnt/tmp/kea/source/kea-2.6.1“ wird verlassen
make: *** [Makefile:543: all] Fehler 2

Running same instructions on a system with boost-1.85 has shown no issues. (not tested with boost-1.86)

There is a patch at Gentoo (​https://bugs.gentoo.org/946488) available as ​https://946488.bugs.gentoo.org/attachment.cgi?id=914160. This patch applies against kea-2.4.1 as Gentoo has not upgraded to kea-2.6.1 for whatever reason. Unfortunately, the patch does not fully apply to 2.6.1.

[thomas]

I've adopted the Gentoo patch to apply on kea-2.6.1. Hopefully it fixes stuff and does not damage something else...

[Xi Ruoyao]

Is the patch reviewed by the upstream? We generally don't take a patch not reviewed by the upstream at all (unless the upstream is dead, or the patch just alters some install paths for LFS configuration, or some other strange circumstances).

[thomas]

Which guideline is this? I'm pretty sure we have a lot patches which are not reviewed by upstream or not even submitted to. They all would be invalid.

Anyway - strange thing is, that Arch does have boost-1.87 now and kea-2.6.1 but either no issues or they have not yet recognized them and produced a patch.

There is(was) a ticket at ISC (​https://gitlab.isc.org/isc-projects/kea/-/issues/3696) which is closed, but it leads to ​https://github.com/isc-projects/kea/pull/143/commits/6300b3c937f91042594cf0944a8ec358d5bed809 which is somehow comparable to the Gentoo stuff.

I'll follow that path for now, maybe there is a solution...

[Xi Ruoyao]

Replying to thomas:

Which guideline is this? I'm pretty sure we have a lot patches which are not reviewed by upstream or not even submitted to. They all would be invalid.

But we should make our practice better. I remember I fixed a bug in an unreviewed BLFS patch several years ago and it'd be considered a security vulnerability if we'd classified security issues that year.

Anyway - strange thing is, that Arch does have boost-1.87 now and kea-2.6.1 but either no issues or they have not yet recognized them and produced a patch.

There is(was) a ticket at ISC (​https://gitlab.isc.org/isc-projects/kea/-/issues/3696) which is closed, but it leads to ​https://github.com/isc-projects/kea/pull/143/commits/6300b3c937f91042594cf0944a8ec358d5bed809 which is somehow comparable to the Gentoo stuff.

Then it's better to directly take the upstream patch. I.e. ​https://github.com/isc-projects/kea/pull/143.patch.

It's already merged so it's just the upstream content.

For Arch the reason is simply they have not rebuilt kea after upgrading boost. Kea was last rebuilt on Dec 22 2024 but boost was updated on Feb 07 2025.

[Xi Ruoyao]

Replying to Xi Ruoyao:

For Arch the reason is simply they have not rebuilt kea after upgrading boost. Kea was last rebuilt on Dec 22 2024 but boost was updated on Feb 07 2025.

(It's similar to how we miss a package broken by a dependency update when we develop BLFS... And AFAIK they even don't have a "tagging" cycle like us.)

[Xi Ruoyao]

And I just mean "we should at least make the upstream take a look." If the upstream insists to reject a good (reviewed & tested by us, and also used by other distros) patch for some stupid reason or it's inactive at all we can just apply the patch.

[thomas]

At the end of the day, i agree we should be quite careful with doing patches on our own. I also didn't felt very good to modify something i do not understand.

I found the 143 patch, too. That patch seems to do the thing, rebuilt Kea with this patch successfully.

(Edit: I've attached that patch. I'll add that patch, which is named kea-2.6.1-fix_boost_1_87-2.patch here in the attachments, as kea-2.6.1-fix_boost_1_87-1.patch to the book)

[thomas]

​https://patch-diff.githubusercontent.com/raw/isc-projects/kea/pull/143.patch

[Douglas R. Reno]

I can confirm that the build is broken and that this patch fixes the issue!

[thomas]

Fixed at BLFS [85d0632167]