Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Andreas Tille]

Hi Wolfgang,

On Mon, Feb 13, 2017 at 07:55:49PM +0100, Wolfgang Fütterer wrote:
> thank you for your reminder. I finally made it to add calculix-cgx to the 
> appropriate task. Unfortunately I get the folloeing error trying to push my 
> changes to the repository:
> 
> remote: error: insufficient permission for adding an object to repository 
> database ./objects
> remote: fatal: failed to write object
> error: unpack failed: unpack-objects abnormal exit
> To git+ssh://git.debian.org/git/blends/projects/science.git
>  ! [remote rejected] master -> master (unpacker error)

Well, DDs and members of Blends team have commit permissions. :-)

I pushed your changes and added you to the team to enable you pushing
yourself next time.
 
> Thank you for your efforts.

You are welcome

 Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Wolfgang Fütterer]

Hi Andreas, 

thank you for your reminder. I finally made it to add calculix-cgx to the 
appropriate task. Unfortunately I get the folloeing error trying to push my 
changes to the repository:

remote: error: insufficient permission for adding an object to repository 
database ./objects
remote: fatal: failed to write object
error: unpack failed: unpack-objects abnormal exit
To git+ssh://git.debian.org/git/blends/projects/science.git
 ! [remote rejected] master -> master (unpacker error)

So I attached the patches to this mail.

The first patch adds calculix-cgx to the appropriate tasks. 
The second patch removes calculix-ccx-test and calculix-ccx-doc from the 
engineering-dev task as the are no development packages. In the packages are 
mainly documentations for how to use the software. I added them as a 
suggestion to the tasks in which calculix-ccx already is.

Thank you for your efforts.

Cheers
Wolfgang

-- 
Jabber: yagha...@jabber.de
IRC #debian-science: wlfuetter>From d8ab618b6a61d8554ccf86818d2a94a519b3c418 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Wolfgang=20F=C3=BCtterer?= 
Date: Mon, 13 Feb 2017 19:16:30 +0100
Subject: [PATCH 1/2] Added calculix-cgx and calculix-cgx-examples

---
 tasks/engineering  | 4 
 tasks/numericalcomputation | 4 
 tasks/simulations  | 4 
 3 files changed, 12 insertions(+)

diff --git a/tasks/engineering b/tasks/engineering
index 78a675c..2353abe 100644
--- a/tasks/engineering
+++ b/tasks/engineering
@@ -144,6 +144,10 @@ Depends: libadios-bin
 
 Depends: calculix-ccx
 
+Depends: calculix-cgx
+
+Suggests: calculix-cgx-examples
+
 Depends: scram
 WNPP: 842766
 
diff --git a/tasks/numericalcomputation b/tasks/numericalcomputation
index 5446c50..08bba3e 100644
--- a/tasks/numericalcomputation
+++ b/tasks/numericalcomputation
@@ -124,6 +124,10 @@ Depends: iep
 
 Depends: calculix-ccx
 
+Depends: calculix-cgx
+
+Suggests: calculix-cgx-examples
+
 Depends: python-sfepy
 
 Depends: python3-bitstring | python-bitstring
diff --git a/tasks/simulations b/tasks/simulations
index d05aad9..379adb2 100644
--- a/tasks/simulations
+++ b/tasks/simulations
@@ -23,6 +23,10 @@ Depends: music-bin
 
 Depends: calculix-ccx
 
+Depends: calculix-cgx
+
+Suggests: calculix-cgx-examples
+
 Depends: python-sfepy
 
 Suggests: libopensurgsim-dev
-- 
2.11.0

>From ef6e0cd82bcae6a69dacd78ecbb5e54442ebc88d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Wolfgang=20F=C3=BCtterer?= 
Date: Mon, 13 Feb 2017 19:21:36 +0100
Subject: [PATCH 2/2] Removed calculix-ccx-doc calculix-ccx-test from
 engineering-dev and added them as a suggestion to engineering,
 numericalcomputation and simulations

---
 tasks/engineering  | 2 ++
 tasks/engineering-dev  | 2 --
 tasks/numericalcomputation | 2 ++
 tasks/simulations  | 2 ++
 4 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/tasks/engineering b/tasks/engineering
index 2353abe..1e903d9 100644
--- a/tasks/engineering
+++ b/tasks/engineering
@@ -144,6 +144,8 @@ Depends: libadios-bin
 
 Depends: calculix-ccx
 
+Suggests: calculix-ccx-doc, calculix-ccx-test
+
 Depends: calculix-cgx
 
 Suggests: calculix-cgx-examples
diff --git a/tasks/engineering-dev b/tasks/engineering-dev
index 1b1bf2d..c01c169 100644
--- a/tasks/engineering-dev
+++ b/tasks/engineering-dev
@@ -93,8 +93,6 @@ Depends: python3-pynfft | python-pynfft
 
 Depends: libplb-dev
 
-Suggests: calculix-ccx-doc, calculix-ccx-test
-
 Depends: libceres-dev
 
 Suggests: ceres-solver-doc
diff --git a/tasks/numericalcomputation b/tasks/numericalcomputation
index 08bba3e..7fa0dce 100644
--- a/tasks/numericalcomputation
+++ b/tasks/numericalcomputation
@@ -124,6 +124,8 @@ Depends: iep
 
