Date: Wed, 19 Apr 2023 23:43:13 +0000 From: Brooks Davis <brooks@freebsd.org> To: Hans Petter Selasky <hselasky@freebsd.org> Cc: Brooks Davis <brooks@freebsd.org>, src-committers@freebsd.org, dev-commits-src-all@freebsd.org, dev-commits-src-main@freebsd.org Subject: Re: git: a7469c9c0a50 - main - libc: bsort_s() requires both __BSD_VISIBLE and __EXT1_VISIBLE Message-ID: <ZEB8kbKhKcdu9KTc@spindle.one-eyed-alien.net> In-Reply-To: <c9798234-3b60-00dd-2c55-a6351d0a0c22@freebsd.org> References: <202304192013.33JKDIrM070521@gitrepo.freebsd.org> <ZEBPQPzdmhhCNQ5I@spindle.one-eyed-alien.net> <c9798234-3b60-00dd-2c55-a6351d0a0c22@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Apr 19, 2023 at 11:35:56PM +0200, Hans Petter Selasky wrote: > On 4/19/23 22:29, Brooks Davis wrote: > > This is a formal request to revert all commits related to bsort. It > > should not have been committed without much broader discussion and IMO > > does not belong in the tree. It certainly should not be in the tree > > under such a generic name. > > > > -- Brooks > > Hi Brooks, > > I don't have an issue reverting my bsort() patch series, but please > clarify your statement first. Who are "we" this time, representing this > formal request for revert? This is my request. I see some review and the thread below, but adding non-standard symbols that are likely to collide with other software[0] to libc should be subject to a higher bar than a few people helping you improve your patch of saying "that's neat". New things added to libc should be in a standard or aligned with one (e.g., strlcpy, etc) and anything not from a standard should have immediate uses where it improves things in the rest of the system. Critically I don't see plans or prototype conversions and I don't see benchmarks of real systems (which could easily be done with LD_PRELOAD. > Regarding "broader discussion" - what do you mean? > > The initial discussion was started last September: > > https://lists.freebsd.org/archives/freebsd-arch/2022-September/000225.html More pushback here probably would be been good, sorry. A heads up before the actual commit might have been a good idea. I personally find your answer to the question, "why not improve qsort instead?" unsatisfactory. It might be that importing your implementation makes sense, but I don't think making is a public symbol we're stuck with forever if we ship it in 14 is a good way to go. > And several people have been asked for review and comments. Please > elaborate what "broader discussion" means? Do you mean like getting > something into ANSI first? I don't get it. I'd like more "I'd use it for X" and less "that's neat". -- Brooks [0] Debian code search finds fewer collisions than I'd feared, but not zero: https://codesearch.debian.net/search?q=%5B%5Ea-z%5Dbsort%5C%28&literal=0
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ZEB8kbKhKcdu9KTc>