Date: Fri, 7 Feb 2020 01:05:56 +0000 From: Mark Linimon <linimon@lonesome.com> To: Mathieu Arnold <mat@FreeBSD.org> Cc: Emmanuel Vadot <manu@bidouilliste.com>, Antoine Brodin <antoine@FreeBSD.org>, ports-committers@freebsd.org, svn-ports-all@freebsd.org, svn-ports-head@freebsd.org Subject: Re: svn commit: r525136 - head/devel Message-ID: <20200207010556.GC15808@lonesome.com> In-Reply-To: <20200204122429.afyvp4w5xx3kb4wk@ogg.in.absolight.net> References: <202002040936.0149aI3E088204@repo.freebsd.org> <20200204104020.cba025dc63af413d7688fd8d@bidouilliste.com> <20200204122429.afyvp4w5xx3kb4wk@ogg.in.absolight.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Feb 04, 2020 at 01:24:29PM +0100, Mathieu Arnold wrote: > This is why we have Tools/script/addport. Allow me to disagree by playing devil's advocate. (disclaimer: I am *not* picking on any one partiular committer here -- although, unfortunately, some of the people I've been sending private email to have taken it that way.) <rant> No, this is why we have *committers*. After I last posted last month about "please folks, test before commit", somewhere between 10 and 20 new ports have been added that simply. did. not. work. I'm *not* complaining that they did not work on powerpc64 -- that's a task that I picked up all by myself; it's my decision whether to do it or not. But: - "make patch" must at least work. This requires almost no time. This means the Makefile has to be sane, the files have to fetch+, all the files intended to be committed must be committed, and the patches must be right. (Suggestion: use "make distclean" then "make patch" to double-check fetching.) - "make describe" has to at least work. This requires no time. - Feeding the port to poudriere++ *must* not fail due to the metadata in <category>/Makefile being missing or misspelled. Otherwise, someone (TM) has to make a commit to "fix bulk -a". (You may have noticed how many there are these days.) By skipping these simple tests, committers push work onto either current portmgrs or other people (including past portmgrs), leading to wasted time, resentment, and burnout. Committers *should* have already accepted that they are the gateway between raw patches and a working ports tree. Please, please, can we now get back to that? </rant> Look, I know we all make mistakes. But the things above result from people being in "too much of a hurry to test". Thanks. mcl + fetch, preferably from an original site, not just a FreeBSD.org echo ++ poudriere is what the FreeBSD package builders use. A commit that builds manually yet not under poudriere may be "technically" correct but users of packages won't think so.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20200207010556.GC15808>