 Depends: calculix-ccx
 
+Suggests: calculix-ccx-doc, calculix-ccx-test
+
 Depends: calculix-cgx
 
 Suggests: calculix-cgx-examples
diff --git a/tasks/simulations b/tasks/simulations
index 379adb2..6899fc3 100644
--- a/tasks/simulations
+++ b/tasks/simulations
@@ -23,6 +23,8 @@ Depends: music-bin
 
 Depends: calculix-ccx
 
+Suggests: calculix-ccx-doc, calculix-ccx-test
+
 Depends: calculix-cgx
 
 Suggests: calculix-cgx-examples
-- 
2.11.0

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Andreas Tille]

Hi Thomas,

On Fri, Feb 10, 2017 at 06:41:22PM +0100, Thomas Schiex wrote:
> > To add a binary (!) package to a task you can simply
> > debcheckout -u your_alioth_login debian-science
> > cd debian-science/tasks
> > and edit the task in question.
> and debcheckout is in devscripts. I just did not realize I could do
> everything through git alone.

I admit I was just to lazy to seek for the Git URL since I think the
above can be memorised more easily. :-)
 
> Anyway, it looks like you did fine from my suggestions.
> 
> I will rely on git to deal with extra details (eg. do not understand why
> coinor-cbc is in logic ONLY, but coinor-symphony is is
> mathematics/numerical computation but not in logic since both are mixed
> integer linear programming solvers).
> 
> I'm also wondering if evolver, a very specialized solver, would not be
> better as a "Suggests" in all cases.

OK, I turnes this into another change set:


diff --git a/tasks/logic b/tasks/logic
index 06e5416..bba78a2 100644
--- a/tasks/logic
+++ b/tasks/logic
@@ -53,7 +53,7 @@ Depends: spass
 X-End-Category: first-order and equational solvers
 
 X-Begin-Category: integer programming solvers
-Depends: coinor-cbc
+Depends: coinor-cbc, coinor-symphony
 
 Suggests: coinor-libcoinmp-dev
 
diff --git a/tasks/mathematics b/tasks/mathematics
index 8d105eb..37d5912 100644
--- a/tasks/mathematics
+++ b/tasks/mathematics
@@ -193,7 +193,7 @@ Remark: Removed from Debian
 
 Depends: python-minieigen
 
-Depends: coinor-symphony
+Depends: coinor-symphony, coinor-cbc
 
 Depends: tetgen
 
@@ -210,6 +210,6 @@ Suggests: gmp-ecm
 
 Suggests: primesieve
 
-Depends: evolver-ogl | evolver-nox
+Suggests: evolver-ogl | evolver-nox
 
 Depends: toulbar2
diff --git a/tasks/numericalcomputation b/tasks/numericalcomputation
index 87b7a16..5446c50 100644
--- a/tasks/numericalcomputation
+++ b/tasks/numericalcomputation
@@ -100,7 +100,7 @@ Suggests: matlab-support
 
 Suggests: libsundials-serial-dev, octave-sundials
 
-Depends: coinor-symphony
+Depends: coinor-symphony, coinor-cbc
 
 Depends: coinor-clp
 
@@ -173,6 +173,6 @@ Suggests: theano-doc
 
 Depends: python3-sparse | python-sparse
 
-Depends: evolver-ogl | evolver-nox
+Suggests: evolver-ogl | evolver-nox
 
 Depends: toulbar2
diff --git a/tasks/physics b/tasks/physics
index 67c27bb..53fccab 100644
--- a/tasks/physics
+++ b/tasks/physics
@@ -203,6 +203,6 @@ Pkg-URL: http://people.debian.org/~tille/packages/fdmnes/
 
 Depends: pyfr
 
-Depends: evolver-ogl | evolver-nox
+Suggests: evolver-ogl | evolver-nox
 
 Depends: toulbar2

 
> As always, thanks for your help.

You are welcome - thanks as well for your help.

Kind regards

   Andreas. 

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Thomas Schiex]

Andreas,

>> I'm sorry,[..] You just need Git to download
>> the repository and the repository has 5MB.  :-P
>> The repositories are on alioth.  You can also
>> perfectly login to Alioth and clone + commit + push. 
Sorry, I'm showing off my mediocre knowledge of debian(science)
organization. I just tried to follow the indication in your mail:
> To add a binary (!) package to a task you can simply
> debcheckout -u your_alioth_login debian-science
> cd debian-science/tasks
> and edit the task in question.
and debcheckout is in devscripts. I just did not realize I could do
everything through git alone.

Anyway, it looks like you did fine from my suggestions.

I will rely on git to deal with extra details (eg. do not understand why
coinor-cbc is in logic ONLY, but coinor-symphony is is
mathematics/numerical computation but not in logic since both are mixed
integer linear programming solvers).

I'm also wondering if evolver, a very specialized solver, would not be
better as a "Suggests" in all cases.

As always, thanks for your help.

Thomas

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Andreas Tille]

Dear Thomas,

thanks for your attempt to help.  Hope I can clarify how you can do this
more easy.

On Fri, Feb 10, 2017 at 12:58:50PM +0100, Thomas Schiex wrote:
> I would have been happy to do the job directly but I currently do not
> have access to a running debian machine I can "control" (have devscripts
> installed) and I lack time/HD space for a VM install.

I'm sorry, but this is no valid excuse.  You just need Git to download
the repository and the repository has 5MB.  :-P

> It perhaps could
> make sense to have it on alioth? (no troll intended - if it can be
> categorized as such)

What do you mean here?  The repositories are on alioth.  You can also
perfectly login to Alioth and clone + commit + push.
 
