Date: Tue, 3 May 2022 22:13:26 +0700 From: Eugene Grosbein <eugen@grosbein.net> To: "Julian H. Stacey" <jhs@berklix.com>, freebsd-stable@FreeBSD.org Cc: freebsd-current@FreeBSD.org, Ed Maste <emaste@FreeBSD.org> Subject: Re: breaking modules Message-ID: <1143af76-e4fe-6169-0cc8-9397b54c2afd@grosbein.net> In-Reply-To: <202205031447.243EkmY5077471@fire.js.berklix.net> References: <202205031447.243EkmY5077471@fire.js.berklix.net>
next in thread | previous in thread | raw e-mail | index | archive | help
03.05.2022 21:46, Julian H. Stacey wrote: > Hi, Reference: >> From: "Julian H. Stacey" <jhs@berklix.com> >> Date: Fri, 29 Apr 2022 23:57:02 +0200 > > "Julian H. Stacey" wrote: >> Ed Maste wrote: >>> On Thu, 28 Apr 2022 at 11:28, Julian H. Stacey <jhs@berklix.com> wrote: >>>> >>>> but that's crude. It's nice to be able to build most modules ready >>>> in case wanted later, so how about a DUDS env. mechanism like ports/ ? >>> >>> I'd rather not add additional complexity to our build infrastructure >>> to address a situation that shouldn't exist. Modules should build & >>> function on an ongoing basis (and, I believe they generally do). CI >>> doesn't report any issues on either stable branch or main at present. >> >> I'm building stable-12 not stable-13. It's broken here. I've seen modules break >> for years, I used to suspect modules werent built by default by >> build engines as often as main src/, so modules had more time to rot against >> changing includes & libs, maybe now build engines might compile >> them as often as eg bin/ls/ ? I don't know; But I'm seeing modules breaks. >> >> I just refetched with git this mid Friday afternoon (TZ=+02:00) 12.3-STABLE >> & the 2 breaks are still present. See below. >> >> Setting a MODULE_DUDS would save work rather than repetitively retro >> patching out the same modules in Makefile after each git pull --ff-only. >> >> I'd happily develop a patch for sys/modules/, but if someone >> else prefers to, that might increase the chance of it being commited. >> I'd be happy to test or develop a fix for sys/modules/Makefile. > > Eugene Grosbein wrote: >> Unfortunately, CI does not catch stand-alone module build failures, >> out of kernel build directory. >> >> For example: >> >> if_em https://cgit.freebsd.org/src/commit/?id=c0460cf2e42d2819c1f191a1d6e1b3dc0c7ea010 >> if_epair https://cgit.freebsd.org/src/commit/?id=7a382e744b0b0ba9b51dc34bfa0cd1515f744f25 >> linuxkpi https://cgit.freebsd.org/src/commit/?id=f5a2e7b0e8483bf51519046fd149a6a31acef6b1 > > > I developed a fix, patch appended, mastered inc. a mini test Makefile at > http://berklix.com/~jhs/src/bsd/fixes/freebsd/src/gen/sys/modules/ > > Filed with https://bugs.freebsd.org/bugzilla/enter_bug.cgi as > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=263758 > > I added cc: current@, Would someone like to try it please ? > > BTW I've not yet but will later read how DUDS is implemented in > /usr/ports/Mk/bsd.port.subdir.mk I filled https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=263750 with a patch fixing standalone build of random_fortuna and random_other. Please test the patch in the PR 263750 and report back if it fixes the problem for you, too.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1143af76-e4fe-6169-0cc8-9397b54c2afd>