[ruby-core:87853] Re: undefined symbol: mjit_init_p

Benoit, I read the bug report. I'm not skilled enough to understand all 
of it, but I agree that testing what you just built before you put it 
into your actual production filesystem is a good idea.

I did "make all install" and have the following after running "make check":

Finished tests in 485.657605s, 40.3268 tests/s, 4724.4972 assertions/s. 

19585 tests, 2294488 assertions, 5 failures, 51 errors, 57 skips

So now I get to figure out if any of those failures, errors, or skips 
affect my code. Or I can undo the tar file of /usr/local/bin and 
/usr/local/lib and be back where I was. Neither option seems great.

Leam


On 07/07/2018 01:09 PM, Benoit Daloze wrote:
> One more reason for https://bugs.ruby-lang.org/issues/13620 maybe? ;)
> The number of ugly hacks because people don't want to "make install'
> by default is huge IMHO :(
> 
> BTW, Ruby can be configured to install to a prefix writable by the
> user and not only root:
> ./configure --prefix=~/prefix/ruby-trunk
> for instance. Then there is no need to run anything as root.
> 
> On Sat, Jul 7, 2018 at 6:13 PM, Takashi Kokubun <[e-mail]> wrote:
>> As I told you, `make install` is needed to make Ruby work. Running
>> `make check` before `make install` is totally wrong with current
>> trunk. Don't run `make check` before `make install`.
>> 2018年7月8日(日) 0:57 Leam Hall <[e-mail]>:
>>>
>>> Thank you for a fast response.
>>>
>>> My process is to run 'make', and then 'make check' as a user and then
>>> run 'make install' as root. The error comes from the 'make check' log.
>>>
>>> What's the best way to test ruby once it is fully built and installed?
>>>
>>> Again, thank you.
>>>
>>> Leam
>>>
>>>
>>> On 07/07/2018 11:33 AM, Takashi Kokubun wrote:
>>>> It's not reproducible on my environment and any CI since mjit_init_p
>>>> does not exist anywhere in the current trunk, but my guess is that you
>>>> didn't execute `make install` and thus you're using an old MJIT
>>>> header. MJIT never works without `make install`.
>>>>
>>>> I always build ruby with `make all install` for that reason.
>>>> 2018年7月8日(日) 0:23 Leam Hall <[e-mail]>:
>>>>>
>>>>> I pulled Ruby trunk on 3 Jul and am now getting errors similar to the
>>>>> following:
>>>>>
>>>>> MJIT warning: failure in loading code from
>>>>> '/tmp/_ruby_mjit_p20781u0.so': /tmp/_ruby_mjit_p20781u0.so: undefined
>>>>> symbol: mjit_init_p
>>>>>
>>>>> The pull before that was 4 June and I didn't get the errors. It looks
>>>>> like this commit may have removed mjit_init_p:
>>>>>
>>>>> https://github.com/ruby/ruby/commit/06f54f03039c225c5c0b5d1709865fb4fff02776#diff-75b3ed34ab948aab8de703d073f34a98
>>>>>
>>>>> I'm not sure what's referencing the lack of the symbol.
>>>>>
>>>>> Platform:
>>>>>           CentOS 6, patched
>>>>>           GCC from rpm,  4.4.7 20120313 (Red Hat 4.4.7-23) (GCC)
>>>>>           Autoconf 2.69
>>>>>
>>>>> Thoughts?
>>>>>
>>>>> Leam
>>>>>
>>>>> Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
>>>>> <http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>
>>>>
>>>> Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
>>>> <http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>
>>>>
>>>
>>
>> Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
>> <http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>
> 
> Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
> <http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>
> 


Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>