> So my suggestions,  on the packages I know/use,  if this can be useful:
> 
> coinmplogic, mathematics, numerical computation

Please not that always *binary* packages (second column of the table)
are mentioned in tasks files.  So I guess you mean coinor-libcoinmp-dev
which I added to mathematics-dev and added Suggests in logic and
numericalcomputation (where we do not yet have development tasks -
probably for the only reason because nobody cared about this so far)
 
> It's a C-API to COIN, a linear/ILP optimization solver (with branch and cut, 
> cut generation...). 
> coinbc, coinsymphony, coincg are there already. It would be logical to put 
> coinmp there too.

I admit I have not found any of the mentioned packages (coinbc,
coinsymphony, coincg) in the Debian package pool.  Please note that I'm
not an expert on this field and never used COIN.  So in case this
might be important I'd need more precise information.
 
> collada-dom   viewing  (given that it is in debian-science)

You mean here the binary packatge libcollada-dom2.4-dp-dev and whether
it is given in Debian Science you can check either in Git[1] or as I
recommended in my mail on the tasks page[2].
 
> collada is a format/initiative for 3D models exchange. I see collada-dom is a 
> "Document Object Model/XML" version. Would be more a "graphics' package 
> for me than a science one (personal opinion, arguable, YMMV).

If something is "arguable" then I tend to set only "Suggests".
 
> evolver   mathematics, physics, numerical computation
> 
> Evolver is a minimal surface optimization tool (variational optimization), 
> math/numerical computation because it is a numerical variational 
> optimization, 
> physics because it allows to model eg. liquid/solid interface.

I've added

  Depends: evolver-ogl | evolver-nox

to all three tasks.

> toulbar2machine learning, physics, logic, mathematics, numerical 
> computation
> 
> optimization on discrete graphical models. Machine learning/Physics because 
> it solves 
> "maximum a posteriori" on Markov Random Fields used in ML and statistical 
> physics. Logic,
> mathematic, numerical computation because it is also a general discrete 
> optimization tool 
> that can solve both logical or numerical NP-complete decision problems.

All in all I did the following changes:


diff --git a/tasks/logic b/tasks/logic
index dd268b2..06e5416 100644
--- a/tasks/logic
+++ b/tasks/logic
@@ -55,6 +55,8 @@ X-End-Category: first-order and equational solvers
 X-Begin-Category: integer programming solvers
 Depends: coinor-cbc
 
+Suggests: coinor-libcoinmp-dev
+
 Depends: minisat+
 X-Comment: minisat2 is only a transitional package
 X-End-Category: integer programming solvers
@@ -74,3 +76,5 @@ Depends: maria
 
 Depends: mona
 X-End-Category: reasoning about automata
+
+Depends: toulbar2
diff --git a/tasks/machine-learning b/tasks/machine-learning
index 754155b..108cd6f 100644
--- a/tasks/machine-learning
+++ b/tasks/machine-learning
@@ -151,3 +151,5 @@ Depends: caffe-cpu | caffe-cuda
 Depends: torch-core-free
 
 Depends: libmlpack-dev
+
+Depends: toulbar2
diff --git a/tasks/mathematics b/tasks/mathematics
index da3be13..8d105eb 100644
--- a/tasks/mathematics
+++ b/tasks/mathematics
@@ -209,3 +209,7 @@ Depends: auto-07p
 Suggests: gmp-ecm
 
 Suggests: primesieve
+
+Depends: evolver-ogl | evolver-nox
+
+Depends: toulbar2
diff --git a/tasks/mathematics-dev b/tasks/mathematics-dev
index b827cd0..58ca854 100644
--- a/tasks/mathematics-dev
+++ b/tasks/mathematics-dev
@@ -246,3 +246,5 @@ Depends: trilinos-all-dev
 Depends: libdeal.ii-dev
 
 Depends: libp4est-dev
+
+Depends: coinor-libcoinmp-dev
diff --git a/tasks/numericalcomputation b/tasks/numericalcomputation
index 2ae005a..87b7a16 100644
--- a/tasks/numericalcomputation
+++ b/tasks/numericalcomputation
@@ -104,6 +104,8 @@ Depends: coinor-symphony
 
 Depends: coinor-clp
 
+Suggests: coinor-libcoinmp-dev
+
 Suggests: libquantum-dev
 
 Depends: metis-edf
@@ -170,3 +172,7 @@ Depends: python3-theano | python-theano
 Suggests: theano-doc
 
 Depends: python3-sparse | python-sparse
+
+Depends: evolver-ogl | evolver-nox
+
+Depends: toulbar2
diff --git a/tasks/physics b/tasks/physics
index 6c11c6d..67c27bb 100644
--- a/tasks/physics
+++ 

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Thomas Schiex]

Dear Andreas,

I would have been happy to do the job directly but I currently do not
have access to a running debian machine I can "control" (have devscripts
installed) and I lack time/HD space for a VM install. It perhaps could
make sense to have it on alioth? (no troll intended - if it can be
categorized as such)

So my suggestions,  on the packages I know/use,  if this can be useful:

coinmp  logic, mathematics, numerical computation

It's a C-API to COIN, a linear/ILP optimization solver (with branch and cut, 
cut generation...). 
coinbc, coinsymphony, coincg are there already. It would be logical to put 
coinmp there too.

