Confused about ARM64 cross-compilation

Karl Denninger karl at denninger.net
Fri Feb 9 19:19:23 UTC 2018


Just rebuilt world and kernel, now on
FreeBSD 11.1-STABLE #1 r329071: Fri Feb  9 12:49:30 CST 2018    
karl at NewFS.denninger.net:/usr/obj/usr/src/sys/KSD-SMP

clang --version did not change

[\u at NewFS /pics]# clang --version
FreeBSD clang version 5.0.1 (tags/RELEASE_501/final 320880) (based on
LLVM 5.0.1)
Target: x86_64-unknown-freebsd11.1
Thread model: posix
InstalledDir: /usr/bin

Herbert, what revision of -HEAD are you building?

On 2/9/2018 12:14, Karl Denninger wrote:
> That's what I'm thinking right now too.  I'm building the
> latest-checked-out 11.1-STABLE at present.
>
> I have:
>
> [karl at NewFS ~]$ clang --version
> FreeBSD clang version 5.0.1 (tags/RELEASE_501/final 320880) (based on
> LLVM 5.0.1)
> Target: x86_64-unknown-freebsd11.1
> Thread model: posix
> InstalledDir: /usr/bin
>
> [karl at NewFS ~]$ uname -v
> FreeBSD 11.1-STABLE #0 r328086: Wed Jan 17 11:49:06 CST 2018    
> root at NewFS.denninger.net:/usr/obj/usr/src/sys/KSD-SMP
>
> He is on a later rev of 11.1 from his build rev, but whether CLANG was
> updated I do not know offhand.
>
>
> On 2/9/2018 12:10, Ian Lepore wrote:
>> On Fri, 2018-02-09 at 11:57 -0600, Karl Denninger wrote:
>>> What is your exact build string for cross-compile?
>>>
>>> This is what I'm using (and having fail) early in the cross-tools
>>> build,
>>> as the bug report shows:
>>>
>>> make TARGET_ARCH=aarch64 SRCCONF=/dev/null __MAKE_CONF=/dev/null
>>> -DNO_CLEAN -DWITHOUT_DEBUG_FILES -DWITHOUT_KERNEL_SYMBOLS
>>> -DWITHOUT_TESTS -j 12 buildworld
>>>
>>> -DNO_CLEAN being gone doesn't change behavior.
>>>
>> Since your failure happens when trying to use the 11.1-stable compiler
>> to build a new compiler, I would suspect the problem is in your base
>> 11.1 compiler.  Maybe H. Skuhra is on a newer revision of 11.1-stable
>> than you, in which some clang bug got fixed?  The output of uname -a
>> and clang --version on working versus failing systems may contain a
>> clue.
>>
>> -- Ian
>>
>>
>>> On 2/9/2018 11:54, Herbert J. Skuhra wrote:
>>>> On Fri, 09 Feb 2018 15:20:25 +0100, Karl Denninger wrote:
>>>>> Yeah, for the last couple of months attempting to build -HEAD on
>>>>> an
>>>>> 11.1-STABLE AMD64 architecture box fails; the current blowup is
>>>>> in the
>>>>> cross-compile tools, so you never get to the actual build itself.
>>>>>
>>>>> I logged a bug on this as it's a showstopper for a project of
>>>>> mine, and
>>>>> I have no workaround other than to spin up a -HEAD instance
>>>>> somewhere
>>>>> (which I assume works, but am not certain it will.)
>>>>>
>>>>> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225521
>>>> I am running 11.1-STABLE r328503 (amd64) and build HEAD for RPI2
>>>> (armv7) and RPI3 (aarch64) frequently.
>>>>
>>>> --
>>>> Herbert
>>>> _______________________________________________
>>>> freebsd-arm at freebsd.org mailing list
>>>> https://lists.freebsd.org/mailman/listinfo/freebsd-arm
>>>> To unsubscribe, send any mail to "freebsd-arm-unsubscribe at freebsd.o
>>>> rg"

-- 
Karl Denninger
karl at denninger.net <mailto:karl at denninger.net>
/The Market Ticker/
/[S/MIME encrypted email preferred]/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4897 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.freebsd.org/pipermail/freebsd-arm/attachments/20180209/c424d93c/attachment.bin>


More information about the freebsd-arm mailing list