Date: Sat, 9 Dec 2017 07:44:17 -0600 From: Mark Linimon <linimon@lonesome.com> To: Ed Schouten <ed@nuxi.nl> Cc: amdmi3@freebsd.org, amutu@amutu.com, "ed@FreeBSD.org" <ed@freebsd.org>, flo@freebsd.org, jjuanino@gmail.com, jlaffaye@freebsd.org, johans@freebsd.org, lev@freebsd.org, Mathieu Arnold <mat@freebsd.org>, Matthew Seaman <matthew@freebsd.org>, mono@freebsd.org, olgeni@freebsd.org, swills@freebsd.org, tcltk@freebsd.org, yuri@rawbw.com, linimon@freebsd.org Subject: Re: [linimon@FreeBSD.org: svn commit: r455813 - in head: audio/zita-convolver biology/phyml databases/percona56-server databases/pg_citus databases/soci devel/arpc devel/critcl devel/cxxtools devel/libffcall devel/ppl devel...] Message-ID: <20171209134417.GA20173@lonesome.com> In-Reply-To: <CABh_MK=yR94m4hs3GOtt5HOZQ%2BXCxF1WPeJoSXidZXMpa9pFDw@mail.gmail.com> References: <20171209004414.GA19019@lonesome.com> <CABh_MK=yR94m4hs3GOtt5HOZQ%2BXCxF1WPeJoSXidZXMpa9pFDw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Dec 09, 2017 at 08:05:36AM +0100, Ed Schouten wrote: > I also wasn't aware that it was broken, as I don't think I ever saw > any pkg-fallout emails. So, this is a fair question. It kind of goes to the difference between how we handle tier-1 and tier-2 architectures. e.g., I don't think portmgr is (yet?) doing pkg-fallout mails for armv6. The problem with armv6/7 is that we are nowhere near parity with amd64 yet. I'm trying to get us closer. I think before we can think about declaring armv6 (or even aarch64) as tier-1, we have a lot of work to do. So, I am going through all the errorlogs and marking things BROKEN so that we can a) stop trying to (re)build hundreds of ports on the cluster and b) make all the problems browsable at a glance. My intention is to issue a call for fixes once I have gone through all the armv6 _and_ armv7 errorlogs. (portmgr is not yet building armv7 so I am building them on my own 2x2 32G amd64 machine.) There are simply too many failing ports for me to fix them all on my own. I am hoping that some 'arm experts' can find common failure cases more quickly than individual maintainers. As a starting point I am looking at the cluster results, then trying to build them on my own system (both for armv6 and armv7) for confirmation. In this particular case the cluster errorlog is: http://beefy8.nyi.freebsd.org/data/head-armv6-default/p453785_s325568/logs/errors/arpc-0.6.log as shown on the page http://beefy8.nyi.freebsd.org/build.html?mastername=head-armv6-default&build=p453785_s325568 which is the last full armv6 -CURRENT build. In that build you'll see 130 Failed ports, and all with the uninformative tags such as "clang", "linker_error", etc. (There should be 16 less on the next run.) In Ignored we have 969, in which 666 of the Reasons match on the string "armv", and thus a rough proxy for architecture-specific. (A quick scan of all Reasons sortedby alpha seems to confirm this.) The Reasons are much more informative than the tags. My own reasoning for doing things in this order on all the tier-2s is that not everyone has access to either hardware, or has a qemu instance set up. (This has been a particular sore point with maintainers w/rt powerpc64 and sparc64, in which case the qemu route is useless.) My goal is to have everything labeled by the end of the year, so we can have fans of each architecture start working through the big lists, and then later on start contacting individual maintainers. Finally, I have not yet tried to build ports that might or might not be broken via qemu bugs, on actual hardware. I intend to try some of these, but it depends on how much time I can allocate. mcl
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20171209134417.GA20173>