collada-dom viewing  (given that it is in debian-science)

collada is a format/initiative for 3D models exchange. I see collada-dom is a 
"Document Object Model/XML" version. Would be more a "graphics' package 
for me than a science one (personal opinion, arguable, YMMV).

evolver mathematics, physics, numerical computation

Evolver is a minimal surface optimization tool (variational optimization), 
math/numerical computation because it is a numerical variational optimization, 
physics because it allows to model eg. liquid/solid interface.

toulbar2machine learning, physics, logic, mathematics, numerical 
computation

optimization on discrete graphical models. Machine learning/Physics because it 
solves 
"maximum a posteriori" on Markov Random Fields used in ML and statistical 
physics. Logic,
mathematic, numerical computation because it is also a general discrete 
optimization tool 
that can solve both logical or numerical NP-complete decision problems.

Kind regards,
Thomas

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Jochen Sprickerhof]

* Andreas Tille  [2017-02-09 16:29]:
> Please categorise *any* of your packages - be it in testing or not.  The
> metapackage creation process will verify whether a package is in testing
> and add Recommends *only* if a package is in testing.  If not the
> package gets a Suggests - which IMHO is very sensible since it might
> point the user to install from unstable if wanted or may be there might
> be a backport.

That makes sense, thanks! I still have the Robotics Blend in mind, you
proposed last time, but I think the dependencies is ok for now.

Cheers Jochen


signature.asc
Description: PGP signature

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Andreas Tille]

Hi Jochen,

On Thu, Feb 09, 2017 at 03:48:20PM +0100, Jochen Sprickerhof wrote:
> 
> I see ros-opencv-apps in the list, which is not in testing (and will not
> make it, due to the freeze). So I assume it shouldn't be categorised as
> well.

Please categorise *any* of your packages - be it in testing or not.  The
metapackage creation process will verify whether a package is in testing
and add Recommends *only* if a package is in testing.  If not the
package gets a Suggests - which IMHO is very sensible since it might
point the user to install from unstable if wanted or may be there might
be a backport.

Moreover the package gets listed on the according tasks page as a
valuable information.  Apropos tasks pages:  The right moment to add a
binary package to the tasks files is once you have created a record in
Vcs (Git or SVN).  This is parsed nightly and the tasks pages will show
a record of your work in the "Packaging has started and developers might
try the packaging code in VCS" section.  So users will be informed that
some work is going on (and might help you).

So I pushed

$ git diff HEAD^
diff --git a/tasks/robotics b/tasks/robotics
index b85b7c0..c01eb4c 100644
--- a/tasks/robotics
+++ b/tasks/robotics
@@ -137,6 +137,8 @@ Depends: ros-desktop-full,
  ros-perception,
  ros-simulators
 
+Depends: ros-opencv-apps
+
 Depends: openrtm
 Homepage: http://www.openrtm.org/
 Responsible: Thomas Moulard 


Please correct it if you do not agree.

> Regarding ros-ros, I think it would make sense to blacklist it, as it is
> only the base of a lot of ROS components already categorised.

Blacklisted.

Thanks for your input

   Andreas.


-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages (second iteration)

[Jochen Sprickerhof]

Hi Andreas,

thank you for your continuous effort!

I see ros-opencv-apps in the list, which is not in testing (and will not
make it, due to the freeze). So I assume it shouldn't be categorised as
well.

Regarding ros-ros, I think it would make sense to blacklist it, as it is
only the base of a lot of ROS components already categorised.

Cheers Jochen


signature.asc
Description: PGP signature

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

On Sat, Jan 21, 2017 at 09:30:41PM +0100, Daniel Stender wrote:
> ... Theano would be belong into numerical computation.
> 
> ---
>  tasks/numericalcomputation | 4 
>  1 file changed, 4 insertions(+)
> 
> diff --git a/tasks/numericalcomputation b/tasks/numericalcomputation
> index cb0a7bb..b0923f5 100644
> --- a/tasks/numericalcomputation
> +++ b/tasks/numericalcomputation
> @@ -164,3 +164,7 @@ Depends: libppl-dev
>  
>  Depends: python-escript | python3-escript | python-escript-mpi | 
> python3-escript-mpi
>  Why: programming tool for implementing mathematical models in Python using 
> the finite element method (FEM)
> +
> +Depends: python3-theano | python-theano
> +
> +Suggests: theano-doc

Done.  Thanks for the hint

Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Daniel Stender]

On 04.01.2017 15:50, Andreas Tille wrote:
> Hi,
> 
> as in every release cycle I'm trying to verify that every package
> maintained in Debian Science team is properly categorised in our Blends
> tasks.  If a package is just a predependency for some other scientific
> software I can add it to a blacklist of packages that should not show
> up in the Debian Science metapackages explicitly.  I have uploaded the
> list of not yet categorised (not yet blacklisted) source packages to
> 
>http://blends.debian.net/tmp/debian-science.txt
> 
> The list also contains the latest uploader - so simply seek for your
> name - everybody in CC is mentioned at least once in the list and should
> definitely have a look.  If other readers here feel competent to
> classify a package for one or more (!) tasks in our task list
> 
>https://blends.debian.org/science/tasks/
> 
> evary suggestion is welcome.
> 
> To add a binary (!) package to a task you can simply
> 
> debcheckout -u your_alioth_login debian-science
> cd debian-science/tasks
> 
> and edit the task in question.  Members of the Debian Pure Blends team
> as well as any DD (ACLs are set but I have heard this does not work
> reliably) have commit permissions.  I'm also fine if you debcheckout
> anonymously and send me `git format-patch` formated changes or just
> answer here to this mail.
> 
> If you are not sure whether a package belongs to a task or not feel
> free to discuss this here.
> 
> Kind regards and thanks for your cooperation
> 
>   Andreas.

