Hi guys,

Thanks for the help.
Since I couldn't fix the issue, then I decided to open an issue in
github (https://github.com/JuliaLang/julia/issues/5920).

Best regards,
Ronan

2014-02-22 16:06 GMT-03:00 Ronan Chagas <[email protected]>:
> I tried to build using `make dist` but I still got the UndefVarError
> problem, just like here
> https://build.opensuse.org/package/live_build_log/home:Ronis_BR/julia/openSUSE_13.1/i586
>
> 2014-02-22 11:51 GMT-03:00 andrew cooke <[email protected]>:
>> ah, thanks; i hadn't understood that.
>>
>>
>> On Saturday, 22 February 2014 11:43:34 UTC-3, Patrick O'Leary wrote:
>>>
>>> To clarify, this only affects the precompiled standard library image,
>>> sys.{so,dylib,dll}. Everything else is fine, and at the cost of startup time
>>> you can delete this file (or use the `make dist` target) and distribute the
>>> result.
>>>
>>> On Saturday, February 22, 2014 5:31:25 AM UTC-6, andrew cooke wrote:
>>>>
>>>> one more point.  ronan, i don't know what you're using opensuse's build
>>>> service for, but there are issues at the moment with binary builds of 
>>>> julia.
>>>> see https://github.com/JuliaLang/julia/issues/5293 which if i understand
>>>> correctly, means that currently 0.3 cannot be pre-built on one machine and
>>>> then used on another (reliably). andrew
>>>>
>>>> On Friday, 21 February 2014 20:23:55 UTC-3, andrew cooke wrote:
>>>>>
>>>>>
>>>>> so i got carried away thinking i had found something useful when i
>>>>> should have read the instructions...
>>>>>
>>>>> with that, and installing m4, it builds fine.
>>>>>
>>>>> andrew
>>>>>
>>>>>
>>>>> On Friday, 21 February 2014 18:58:33 UTC-3, Elliot Saba wrote:
>>>>>>
>>>>>> The first thing you should try is restricting the CPU features
>>>>>> OpenBLAS is trying to use.  Try passing one of the following options
>>>>>> to `make` when building Julia:
>>>>>>
>>>>>> * `OPENBLAS_TARGET_ARCH=BARCELONA` (AMD CPUs)
>>>>>>
>>>>>> * `OPENBLAS_TARGET_ARCH=NEHALEM` (Intel CPUs)
>>>>>>
>>>>>> These will force OpenBLAS to build for a certain architecture that
>>>>>> doesn't have things like AVX instructions, etc.... which can cause
>>>>>> problems.
>>>>>> -E
>>>>>>
>>>>>> On Fri, Feb 21, 2014 at 11:05 AM, andrew cooke <[email protected]>
>>>>>> wrote:
>>>>>> > ok, so i've spent a large chunk of the day playing with qemu.  it's
>>>>>> > horrible.
>>>>>> >
>>>>>> > i finally got a system running with network access using the recipe
>>>>>> > at
>>>>>> > http://acooke.org/cute/StartingQe0.html
>>>>>> >
>>>>>> > when i duplicate the work from yesterday there (git clone; make) i
>>>>>> > get an
>>>>>> > error in the middle of building cblas.  interestingly, it looks like
>>>>>> > the
>>>>>> > problems start in a routine called getarch_2nd.c - i guess these are
>>>>>> > the
>>>>>> > cpu/architecture detection issues mentioned earlier.
>>>>>> >
>>>>>> > the way that i started qemu is very different to the command you
>>>>>> > gave.  i
>>>>>> > have no idea whether either is wrong in some way.  i really am not at
>>>>>> > all
>>>>>> > happy with either qemu or libvirt - they seem to be extremely flakey
