Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 25 Apr 2021 00:39:57 +0200
From:      Willem Jan Withagen <wjw@digiware.nl>
To:        Adriaan de Groot <adridg@freebsd.org>, freebsd-ports@freebsd.org
Subject:   Re: freebsd-ports Digest, Vol 933, Issue 7
Message-ID:  <0a6c7864-e9e7-dc47-793c-914be52ca582@digiware.nl>
In-Reply-To: <2544300.TYJnH3iKXO@beastie.bionicmutton.org>
References:  <mailman.88.1618747201.1748.freebsd-ports@freebsd.org> <2544300.TYJnH3iKXO@beastie.bionicmutton.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On 18-4-2021 16:51, Adriaan de Groot wrote:
> On Sunday, 18 April 2021 14:00:01 CEST freebsd-ports-request@freebsd.org
> wrote:
>>>> So perhaps that is the best way to avoid having to deal with ABI/API
>>>> breakage...
>>>> After that it is up to the maintainers of the dependant packages to
>>>> update their package and start using boost-1.75.
>>> There is the implicit assumption that a patch that updates
>>> boost for all the dependent ports should also provide fixes
>>> if those ports fail to build after the update. That is
>>> the major task.
>> There are "only" 490 ports that have boost in their Makefile.
> Compare this to CMake, which has about 1900 direct consumers. While there's
> less ABI-breakage and more configure- and build-time breakage, a CMake update
> looks like this:
> - build all the consumers with the current version of CMake
> - fix the ones that don't build, or mark BROKEN
> - bump CMake locally
> - build everything again
> - fix the ones that don't build
>
> The actual CMake parts are the easy bits. Dealing with 15-year old C++ code
> that happens to use CMake and falls over for totally unrelated reasons is the
> real time sink. Ceph, too, is one of those things that eventually gets added
> to my ignore list as "doesn't build in any recent ports checkout":
>
> /usr/local/bin/ld: ../../../lib/libkv.a(LevelDBStore.cc.o):
> (.data.rel.ro._ZTI17CephLevelDBLogger[_ZTI17CephLevelDBLogger]+0x10):
> undefined reference to `typeinfo for leveldb::Logger'
>
> I fix what I can, send mail to maintainers when I can't and so we hope that
> progress happens
Hi Adriaan,

I understand your pain.

I'm in the squeeze that Ceph has moved to new version, which I do get to 
build/test
in my jenkins builders. But I'm/was sort of waiting for the move to git 
with ports.
And then figure out how to get things to work again with the latest 
release.
Unfortunately I got little time at the moment, so progress there is slow.

And you are right that it hard to find the cause for this, since so many 
things
have changed: Ceph source, Cmake config, llvm compiler and linker and 
leveldb.
So the permutations are plenty.

If you want to fix the above (which I also got in my builders, and have 
not found the
root cause for) you could add `-D WITH_LEVELDB=OFF` to the CMAKE options.
Hopefully I get around to submitting a patch to get ceph14 to v14.2.20 
real soon now.

--WjW




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?0a6c7864-e9e7-dc47-793c-914be52ca582>