... Theano would be belong into numerical computation.

Best,
DS

-- 
4096R/DF5182C8
Debian Developer (sten...@debian.org)
LPIC-1 (LPI000329859 64mz6f7kt4)
http://www.danielstender.com/
>From d864bf94d79aeb57d8fc61264fbfe9e185b01762 Mon Sep 17 00:00:00 2001
From: Daniel Stender 
Date: Sat, 21 Jan 2017 21:26:42 +0100
Subject: [PATCH] add theano to numericalcomputation

---
 tasks/numericalcomputation | 4 
 1 file changed, 4 insertions(+)

diff --git a/tasks/numericalcomputation b/tasks/numericalcomputation
index cb0a7bb..b0923f5 100644
--- a/tasks/numericalcomputation
+++ b/tasks/numericalcomputation
@@ -164,3 +164,7 @@ Depends: libppl-dev
 
 Depends: python-escript | python3-escript | python-escript-mpi | python3-escript-mpi
 Why: programming tool for implementing mathematical models in Python using the finite element method (FEM)
+
+Depends: python3-theano | python-theano
+
+Suggests: theano-doc
-- 
2.11.0

Re: Please categorise your packages for the Debian Science metapackages

[Sébastien Villemot]

Le mardi 10 janvier 2017 à 12:05 +0100, Andreas Tille a écrit :
> 
> On Tue, Jan 10, 2017 at 11:29:14AM +0100, Sébastien Villemot wrote:

> > If it is the former, it makes senses to remove r-base (+ octave and
> > julia). But if it is the latter (and it is what I was implicitly
> > assuming), then we should leave those packages.
> 
> I'm perfectly fine if you want r-base back if its really intended.

Thanks. I reverted your change.

Cheers,

-- 
 .''`.Sébastien Villemot
: :' :Debian Developer
`. `' http://sebastien.villemot.name
  `-  GPG Key: 4096R/381A7594



signature.asc
Description: This is a digitally signed message part

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi Sébastien,

On Tue, Jan 10, 2017 at 11:29:14AM +0100, Sébastien Villemot wrote:
> > BTW, I removed r-base from the economics task since it seemed to
> > unspecific for me and it will be pulled in by the other r-* packages
> > anyway.  I hope you agree with this - if not feel free to revert.
> 
> Actually the task also contains octave and julia, which are also
> generic-purpose computing environments. So I think we should either
> remove them, or put back r-base, for consistency.

I'm fine with whatever you decide - but please stick to open discussion
on Debian Science list (in CC).
 
> This leads me to wonder: what exactly should be in the tasks? The
> packages that are specific to a given field? or the packages that may be
> useful for a user working a given field, and that are not installed by
> default?

The latter (=may be useful).  In the case of r-base I simply decided
because there are specific packages that are pulling in r-base anyway
(=installed by default installation of the sciene-economics).
 
> If it is the former, it makes senses to remove r-base (+ octave and
> julia). But if it is the latter (and it is what I was implicitly
> assuming), then we should leave those packages.

I'm perfectly fine if you want r-base back if its really intended.

Kind regards

   Andreas. 

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi Tino,

On Fri, Jan 06, 2017 at 12:17:38PM +0100, Tino Didriksen wrote:
> 
> giella-sme and everything built from hfst, hfst-ospell, lttoolbox fit in
> linguistics, given that there isn't a linguistics-dev task.

Since we do not have a linguistics-dev task I have done

 $ git diff HEAD^
diff --git a/tasks/linguistics b/tasks/linguistics
index 197665f..d423f28 100644
--- a/tasks/linguistics
+++ b/tasks/linguistics
@@ -130,3 +130,6 @@ Depends: python3-pynlpl | python-pynlpl
 
 Depends: uctodata
 
+Depends: giella-sme
+
+Depends: hfst, hfst-ospell, lttoolbox


It might make sense to decide whether to split up linguistics and
linguistics-dev.
 
> giella-core and giella-sme-dev should just be hidden, as they're primarily
> for build-depends.

OK, it is excluded now.

Kind regards

   Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Tino Didriksen]

On 4 January 2017 at 15:50, Andreas Tille  wrote:

> as in every release cycle I'm trying to verify that every package
> maintained in Debian Science team is properly categorised in our Blends
> tasks.  If a package is just a predependency for some other scientific
> software I can add it to a blacklist of packages that should not show
> up in the Debian Science metapackages explicitly.



giella-sme and everything built from hfst, hfst-ospell, lttoolbox fit in
linguistics, given that there isn't a linguistics-dev task.

giella-core and giella-sme-dev should just be hidden, as they're primarily
for build-depends.

-- Tino Didriksen

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Drew Parsons]

On Thu, 2017-01-05 at 21:36 +0100, Andreas Tille wrote:

> I'd like you something else to consider:  A user *application* should
> not contain the programming language it was written in its package
> name.
> I have not checked but dolphin-bin sounds way more like a user
> application than python-dolphin.

dolfin-bin is just a handful of little utility scripts.  Not
necessarily used in day-to-day use of dolfin.

