The rc.c/ceph script uses too many paramters. Index: net/ceph/files/ceph.in =================================================================== --- net/ceph/files/ceph.in (revision 476830) +++ net/ceph/files/ceph.in (working copy) @@ -32,5 +32,5 @@ restart_cmd="%%PREFIX%%/bin/init-ceph $*" condrestart_cmd="%%PREFIX%%/bin/init-ceph $*" -run_rc_command "$1" "$*" +run_rc_command "$1"
I'll take this to avoid bumping PORTREVISON twice for some other issues listed below. The checksum and size for ceph-ceph-erasure-code-corpus-2d7d78b_GH0.tar.gz is wrong. It looks like it got changed in r474922 even though the git hash did not change. I am coming up with what the port had before: SHA256 (ceph-ceph-erasure-code-corpus-2d7d78b_GH0.tar.gz) = 466f7185015df8d13f8b2b9a17ee30ab419bcd667284ce2b6d32a1128c4640f1 SIZE (ceph-ceph-erasure-code-corpus-2d7d78b_GH0.tar.gz) = 3634266 Also, I am working an update for security/cryptopp. This port has it listed in LIB_DEPENDS even though it does not use it. In the main CMakeLists.txt for ceph, it will only use cryptopp if nss is explicitly disabled. I plan on removing the cryptopp dependency because there are API changes and this will be one less port to worry about. You could make an option for it, but ceph would have to check the new pkgconfig file that will be provided by cryptopp to see whether it was built with assembly optimizations or not. Next time, please add your patches as an attachment and not as a comment. Thanks!
(In reply to Jason E. Hale from comment #1) Hi Jason, Thanx for picking this up. I will check again in my builds but could very well be that you are correct, and that this is a leftover from previous requirement. And I never noticed that it became a possible selection for either one. On the other hand, if it builds, it builds. ;-) If you fix it, I'll work from that further. This has gone thru poudriere, which should have complained about the wrong size and checksum. So I have no idea where that went wrong.
(In reply to Willem Jan Withagen from comment #2) Poudriere doesn't re-download and check distfiles. If you had a corrupt distfile in your distfiles cache and it matched the distinfo file, then poudriere would not complain as long as it extracted. When doing an update, it is usually a good idea after running 'make makesum' to run 'make distclean' and download again to make sure the checksums match. Before doing that, you should backup your old distfiles for the port so you can check for corruption or security compromise against the new ones. https://www.freebsd.org/doc/en/books/porters-handbook/dads-rerolling-distfiles.html
(In reply to Jason E. Hale from comment #1) I also found the cause for the "removal" of cryptopp: From: https://github.com/ceph/ceph/pull/23650 add Findcryptopp.cmake back cryptopp support was dropped in #20015, but it's required by src/msg/async/dpdk/TCP.h, which #include <cryptopp/md5.h> so, to fix the FTBFS of WITH_DPDK=ON, we need to bring Findcryptopp.cmake back. it was also removed in #20015. And https://github.com/ceph/ceph/pull/20015 goes: our cryptopp implementation was broken with the change to c++17, and the cost of fixing it likely outweighs the utility of keeping it around Now we currently do not build with DPDK, so it is safe to remove for the time being.
A commit references this bug: Author: jhale Date: Fri Aug 24 07:45:54 UTC 2018 New revision: 477944 URL: https://svnweb.freebsd.org/changeset/ports/477944 Log: - Fix rc.d script (too many parameters) - Fix distinfo (checksum and size were corrupted in last update) - Remove security/crytopp dependency that the port does not use - Fix typo in IGNORE message - Move pkgconf dependency to USES - Depend on python ports correctly - Remove CMAKE_BUILD_TYPE - it already is set to "Release" by the framework - Bump PORTREVISION for dependency changes PR: 230665 Submitted by: Willem Jan Withagen <wjw@digiware.nl> (maintainer) Changes: head/net/ceph/Makefile head/net/ceph/distinfo head/net/ceph/files/ceph.in
Committed, thanks!
(In reply to Jason E. Hale from comment #6) You too for patching up all the rough edges.