>>>>>> > pieces
>>>>>> > of software :o(
>>>>>> >
>>>>>> > is there an issue raised for this?  what info can i provide that
>>>>>> > would be
>>>>>> > useful?  i can't really upload the VM as it's 8GB.  if someone else
>>>>>> > wants to
>>>>>> > duplicate what i see (which isn;t what the original error was, but
>>>>>> > might be
>>>>>> > related) just follow the recipe in the link i gave, use yast to
>>>>>> > install
>>>>>> > make, gcc, g+, patch, ncurses-devel, gfortran and git, then clone
>>>>>> > julia and
>>>>>> > type make...
>>>>>> >
>>>>>> > i guess at a minimum i can make > log 2>&1 which will give some idea
>>>>>> > of what
>>>>>> > went wrong.  what other info would help?
>>>>>> >
>>>>>> > andrew
>>>>>> >
>>>>>> >
>>>>>> > On Friday, 21 February 2014 09:23:04 UTC-3, Ronan Chagas wrote:
>>>>>> >>
>>>>>> >> Yes I have:
>>>>>> >>
>>>>>> >> /usr/bin/qemu-kvm -no-reboot -nographic -vga none -net none -cpu
>>>>>> >> host
>>>>>> >> -kernel /boot/vmlinuz -initrd /boot/initrd-3.11.10-7-desktop-virtio
>>>>>> >> -append root=/dev/sda panic=1 quiet no-kvmclock nmi_watchdog=0 rw
>>>>>> >> elevator=noop console=ttyS0 init=/.build/build -hda
>>>>>> >> /var/tmp/build-root/openSUSE_13.1-x86_64/img -drive
>>>>>> >>
>>>>>> >>
>>>>>> >> file=/var/tmp/build-root/openSUSE_13.1-x86_64/swap,if=ide,index=1,cache=unsafe
>>>>>> >> -smp 8
>>>>>> >>
>>>>>> >> Best regards,
>>>>>> >> Ronan
>>>>>> >>
>>>>>> >> 2014-02-21 9:05 GMT-03:00 andrew cooke <[email protected]>:
>>>>>> >> >
>>>>>> >> > ok, so running on vbox overnight worked.
>>>>>> >> >
>>>>>> >> > i need to do some (paid) work now, but when i get free time (lunch
>>>>>> >> > or
>>>>>> >> > evening) i'll try to run the same image under qemu/kvm and repeat.
>>>>>> >> >
>>>>>> >> > do you have the command you used to start qemu?
>>>>>> >> >
>>>>>> >> > thanks,
>>>>>> >> > andrew
>>>>>> >> >
>>>>>> >> >
>>>>>> >> >
>>>>>> >> > On Thursday, 20 February 2014 22:58:47 UTC-3, Ronan Chagas wrote:
>>>>>> >> >>
>>>>>> >> >> Hi Andrew,
>>>>>> >> >>
>>>>>> >> >> Em 20-02-2014 21:54, andrew cooke escreveu:
>>>>>> >> >> >
>>>>>> >> >> > Still running.  I need to go to bed, but assuming this works
>>>>>> >> >> > then
>>>>>> >> >> > tomorrow I think that I can start up the same VM in qemu and
>>>>>> >> >> > try
>>>>>> >> >> > again
>>>>>> >> >> > (I have never used qemu, so if someone can post an example use
>>>>>> >> >> > that
>>>>>> >> >> > would be great).  If that then fails then there would be clear
>>>>>> >> >> > evidence
>>>>>> >> >> > that it's KVM (and of course it may still fail tonight inside
>>>>>> >> >> > VB).
>>>>>> >> >> > Andrew
>>>>>> >> >> >
>>>>>> >> >>
>>>>>> >> >> Thanks! I'll wait for the results. I tried to build Julia using
>>>>>> >> >> all
>>>>>> >> >> deps
>>>>>> >> >> downloaded by the script. The result: the same error...
>>>>>> >> >>
>>>>>> >> >> Thus, we can conclude that it is not something related with the
>>>>>> >> >> libraries shipped with openSUSE 13.1.
>>>>>> >> >>
>>>>>> >> >> Best regards,
>>>>>> >> >>
>>>>>> >> >> Ronan

Reply via email to