Drew

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Andreas Tille]

Hi again,

On Fri, Jan 06, 2017 at 01:40:52AM +0800, Drew Parsons wrote:
> 
> What I mean is that there are a number of component packages which
> together make up FENiCS:
> dolfin-bin
> python-dolfin
> python-ffc
> python-ufl
> python-dijitso
> python-instant
> python-mshr
> 
> It doesn't really make sense to list these packages as separate entries
> in the tasks. python-dolfin ultimately is the end-user package, but we
> have a metapackage, fenics, to keep track of the whole collection.
> 
> These packages (in particular python-dolfin) are ready to be used by an
> end user, so the mathematics task is the right one.
> 
> So my recommendation is that the mathematics task should just list the
> one package, fenics.
> 
> Conceivably some of them could also be used by mathematics developers,
> e.g. the developers of Firedrake use FENiCS's UFL (python-ufl) [1]. So
> it's reasonable to also list the fenics package under mathematics-dev.

I admit I remain unsure what actually should be done.  It might make
sense to give users hints on the FENiCS components even if they are only
parts of a suite.  Its not an exclusion criterion for a package from a
metapackage that it is in the list of dependencies of another package
inside the same metapackage.  Could you please provide a patch for
mathematics and mathematics-dev tasks files to make sure it will be
exactly as you want it to be?

I'd like you something else to consider:  A user *application* should
not contain the programming language it was written in its package name.
I have not checked but dolphin-bin sounds way more like a user
application than python-dolphin.

Kind regards

   Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Drew Parsons]

On Thu, 2017-01-05 at 14:01 +0100, Andreas Tille wrote:
> Hi Drew,
> 
> I'm not sure whether I fully understand your suggestion.

What I mean is that there are a number of component packages which
together make up FENiCS:
dolfin-bin
python-dolfin
python-ffc
python-ufl
python-dijitso
python-instant
python-mshr

It doesn't really make sense to list these packages as separate entries
in the tasks. python-dolfin ultimately is the end-user package, but we
have a metapackage, fenics, to keep track of the whole collection.

These packages (in particular python-dolfin) are ready to be used by an
end user, so the mathematics task is the right one.

So my recommendation is that the mathematics task should just list the
one package, fenics.

Conceivably some of them could also be used by mathematics developers,
e.g. the developers of Firedrake use FENiCS's UFL (python-ufl) [1]. So
it's reasonable to also list the fenics package under mathematics-dev.

Cheers,
Drew

[1] http://www.firedrakeproject.org/




> On Thu, Jan 05, 2017 at 01:26:46PM +0800, Drew Parsons wrote:
> > Hi Andreas, there are a couple more fenics components in the 
> > mathematics-dev task:
> > 
> >   python-dolfin (and dolfin-dev)
> >   python-ffc
> >   python-ufl
> > 
> > I'm thinking it's redundant to list these separately under 
> > mathematics-dev if there's already an entry in the mathematics
> > task.
> 
> Do you want these three packages to be under mathematics-dev or
> mathematics in other words are these more developer oriented or user
> oriented packages.  In principle it is perfectly fine to list
> packages
> in more than one task if this makes sense.
>  
> > Likewise in your not-yet categorised list:
> >   dijitso
> > 
> > python-dijitso is another of the FENiCs components that I think
> > doesn't
> > need to be listed separately.
> 
> $ git diff HEAD^
> diff --git a/misc/sql/0-uncategorised.sh b/misc/sql/0-
> uncategorised.sh
> index c97012c..be73f1a 100755
> --- a/misc/sql/0-uncategorised.sh
> +++ b/misc/sql/0-uncategorised.sh
> @@ -10,7 +10,7 @@ case "$1" in
>  debian-science)
>  team="'debian-science-maintain...@lists.alioth.debian.org','
> pkg-scicomp-de...@lists.alioth.debian.org'"
>  ignore="'cbp2make','coinor-cgl','coinor-osi','coinutils',
> -'dh-r',
> +'dh-r','dijitso',
>  'dune-common', 'dune-geometry', -- Dependencies of
> dune-grid
>  'gmp',  -- very basic, used
> in lots of other packages as dependency
>  'imview-doc', 'instant',
> 
>  
> > How is "blacklisting" done? Is it just a matter of removing the
> > entries
> > in the files in debian-science/tasks ?
> 
> Its a bit hidden "tool" - or rather a quickly hacked script available
> here:
> 
> https://anonscm.debian.org/cgit/blends/website.git/tree/misc/sql/
> 0-uncategorised.sh
> 
> To make a proper tool the excluded source packages should be either
> in a
> config file or a database table - but for the moment it worked for me
> and so far nobody else showed interest to run it. :-)
> 
> Thanks for the contribution
> 
>   Andreas.
> 

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Johannes Ring]

On Thu, Jan 5, 2017 at 6:23 PM, Drew Parsons  wrote:
> On Thu, 2017-01-05 at 18:11 +0100, Johannes Ring wrote:
>> On Thu, Jan 5, 2017 at 6:03 PM, Drew Parsons 
>> wrote:
>> >
>> > No worries, thanks Johannes. We'll keep the fenics metapackage.
>> >
>> > In regards to Andreas' task package, is there any sense putting in
>> > two
>> > references, one for dolfin and one for fenics?
>> >
>> > It makes sense to me to just provide one. In that case I'd say the
>> > fenics package is the one to list (in the Debian mathematics task),
>> > grabbing the whole set of fenics components.
>>
>> Yes, the fenics package is the one we recommend our users to install.
>> In the FEniCS PPA on Launchpad, the fenics meta package also includes
>> mshr.
>>
>
> Thanks, I'll pass that recommendation on to Andreas.
>
> In our fenics package, we could set
>   Suggests: mshr
> (until CGAL liberation is ready in the mshr package)
>
> That won't get in the way of anything, but it will signal that we do
> have mshr available for Debian (in a sense). Users can easily build it
> from the git repo.

