Debian packages for OpenCog

[Masayuki Hatta]

Hi,

I'm trying out OpenCog now, and cooked up Debian packages.  Currently I have cogutils (libcogutil-dev), opencog-atomspace, opencog-moses.  The main opencog repo can be packaged, but not yet.

Files are available at https://people.debian.org/~mhatta/debian/opencog/

(APT repository is not set up yet)

Basically I made them for my own convenience,  and still need to be polished, but I think those are ok quality.  I'm a Debian developer, so I can upload them to the official Debian archive if you want.

Some changes are made to the source code to build with the current g++.  One thing is "Ambiguous overload call to abs()"  errors, and I changed abs to fabs(I'm not sure this solution is correct). 

Also I killed BackwardChainerUTest in atomspace for now.

If there are Debian users, please let me know your impression.

Best regards,
MH

[Ben Goertzel]

hey, very cool!

I will be in Tokyo meeting with some AGI researchers on Oct 9, maybe
you should join us ... do you know Naoya Arakawa?

ben

> --
> You received this message because you are subscribed to the Google Groups
> "opencog" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to opencog+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> Visit this group at https://groups.google.com/group/opencog.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/opencog/61ed7380-c150-4edc-93c9-c3747a8370ff%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.



--
Ben Goertzel, PhD
http://goertzel.org

"I am God! I am nothing, I'm play, I am freedom, I am life. I am the
boundary, I am the peak." -- Alexander Scriabin

[Masayuki Hatta]

Hi,

Seems interesting, but I should say I'm pretty much AGI newbie...I
heard about Dr. Arakawa and WBAI, but I've never met him in person.

Best regards,
MH

> You received this message because you are subscribed to a topic in the Google Groups "opencog" group.
> To unsubscribe from this topic, visit https://groups.google.com/d/topic/opencog/GfIIDgH0sww/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to opencog+u...@googlegroups.com.

> To post to this group, send email to ope...@googlegroups.com.
> Visit this group at https://groups.google.com/group/opencog.

> To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CACYTDBdhTPjXUD5PZCw4OT8Z-heFHBKLF9fTPL33pTQ8%2BEyUFA%40mail.gmail.com.

> For more options, visit https://groups.google.com/d/optout.



--

Masayuki Hatta
Associate Professor, Faculty of Economics and Management, Surugadai
University, Japan

http://about.me/mhatta

mha...@gnu.org / mha...@debian.org / mha...@opensource.jp /
hatta.m...@surugadai.ac.jp

[Ben Goertzel]

Well AGI is a pretty wide-open field and can use newbies with new
perspectives, for sure...

I noticed your interest in open innovation and IP law so forth -- it
is valuable to have people with this sort of background involved in an
OSS project for sure...!

ben

> To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAECZVPDp5RcfUq%2B%3DxzGcAqzJ-C5XjjLMXW%3DtYHpuDV9PnMOfoA%40mail.gmail.com.

[Linas Vepstas]

Cool!

On Tue, Aug 8, 2017 at 2:55 AM, Masayuki Hatta <mha...@gmail.com> wrote:

Hi,

I'm trying out OpenCog now, and cooked up Debian packages.  Currently I have cogutils (libcogutil-dev), opencog-atomspace, opencog-moses.  The main opencog repo can be packaged, but not yet.

Files are available at https://people.debian.org/~mhatta/debian/opencog/

(APT repository is not set up yet)

Basically I made them for my own convenience,  and still need to be polished, but I think those are ok quality.  I'm a Debian developer, so I can upload them to the official Debian archive if you want.

Wow. That would be .. not sure. Good and bad, mostly because we are not that stable, and do not use version numbers ...
 

Some changes are made to the source code to build with the current g++.  One thing is "Ambiguous overload call to abs()"  errors, and I changed abs to fabs(I'm not sure this solution is correct). 

Can you provide a pull req, or open a bug report? I recall seeing those, but they went away at some point (on debian unstable half a year ago) 

Also I killed BackwardChainerUTest in atomspace for now.

