Tensorflow port
Li-Wen Hsu
lwhsu at freebsd.org
Sun Feb 11 12:08:57 UTC 2018
Just checked building devel/bazel in poudriere and it built fine. It seems
that you have BUILD_AS_NON_ROOT in poudriere.conf so it cannot stage the
port, you might want to disable it or set sysctl
security.bsd.hardlink_check_uid and/or security.bsd.hardlink_check_gid to 1
(not recommended).
I also tested attached bazel5 but it doesn't have good luck:
https://gist.github.com/02694ad2f9f9abf78078c85362d3efe4
It looks we have some Java problems here.
May I ask why tensorflow needs an old version of bazel?
Best,
Li-Wen
On Sun, Feb 11, 2018 at 10:02 AM, Jov <amutu at amutu.com> wrote:
> When I try to create bazel5 port ,I get this error :
>
> poudriere testport -vv -j 111amd64 devel/bazel5
> =>> Debug: Reading /usr/local/etc/poudriere.conf
> [00:00:00] Warning: You have BUILD_AS_NON_ROOT set to 'yes' (c.f.
> poudriere.conf),
> [00:00:00] Warning: and 'security.bsd.hardlink_check_uid' or
> 'security.bsd.hardlink_check_gid' are not set to '0'.
> [00:00:00] Error: Poudriere will not be able to stage some ports. Exiting.
>
> This is a fresh ports tree. Testport for devel/bazel also report this
> error.
>
> I can not search any info for this error.
>
> Att is the bazel5 port, extract from the last known good 5.x port(ports
> r448218).
>
> As my time is limited, I will not do the work further.
>
> Jov
>
> 2018-02-08 9:06 GMT+08:00 Jason Bacon <bacon4000 at gmail.com>:
>
>>
>> Thanks, Li-Wen! That will be helpful until someone finds time to upgrade
>> it.
>>
>> On 02/02/18 02:10, Li-Wen Hsu wrote:
>>
>>> Hi Jov,
>>>
>>> I just noticed this issue and I think it is an acceptable way to create
>>> new bazel5 port if TensorFlow really needs it. I added the TensorFlow port
>>> so I am happy to work with you to fix it.
>>>
>>> Best,
>>> Li-Wen
>>>
>>> On Fri, Dec 15, 2017 at 10:06 AM, Jov <amutu at amutu.com <mailto:
>>> amutu at amutu.com>> wrote:
>>>
>>> I tried to fix and update the py-tensorflow last week but failed
>>> because
>>> the bazel and tensorflow upstream seem heavily development and a
>>> lot of
>>> things changed. My time for the ports work is limited so I will
>>> give up
>>> this port. Anyone can take it and I am fine to answer any question.
>>>
>>> The easiest fixing is to create a new port bazel5 with 0.5.3 as
>>> version and
>>> change the tensorflow depend on this new port.
>>>
>>> Jov
>>>
>>> 2017-12-15 9:21 GMT+08:00 Jason Bacon <bacon4000 at gmail.com
>>> <mailto:bacon4000 at gmail.com>>:
>>>
>>> >
>>> > Hi Jov,
>>> >
>>> > Are you aware that your py-tensorflow port is marked BROKEN?
>>> >
>>> > Thanks for porting this, BTW. We have researchers interested in
>>> it and we
>>> > use FreeBSD quite a bit here.
>>> >
>>> > Cheers,
>>> >
>>> > Jason
>>> >
>>> > --
>>> > Earth is a beta site.
>>> >
>>> >
>>> _______________________________________________
>>> freebsd-ports at freebsd.org <mailto:freebsd-ports at freebsd.org>
>>> mailing list
>>> https://lists.freebsd.org/mailman/listinfo/freebsd-ports
>>> <https://lists.freebsd.org/mailman/listinfo/freebsd-ports>
>>> To unsubscribe, send any mail to
>>> "freebsd-ports-unsubscribe at freebsd.org
>>> <mailto:freebsd-ports-unsubscribe at freebsd.org>"
>>>
>>>
>>>
>>>
>>> --
>>> Li-Wen Hsu <lwhsu at FreeBSD.org>
>>> https://lwhsu.org
>>>
>>
>>
>> --
>> Earth is a beta site.
>>
>>
>
--
Li-Wen Hsu <lwhsu at FreeBSD.org>
https://lwhsu.org
More information about the freebsd-ports
mailing list