Yes, that is a good idea.

Johannes

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Drew Parsons]

On Thu, 2017-01-05 at 18:11 +0100, Johannes Ring wrote:
> On Thu, Jan 5, 2017 at 6:03 PM, Drew Parsons 
> wrote:
> > 
> > No worries, thanks Johannes. We'll keep the fenics metapackage.
> > 
> > In regards to Andreas' task package, is there any sense putting in
> > two
> > references, one for dolfin and one for fenics?
> > 
> > It makes sense to me to just provide one. In that case I'd say the
> > fenics package is the one to list (in the Debian mathematics task),
> > grabbing the whole set of fenics components.
> 
> Yes, the fenics package is the one we recommend our users to install.
> In the FEniCS PPA on Launchpad, the fenics meta package also includes
> mshr.
> 

Thanks, I'll pass that recommendation on to Andreas.

In our fenics package, we could set
  Suggests: mshr
(until CGAL liberation is ready in the mshr package)

That won't get in the way of anything, but it will signal that we do
have mshr available for Debian (in a sense). Users can easily build it
from the git repo.

Drew

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Johannes Ring]

On Thu, Jan 5, 2017 at 6:03 PM, Drew Parsons  wrote:
> On Thu, 2017-01-05 at 14:13 +0100, Johannes Ring wrote:
>> On Thu, Jan 5, 2017 at 6:04 AM, Drew Parsons 
>> wrote:
>> > dolfin is the front-end of the FENiCS system, and I think the
>> > fenics
>> > package is now deprecated.  Johannes, did you intend to update the
>> > fenics Debian metapackage, or should we remove it from the Debian
>> > archive?
>>
>> I have already updated the fenics meta package, but was waiting for
>> dolfin to be available in unstable before uploading. Now that dolfin
>> has been uploaded (thanks Drew!), I will go ahead and upload the
>> fenics meta package.
>>
>> > We could handle it in the dolfin source package to save
>> > maintenance efforts if you want to keep it.
>>
>> I prefer to keep it as it is if that is okay.
>
> No worries, thanks Johannes. We'll keep the fenics metapackage.
>
> In regards to Andreas' task package, is there any sense putting in two
> references, one for dolfin and one for fenics?
>
> It makes sense to me to just provide one. In that case I'd say the
> fenics package is the one to list (in the Debian mathematics task),
> grabbing the whole set of fenics components.

Yes, the fenics package is the one we recommend our users to install.
In the FEniCS PPA on Launchpad, the fenics meta package also includes
mshr.

Johannes

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Drew Parsons]

On Thu, 2017-01-05 at 14:13 +0100, Johannes Ring wrote:
> On Thu, Jan 5, 2017 at 6:04 AM, Drew Parsons 
> wrote:
> > dolfin is the front-end of the FENiCS system, and I think the
> > fenics
> > package is now deprecated.  Johannes, did you intend to update the
> > fenics Debian metapackage, or should we remove it from the Debian
> > archive?
> 
> I have already updated the fenics meta package, but was waiting for
> dolfin to be available in unstable before uploading. Now that dolfin
> has been uploaded (thanks Drew!), I will go ahead and upload the
> fenics meta package.
> 
> > We could handle it in the dolfin source package to save
> > maintenance efforts if you want to keep it.
> 
> I prefer to keep it as it is if that is okay.

No worries, thanks Johannes. We'll keep the fenics metapackage.

In regards to Andreas' task package, is there any sense putting in two
references, one for dolfin and one for fenics?  

It makes sense to me to just provide one. In that case I'd say the
fenics package is the one to list (in the Debian mathematics task),
grabbing the whole set of fenics components.

Drew

Re: Please categorise your packages for the Debian Science metapackages

[Thibaut Paumard]

Dear Andreas,

Le 05/01/2017 à 14:39, Andreas Tille a écrit :
> Since this script has the logic to check whether a package that is
> maintained by the Debian Science team is mentioned in Debian Science
> tasks the best way to do the exclusion would be rather if you move your
> packages into Debian Astro team.
> 
> I have also *not* included yorick-mira into any Debian Astro which you
> definitely should if you think it should be part of some metapackage.

Yes, that's actually what I meant. I'll do that with the next upload in
the next few days.

Regards, Thibaut.

Re: Please categorise your packages for the Debian Science metapackages

[Ole Streicher]

On 05.01.2017 16:10, Andreas Tille wrote:
> For Debian Med its a bit different since its not only Biology.  So
> adding the biology related metapackages seems somehow reasonable.  I
> think the current approach works (at least I do not see any problem with
> it) so from my point of view we can leave it but I have no really strong
> opinion about it.