Hmm. That always passes for me on the current Debian stable, always.  I'm guessing now that maybe it is a guile-2.0 vs guile-2.2 bug.   I bet it will pass with guile-2.2

https://github.com/opencog/atomspace/issues/1318

In my experience, guile-2.2 is less buggy than guile-2.0  Its not perfect, its got some major flaws, but its better.

-- linas

If there are Debian users, please let me know your impression.

Best regards,
MH

--

You received this message because you are subscribed to the Google Groups "opencog" group.

To unsubscribe from this group and stop receiving emails from it, send an email to opencog+unsubscribe@googlegroups.com.

[Masayuki Hatta]

Hi, thanks for comment!

2017-08-10 5:02 GMT+09:00 Linas Vepstas <linasv...@gmail.com>:

>> Basically I made them for my own convenience, and still need to be
>> polished, but I think those are ok quality. I'm a Debian developer, so I
>> can upload them to the official Debian archive if you want.
>
> Wow. That would be .. not sure. Good and bad, mostly because we are not that
> stable, and do not use version numbers ...

Yes, that's what I meant "if you want". I can put them into
"experimental", but I guess most suitable way for now is setting up a
PPA(personal package archive). Then we can do "apt-get install
opencog". I can setup one at my own server, but if you want I can
help setting up it somewhere under opencog.org.

Building opencog & friends is quite complicated for newbies like me,
so setting up an official PPA would help them a lot.

BTW, many software in Debian these days doesn't have version numbers
or the concept of "release" (simply uses Git repo snapshot), so that's
ok.

>> Some changes are made to the source code to build with the current g++.
>> One thing is "Ambiguous overload call to abs()" errors, and I changed abs
>> to fabs(I'm not sure this solution is correct).
>
> Can you provide a pull req, or open a bug report? I recall seeing those, but
> they went away at some point (on debian unstable half a year ago)

Done.

https://github.com/opencog/moses/pull/49
https://github.com/opencog/cogutils/pull/80

I'm not sure these fixes are appropriate, so please review carefully.

>> Also I killed BackwardChainerUTest in atomspace for now.
>
> Hmm. That always passes for me on the current Debian stable, always. I'm
> guessing now that maybe it is a guile-2.0 vs guile-2.2 bug. I bet it will
> pass with guile-2.2
>
> https://github.com/opencog/atomspace/issues/1318

Yes, as I said in that thread, I'm still using guile 2.0.13. Now
guile-2.2 packages are entering into Debian unstable (still in NEW
queue), so I will try it later.

Best regards,
MH

[Masayuki Hatta]

Hi,

2017-08-09 14:41 GMT+09:00 Ben Goertzel <b...@goertzel.org>:
> Well AGI is a pretty wide-open field and can use newbies with new
> perspectives, for sure...
>
> I noticed your interest in open innovation and IP law so forth -- it
> is valuable to have people with this sort of background involved in an
> OSS project for sure...!

Well, then if you don't mind, please let me know the detail of Oct. 9
event. I couldn't find any information...lecture or workshop?

Best regards,
MH

[Dagim Sisay]

This is awesome. As Linas said, opencog isn't stable enough to have a 'release' but I have been working on opencog for RPI3 on raspbian and I have a script generated debian package to make installation on the PI easy enough, because compiling the entire thing on a PI is a full day task. The deb is at 144.76.153.5:/opencog/opencog-dev_1.0-1_armhf.deb
I have zero experience with debian packaging and so it's the most minimal package. Plus base dependencies aren't installed with this, that's done by the script. Would be great if u have pointers

[AmeBel]

There is https://launchpad.net/~opencog-dev/+archive/ubuntu/ppa. Don't know who have access to it.

[Linas Vepstas]

On Wed, Aug 9, 2017 at 7:56 PM, Masayuki Hatta <mha...@gmail.com> wrote:

Hi, thanks for comment!

