From patrick at gentoo.org Sun Aug 18 03:57:50 2013 From: patrick at gentoo.org (Patrick Lauer) Date: Sun, 18 Aug 2013 11:57:50 +0800 Subject: [Aboriginal] Kernel patching failure with 1.2.4.1 Message-ID: <5210463E.8070407@gentoo.org> Trying to build 1.2.4.1 fails when fetching/patching kernel: Confirmed linux-3.10.tar.bz2 === linux (host download) Extracting 'linux'........................................................................................................................................................................................$ Applying /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-arm.patch patching file arch/arm/mach-versatile/Kconfig patching file arch/arm/mm/Kconfig Hunk #5 succeeded at 359 (offset 4 lines). patching file arch/arm/mach-versatile/Kconfig patching file arch/arm/mach-versatile/pci.c Applying /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-fixsh4-2.patch patching file drivers/tty/serial/sh-sci.c Applying /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-fixsh4.patch patching file arch/sh/boot/compressed/head_32.S patching file arch/sh/include/asm/entry-macros.S patching file arch/sh/kernel/cpu/sh3/entry.S Hunk #1 succeeded at 258 (offset 1 line). patching file arch/sh/kernel/entry-common.S Hunk #1 succeeded at 104 (offset 6 lines). patching file arch/sh/lib/__clear_user.S Hunk #1 succeeded at 11 (offset -46 lines). Applying /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-fixuClibc.patch patching file include/uapi/linux/netlink.h Applying /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-hack.patch patching file init/Kconfig Hunk #1 succeeded at 1797 (offset 131 lines). Applying /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-noperl-timeconst.patch The next patch would create the file kernel/Makefile, which already exists! Assume -R? [n] Apply anyway? [n] Skipping patch. 1 out of 1 hunk ignored patching file kernel/timeconst.bc patching file kernel/Makefile Hunk #1 succeeded at 126 (offset 1 line). Exiting due to errors (host download linux) Exiting due to errors ( ) 1376798270.0 From prazzb at gmail.com Mon Aug 19 15:32:54 2013 From: prazzb at gmail.com (Prasanna Balan) Date: Mon, 19 Aug 2013 21:02:54 +0530 Subject: [Aboriginal] Kernel patching failure with 1.2.4.1 In-Reply-To: <5210463E.8070407@gentoo.org> References: <5210463E.8070407@gentoo.org> Message-ID: <20130819153254.GA7355@desktop> On 18/08/13, Patrick Lauer wrote: > Trying to build 1.2.4.1 fails when fetching/patching kernel: > > Confirmed linux-3.10.tar.bz2 > === linux (host download) > Extracting It does build cleanly on my gentoo box here. Confirmed means it got downloaded, there must be some issue during patching.Also this step (as much I understand) uses toybox patch and not the one from your host and (obviously) is architecture independent. Are you doing something different than default ?? Which distro ?? Are you using EXTRACT_ALL=1 ?? -prazzb 1376926374.0 From rob at landley.net Tue Aug 20 12:03:38 2013 From: rob at landley.net (Rob Landley) Date: Tue, 20 Aug 2013 07:03:38 -0500 Subject: [Aboriginal] Kernel patching failure with 1.2.4.1 In-Reply-To: <5210463E.8070407@gentoo.org> (from patrick@gentoo.org on Sat Aug 17 22:57:50 2013) Message-ID: <1377000218.2737.70@driftwood> On 08/17/2013 10:57:50 PM, Patrick Lauer wrote: > Trying to build 1.2.4.1 fails when fetching/patching kernel: Odd. It works here and I tried it on your gentoo server too. Did a fresh one in a new directory from the release tarball, and the kernel patching part went: Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-arm.patch patching file arch/arm/mach-versatile/Kconfig patching file arch/arm/mm/Kconfig Hunk #5 succeeded at 359 (offset 4 lines). patching file arch/arm/mach-versatile/Kconfig patching file arch/arm/mach-versatile/pci.c Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-fixsh4-2.patch patching file drivers/tty/serial/sh-sci.c Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-fixsh4.patch patching file arch/sh/boot/compressed/head_32.S patching file arch/sh/include/asm/entry-macros.S patching file arch/sh/kernel/cpu/sh3/entry.S Hunk #1 succeeded at 258 (offset 1 line). patching file arch/sh/kernel/entry-common.S Hunk #1 succeeded at 104 (offset 6 lines). patching file arch/sh/lib/__clear_user.S Hunk #1 succeeded at 11 (offset -46 lines). Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-fixuClibc.patch patching file include/uapi/linux/netlink.h Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-hack.patch patching file init/Kconfig Hunk #1 succeeded at 1797 (offset 131 lines). Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-noperl-timeconst.patch patching file kernel/mktimeconst.c patching file kernel/timeconst.bc patching file kernel/Makefile Hunk #1 succeeded at 126 (offset 1 line). Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-pmacorder.patch patching file drivers/ide/Makefile Applying /home/landley/aboriginal/aboriginal-1.2.4.1/sources/patches/linux-sh4-fixifconfig.patch patching file arch/sh/mm/consistent.c Hunk #1 succeeded at 84 (offset 2 lines). So that worked fine... > Confirmed linux-3.10.tar.bz2 > === linux (host download) > Extracting > 'linux'...............................................$ > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-arm.patch > patching file arch/arm/mach-versatile/Kconfig > patching file arch/arm/mm/Kconfig > Hunk #5 succeeded at 359 (offset 4 lines). > patching file arch/arm/mach-versatile/Kconfig > patching file arch/arm/mach-versatile/pci.c > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-fixsh4-2.patch > patching file drivers/tty/serial/sh-sci.c > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-fixsh4.patch > patching file arch/sh/boot/compressed/head_32.S > patching file arch/sh/include/asm/entry-macros.S > patching file arch/sh/kernel/cpu/sh3/entry.S > Hunk #1 succeeded at 258 (offset 1 line). > patching file arch/sh/kernel/entry-common.S > Hunk #1 succeeded at 104 (offset 6 lines). > patching file arch/sh/lib/__clear_user.S > Hunk #1 succeeded at 11 (offset -46 lines). > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-fixuClibc.patch > patching file include/uapi/linux/netlink.h > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-hack.patch > patching file init/Kconfig > Hunk #1 succeeded at 1797 (offset 131 lines). > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-noperl-timeconst.patch > The next patch would create the file kernel/Makefile, > which already exists! Assume -R? [n] Yes, the file already exists. But the patch _isn't_ trying to create it, it starts: --- linux/kernel/Makefile +++ linux/kernel/Makefile @@ -125,19 +125,13 @@ $(obj)/time.o: $(obj)/timeconst.h -quiet_cmd_hzfile = HZFILE $@ - cmd_hzfile = echo "hz=$(CONFIG_HZ)" > $@ - -targets += hz.bc -$(obj)/hz.bc: $(objtree)/include/config/hz.h FORCE - $(call if_changed,hzfile) - How you _create_ a file starting at line 125 is a strainge question. And the old one is linux/kernel/Makefile not /dev/null (and since there's no date, it's not dated the epoch which is the other way of saying file may not actually have existed). I.E. the patch isn't asking to create a new file, and there seems to be something wrong with your host's "patch" command for thinking it is. The patch command in ubuntu likes it, the patch command in toybox likes it (I wrote that one, which is why I know so much about the minutiae of patch format), and the busybox patch command likes it (although since the current version's a port of the toybox patch command, that's not saying much). Rob 1377000218.0 From rob at landley.net Tue Aug 20 12:58:15 2013 From: rob at landley.net (Rob Landley) Date: Tue, 20 Aug 2013 07:58:15 -0500 Subject: [Aboriginal] Kernel patching failure with 1.2.4.1 In-Reply-To: <20130819153254.GA7355@desktop> (from prazzb@gmail.com on Mon Aug 19 10:32:54 2013) References: <5210463E.8070407@gentoo.org> <20130819153254.GA7355@desktop> Message-ID: <1377003495.2737.72@driftwood> On 08/19/2013 10:32:54 AM, Prasanna Balan wrote: > On 18/08/13, Patrick Lauer wrote: > > Trying to build 1.2.4.1 fails when fetching/patching kernel: > > > > Confirmed linux-3.10.tar.bz2 > > === linux (host download) > > Extracting > > It does build cleanly on my gentoo box here. > Confirmed means it got downloaded, there must be some issue during > patching.Also this step (as much I understand) uses toybox patch and > not the one from your host and (obviously) is architecture > independent. Depends how you do it. If you run "EXTRACT_ALL=1 ./download.sh" then it extracts everything into the package cache (build/packages) and applies the patches up front, before running host-tools.sh. Later builds reuse the package cache if the sources/patches and packages directories haven't changed. If you just do ./build.sh with an empty/nonexistent build directory, it extracts packages as needed, so it builds host-tools long before extracting the kernel. So in that case, it would use the toybox patch. Both should work, though. > Are you doing something different than default ?? Which distro ?? Are > you using EXTRACT_ALL=1 ?? If you use more/buildall.sh it'll do EXTRACT_ALL=1 as part of the setup. (That way it can do the builds in parallel without the package cache initialization causing race conditions.) Rob 1377003495.0 From rob at landley.net Thu Aug 22 20:49:43 2013 From: rob at landley.net (Rob Landley) Date: Thu, 22 Aug 2013 15:49:43 -0500 Subject: [Aboriginal] Linux From Scratch build. Message-ID: <1377204583.2737.109@driftwood> The existing LFS build got broken by the posix_spawn() patch, because gnu/dammit m4 is detecting it's there and trying to grab gnu-specific undocumented fields out of the guts of some structure in its implementation. Those fields aren't there in the backported uClibc patch, so the build breaks. How you screw up something as simple as posix_spawn() (it's fork+exec, it's a non-blocking system() that doesn't go through a shell), I couldn't tell you. But they managed. I've tried it with and without the #define in the headers announcing posix_spawn's presence. (Why you need a #define in the headers to announce a posix feature when you're already claiming posix-2008... let's just gloss over that for now). I'm thinking this might be a good excuse to upgrade from LFS 6.8 to LFS 7.4. The -rc1 of that is out, and they're actually doing a matching BLFS release for the first time since 2008. Unfortunately, this would mean I'm not switching _from_ a known working environment, although it works fine if you yank the posix_spawn patch. So the downside would be it not fixing the bug _and_ me winding up with a whole lot more work on my plate, but it's work I've been meaning to do anyway... So there's that. I also need to test the 3.11 kernel. Toybox grep is finally working in the build (exit code was wrong for everything but -q mode, fixed it). Still havent' gotten back to the ccwrap rewrite for musl. It's at about 50% done, and has been for ~2 weeks... Rob 1377204583.0 From maillist-aboriginal at barfooze.de Thu Aug 22 22:09:43 2013 From: maillist-aboriginal at barfooze.de (John Spencer) Date: Fri, 23 Aug 2013 00:09:43 +0200 Subject: [Aboriginal] Linux From Scratch build. In-Reply-To: <1377204583.2737.109@driftwood> References: <1377204583.2737.109@driftwood> Message-ID: <52168C27.3090309@barfooze.de> On 08/22/2013 10:49 PM, Rob Landley wrote: > The existing LFS build got broken by the posix_spawn() patch, because > gnu/dammit m4 is detecting it's there and trying to grab gnu-specific > undocumented fields out of the guts of some structure in its > implementation. Those fields aren't there in the backported uClibc > patch, so the build breaks. IIRC the problem is due to m4's autoconf test omitting the necessary -lrt LDFLAGS. 1377209383.0 From rob at landley.net Sun Aug 25 05:39:24 2013 From: rob at landley.net (Rob Landley) Date: Sun, 25 Aug 2013 00:39:24 -0500 Subject: [Aboriginal] Linux From Scratch build. In-Reply-To: <52168C27.3090309@barfooze.de> (from maillist-aboriginal@barfooze.de on Thu Aug 22 17:09:43 2013) References: <1377204583.2737.109@driftwood> <52168C27.3090309@barfooze.de> Message-ID: <1377409164.2737.114@driftwood> On 08/22/2013 05:09:43 PM, John Spencer wrote: > On 08/22/2013 10:49 PM, Rob Landley wrote: >> The existing LFS build got broken by the posix_spawn() patch, because >> gnu/dammit m4 is detecting it's there and trying to grab gnu-specific >> undocumented fields out of the guts of some structure in its >> implementation. Those fields aren't there in the backported uClibc >> patch, so the build breaks. > > IIRC the problem is due to m4's autoconf test omitting the necessary > -lrt LDFLAGS. I fixed it by making the patch add it to libc instead of -lrt. (I prefer not to patch packages that built against glibc, on the theory they're not the only one doing this so I should fix the build environment.) I'm finally migrating to the new much more powerful laptop, and there's likely to be some email disruption while I do this. I rsynced my files over but am _not_ setting balsa (which is crap) up on the new machine, so need to rail against open source's horrible inability to do UI before grudgingly reinstalling thunderbird find a new mail client and set it up. I also have two copies of my files and am trying to make sure not to accidentally delete updates to either one while keeping them in sync, until I can decommission the netbook once I'm no longer using it for email... Rob 1377409164.0 From rob at landley.net Sun Aug 25 05:43:23 2013 From: rob at landley.net (Rob Landley) Date: Sun, 25 Aug 2013 00:43:23 -0500 Subject: [Aboriginal] Linux From Scratch build. In-Reply-To: <52168C27.3090309@barfooze.de> (from maillist-aboriginal@barfooze.de on Thu Aug 22 17:09:43 2013) References: <1377204583.2737.109@driftwood> <52168C27.3090309@barfooze.de> Message-ID: <1377409403.2737.115@driftwood> On 08/22/2013 05:09:43 PM, John Spencer wrote: > On 08/22/2013 10:49 PM, Rob Landley wrote: >> The existing LFS build got broken by the posix_spawn() patch, because >> gnu/dammit m4 is detecting it's there and trying to grab gnu-specific >> undocumented fields out of the guts of some structure in its >> implementation. Those fields aren't there in the backported uClibc >> patch, so the build breaks. > > IIRC the problem is due to m4's autoconf test omitting the necessary > -lrt LDFLAGS. Oh, status report. The new laptop build aboriginal for all targets and built LFS under all those targets except sparc, which broke in the "bison" build due to yet more posix_spawn() confusion that seems to only hit sparc? Weeeeird. Not quite sure what's up there, but not hugely bothered by it either since sparc's new and not as tested as the others. Well, and these targets don't really work either: m68k: still no emulator to run it. sh4: still only 64 megs on the board. armeb: still unfinished. alpha: half-implemented. mips64: I'm using the old API which seems to have bit rotted, need to redo. But arm, power, mips, and x86 all built. So releasing with linux for workgroups should be fine when it comes out. Rob 1377409403.0 From maillist-aboriginal at barfooze.de Sun Aug 25 20:51:33 2013 From: maillist-aboriginal at barfooze.de (John Spencer) Date: Sun, 25 Aug 2013 22:51:33 +0200 Subject: [Aboriginal] Linux From Scratch build. In-Reply-To: <1377409403.2737.115@driftwood> References: <1377204583.2737.109@driftwood> <52168C27.3090309@barfooze.de> <1377409403.2737.115@driftwood> Message-ID: <521A6E55.9090309@barfooze.de> glad to hear the issue is resolved. On 08/25/2013 07:43 AM, Rob Landley wrote: > > But arm, power, mips, and x86 all built. So releasing with linux for > workgroups should be fine when it comes out. another release before your switch to musl is very welcome, as i fear the port will take a lot of time ... > > Rob --JS 1377463893.0 From patrick at gentoo.org Fri Aug 30 05:58:35 2013 From: patrick at gentoo.org (Patrick Lauer) Date: Fri, 30 Aug 2013 13:58:35 +0800 Subject: [Aboriginal] Kernel patching failure with 1.2.4.1 In-Reply-To: <5210463E.8070407@gentoo.org> References: <5210463E.8070407@gentoo.org> Message-ID: <5220348B.1040202@gentoo.org> On 08/18/2013 11:57 AM, Patrick Lauer wrote: > Trying to build 1.2.4.1 fails when fetching/patching kernel: > patching file init/Kconfig > Hunk #1 succeeded at 1797 (offset 131 lines). > Applying > /home/me/code/aboriginal-1.2.4.1/sources/patches/linux-noperl-timeconst.patch > The next patch would create the file kernel/Makefile, > which already exists! Assume -R? [n] > Apply anyway? [n] > Skipping patch. > 1 out of 1 hunk ignored > patching file kernel/timeconst.bc > patching file kernel/Makefile > Hunk #1 succeeded at 126 (offset 1 line). > > Exiting due to errors (host download linux) > > Exiting due to errors ( ) This is due to a change in behaviour with patch - version <=2.6.1 work fine, >=2.7 fail. 1377842315.0