Date: Tue, 17 Sep 2013 06:58:19 -0400 From: George Mitchell <george+freebsd@m5p.com> To: Chie Taguchi <taguchi.ch@gmail.com> Cc: "freebsd-arm@freebsd.org" <freebsd-arm@freebsd.org> Subject: Re: What's the recipe? Message-ID: <523835CB.2070900@m5p.com> In-Reply-To: <8B750CD2-53BB-478A-8C7E-369226AA9997@gmail.com> References: <522229C0.5030504@m5p.com> <5235D207.5070400@m5p.com> <6AE316FC-8B08-4F5D-B918-7F97C8C4CB60@gmail.com> <5237BC28.8030709@m5p.com> <8B750CD2-53BB-478A-8C7E-369226AA9997@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 09/17/13 00:07, Chie Taguchi wrote:
> It will cause following issue in building devel/glib20:
>
> /bin/mkdir -p '/usr/local/include/glib-2.0/gio/'
> <jemalloc>: jemalloc_arena.c:380: Failed assertion: "p[i] == 0"
> Abort trap (core dumped)
>
> This is a very annoy problem, and glib20 ignore this error.
> As a result, glib20 fail to install some header files.
>
> This workaround is rebuild your image with "MALLOC_PRODUCTION=yes".
>
> Thanks.
>
> C.Taguchi
Thanks! I'll try that tonight. -- George
>
>
> On 2013/09/17, at 11:19, George Mitchell <george+freebsd@m5p.com> wrote:
>
>> On 09/16/13 22:00, Chie Taguchi wrote:
>>> Have you tried ports/181140?
>>>
>>> C.Taguchi
>>
>> Yes; google pointed me there a little earlier today. But the syntax
>> of the patch appears to be slightly off. Here's what works for me:
>>
>> ...
>> .if ! {PORT_OPTIONS:MSIMD}
>> CONFIGURE_ARGS+= --disable-vmx --disable-arm-simd
>> .if ! {MACHINE_CPU:Mmmx}
>> CONFIGURE_ARGS+= --disable-mmx
>> .endif
>> .if ! {MACHINE_CPU:Msse2}
>> CONFIGURE_ARGS+= --disable-sse2
>> .endif
>> .endif
>>
>> .if ${ARCH} == arm || ${ARCH} == armv6
>> CONFIGURE_ARGS+= --disable-arm-neon
>> .endif
>> ...
>>
>> (The patch had "${MACHINE_CPU:Mmmx}" and "${MACHINE_CPU:Msse2}", but
>> make didn't like the dollar signs.)
>>
>> My next problem is that my devel/glib20 build somehow didn't populate
>> /usr/local/include/glib-2.0/gio, so devel/gobject-introspection couldn't
>> build. I'm rerunning the devel/glib20 build to see if it works better
>> the second time. -- George
>>
>>>
>>> On 2013/09/16, at 0:28, George Mitchell <george+freebsd@m5p.com> wrote:
>>>
>>>> On 08/31/13 13:37, George Mitchell wrote:
>>>>> Have you built a working Raspberry Pi image recently? If so, for the
>>>>> benefit of the rest of us, could you share a few secrets?
>>>>>
>>>>> 1. What system did you do the build on? If it was an i386 or amd64,
>>>>> what svn version was it built with?
>>>>
>>>> Built on amd64, svn version r255361M. (Two changes: I still distrust
>>>> SCHED_ULE, and there's a patch [attached] required for the ARM xdev
>>>> build to complete.)
>>>>>
>>>>> 2. What did you have in /etc/src.conf and /etc/make.conf, both for
>>>>> building the build system itself and for building the RPi?
>>>>
>>>> I finally bit the bullet and went to clang. /etc/src.conf is empty;
>>>> /etc/make.conf contains "WITH_PKGNG=yes".
>>>>>
>>>>> 3. What svn version of /usr/src did you use in building the RPi image?
>>>>>
>>>> Same as above (r255361M).
>>>>
>>>>> 4. Did you use crochet? If so, what was the last commit in your git
>>>>> log?
>>>> commit b154a2f1252eaa5f4d821439d0d9005cea94e580
>>>> Merge: a6f4cc5 8d03c23
>>>> Author: Tim Kientzle <kientzle@acm.org>
>>>> Date: Sat Aug 17 12:00:52 2013 -0700
>>>>
>>>>>
>>>>> When I say "working," I'm hoping for the ability to run stably for a
>>>>> number of days, running NFS and CUPS. I've been doing this since
>>>>> January with a precompiled image I downloaded then which worked
>>>>> wonderfully with one of my printers, but not the other one. Now
>>>>> there's a patch that enables both printers to work, and I would love
>>>>> to build a new image. So I've been thrashing around trying to find
>>>>> the answers to the questions above without success. Thanks for any
>>>>> help you can give! -- George
>>>>
>>>> This build is marginally stable. I've been working for a week to get
>>>> a complete build of print/cups and I'm not quite there. From the RPi,
>>>> /usr/ports is an NFS mount from another machine with version 326306 of
>>>> the ports tree (just before the iconv incident). The RPi likes to get
>>>> messages of this sort:
>>>>
>>>> http://www.m5p.com/~george/pi/DSCN1282.JPG
>>>>
>>>> primarily during the period that portmaster is discovering everybody's
>>>> dependencies. (Oddly, if I make it to the build stage, the build
>>>> completes about 90% of the time. devel/glib20 built, including its
>>>> dependencies on perl and python, in one overnight session.)
>>>>
>>>> Once, instead of the fatal kernel mode prefetch abort, I got:
>>>>
>>>> panic: __rw_wlock_hard: recursing but non-recursive rw pmap pw @ /usr/src/sys/arm/arm/pmap-v6.c:1406
>>>> KDB: enter: panic [ thread pid 906 tid 100076 ]
>>>> stopped at $d: ldrb r15, [r15, 15, ror r15]
>>>>
>>>> (Sorry for the transcription by hand; I have an AdaFruit serial cable
>>>> on order.)
>>>>
>>>> My current problem is that x11/pixman fails to build:
>>>>
>>>> ===> Building for pixman-0.30.0
>>>> [...]
>>>> CC pixman-arm-neon.lo
>>>> CPPAS pixman-arm-neon-asm.lo
>>>> /usr/include/float.h: Assembler messages:
>>>> /usr/include/float.h:43: Error: bad instruction `extern int __flt_rounds(void)'
>>>> [...]
>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?523835CB.2070900>