2017-08-10 5:02 GMT+09:00 Linas Vepstas <linasv...@gmail.com>:

>> Basically I made them for my own convenience,  and still need to be
>> polished, but I think those are ok quality.  I'm a Debian developer, so I
>> can upload them to the official Debian archive if you want.
>
> Wow. That would be .. not sure. Good and bad, mostly because we are not that
> stable, and do not use version numbers ...

Yes, that's what I meant "if you want". 

OK, yes, I think I want that. The positives outweigh the negatives.
 

I can put them into
"experimental", but I guess most suitable way for now is setting up a
PPA(personal package archive).  Then we can do "apt-get install
opencog".  I can setup one at my own server, but if you want I can
help setting up it somewhere under opencog.org.

Whatever is easiest...

Building opencog & friends is quite complicated for newbies like me,
so setting up an official PPA would help them a lot.

Yes, that is one of the big positives to having a package.

BTW, many software in Debian these days doesn't have version numbers
or the concept of "release" (simply uses Git repo snapshot), so that's
ok.

OK. We are constantly fixing bugs or adding features, so if someone complains
"this doesn't work" , our standard answer is "use the latest".

https://github.com/opencog/moses/pull/49
https://github.com/opencog/cogutils/pull/80

I'm not sure these fixes are appropriate, so please review carefully.

Nil merged them. They looked OK to me.
 

>> Also I killed BackwardChainerUTest in atomspace for now.
>
> Hmm. That always passes for me on the current Debian stable, always.  I'm
> guessing now that maybe it is a guile-2.0 vs guile-2.2 bug.   I bet it will
> pass with guile-2.2
>
> https://github.com/opencog/atomspace/issues/1318

Yes, as I said in that thread, I'm still using guile 2.0.13.  Now
guile-2.2 packages are entering into Debian unstable (still in NEW
queue), so I will try it later.

OK. Tell the Debian folks that 2.2 really is much better than 2.0. I've been using
it for 2+ years now, from when it was 2.1.x and it was consistently better in just

about every way. (I can also tell you where its broken, but that's a different matter).

--linas

[AmeBel]

Hi,

 

BTW, many software in Debian these days doesn't have version numbers
or the concept of "release" (simply uses Git repo snapshot), so that's
ok.

If there are multiple builds in a day then how will the git-sha help denote the ordered relationship, that is without having access to the git repo.

[Masayuki Hatta]

Hi,

2017年8月16日水曜日 20時46分00秒 UTC+9 AmeBel:

BTW, many software in Debian these days doesn't have version numbers
or the concept of "release" (simply uses Git repo snapshot), so that's
ok.

If there are multiple builds in a day then how will the git-sha help denote the ordered relationship, that is without having access to the git repo.

 
Yes, the package version number itself doesn't guarantee the order.  There may be, say, 1.0~git20170817.aaaaaaa and 1.0~git20170817.bbbbbbb, and it's quite possible that the former is actually the newer one.  The package maintainer should be careful for that.

(or, honestly, I'm not sure what your question means.  Could you elaborate a bit?)

Best regards,
MH

[Masayuki Hatta]

Hi,

2017年8月10日木曜日 14時56分41秒 UTC+9 Dagi S:

This is awesome. As Linas said, opencog isn't stable enough to have a 'release' but I have been working on opencog for RPI3 on raspbian and I have a script generated debian package to make installation on the PI easy enough, because compiling the entire thing on a PI is a full day task. The deb is at 144.76.153.5:/opencog/opencog-dev_1.0-1_armhf.deb
I have zero experience with debian packaging and so it's the most minimal package. Plus base dependencies aren't installed with this, that's done by the script. Would be great if u have pointers

I'll be very glad if you can make my packages rebuild for Rasbian (I have no knowledge about Rasberry Pi, sorry).  I set build dependencies carefully, so it should not be so difficult to rebuild packages on RPI3, or even cross-compiling.  Please let me know if you have any difficulties.

Best regards,
MH