I just added DebiChem, Debian Med and Debian Astro as "related Blends"
to the wiki. This should prevent people from assuming there is no
astronomy here :-)

Cheers

Ole

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi Ole,

On Thu, Jan 05, 2017 at 03:34:52PM +0100, Ole Streicher wrote:
> 
> When I removed the astronomy tasks from Debian Science, I added two
> transitional packages directly to debian/control.stub: science astronomy
> and science-astronomy-dev, to provide a smooth transition between the
> old tasks and Debian Astro.

Ahhh, yes, now I remember. :-)
 
> I think, now Debian Astro should just (manually) mentioned somewhere on
> the Debian Science Web page and linked to the home page. Meta-tasks that
> pull other Pure Blends completely is IMO a bit overkill. But we should
> make a common decision here together with the other Blends that somehow
> relate to Science: DebianMed, DebiChem, etc.

For Debian Med its a bit different since its not only Biology.  So
adding the biology related metapackages seems somehow reasonable.  I
think the current approach works (at least I do not see any problem with
it) so from my point of view we can leave it but I have no really strong
opinion about it.

Kind regards

  Andreas. 

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Thomas Schiex]

Agreed. This is very weird from a Debian user point of view.

Astronomy is not Science ?

Thomas

> I realised that the astronomy task fully vanished from Debian Science
> since now there is a Debian Astro Blend.  IMHO it makes sense to
> consider at least a task that sumarises all Debian Astro metapackages.
> Currently it looks as if Debian does not have any astronomy packages if
> somebody would have an uneducated look at Debian Science. 

-- 
New address: thomas.sch...@inra.fr

Re: Please categorise your packages for the Debian Science metapackages

[Ole Streicher]

Hi Thibaut, Andreas,

On 05.01.2017 14:39, Andreas Tille wrote:
> I realised that the astronomy task fully vanished from Debian Science
> since now there is a Debian Astro Blend.  IMHO it makes sense to
> consider at least a task that sumarises all Debian Astro metapackages.
> Currently it looks as if Debian does not have any astronomy packages if
> somebody would have an uneducated look at Debian Science. 

When I removed the astronomy tasks from Debian Science, I added two
transitional packages directly to debian/control.stub: science astronomy
and science-astronomy-dev, to provide a smooth transition between the
old tasks and Debian Astro.

I think, now Debian Astro should just (manually) mentioned somewhere on
the Debian Science Web page and linked to the home page. Meta-tasks that
pull other Pure Blends completely is IMO a bit overkill. But we should
make a common decision here together with the other Blends that somehow
relate to Science: DebianMed, DebiChem, etc.

> I have also *not* included yorick-mira into any Debian Astro which you
> definitely should if you think it should be part of some metapackage.

You (Thibaut) could put yorick-mira into the "datareduction" task of
Debian Astro. Yorick itself (yorick-full) is already listed under
"frameworks":

https://blends.debian.org/astro/tasks/

We also have yorick-cubeview ad yorick-spydr in viewers, and yorick-yao
in simulation. Please adjust these as well if you like.

https://anonscm.debian.org/cgit/debian-astro/packages/debian-astro.git/

Best

Ole

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi Jose,

On Thu, Jan 05, 2017 at 03:21:39PM +0100, Jose Luis Rivero wrote:
> Thanks Andreas for the reminder. I've updated gazebo/sdformat versions
> and included gazebo into the simulation category. git format-patch attached.

Patch pushed. :-)
 
> Happy new year.

Same to you

Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

On Wed, Jan 04, 2017 at 05:47:50PM +0100, Maarten van Gompel wrote:
> 
> python-pynlpl and uctodata should go to 'linguistics'

I did the following change:

$ git diff HEAD^
diff --git a/tasks/linguistics b/tasks/linguistics
index e414e94..197665f 100644
--- a/tasks/linguistics
+++ b/tasks/linguistics
@@ -125,3 +125,8 @@ Suggests: libcg3-dev
 Depends: libcld2-dev
 
 Suggests: r-cran-tm
+
+Depends: python3-pynlpl | python-pynlpl
+
+Depends: uctodata
+


Thanks for your cooperation

Andreas. 

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Jose Luis Rivero]

On 04/01/17 15:50, Andreas Tille wrote:
> Hi,
> 
> as in every release cycle I'm trying to verify that every package
> maintained in Debian Science team is properly categorised in our Blends
> tasks.  If a package is just a predependency for some other scientific
> software I can add it to a blacklist of packages that should not show
> up in the Debian Science metapackages explicitly.  I have uploaded the
> list of not yet categorised (not yet blacklisted) source packages to
> 
>http://blends.debian.net/tmp/debian-science.txt
> 

Thanks Andreas for the reminder. I've updated gazebo/sdformat versions
and included gazebo into the simulation category. git format-patch attached.

Happy new year.

-- 
Jose Luis Rivero 
>From 60907a695f95ec9290c040e318361228287f858d Mon Sep 17 00:00:00 2001
From: Jose Luis Rivero 
Date: Thu, 5 Jan 2017 15:17:18 +0100
Subject: [PATCH] Update gazebo and sdformat versioned packages.

Update gazebo and sdformat versioned pacakges to the latest version and
added gazebo to the tasks/simulation blend.
---
 tasks/robotics | 2 +-
 tasks/robotics-dev | 4 ++--
 tasks/simulations  | 2 ++
 3 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/tasks/robotics b/tasks/robotics
index a09b595..b85b7c0 100644
--- a/tasks/robotics
+++ b/tasks/robotics
@@ -176,7 +176,7 @@ Pkg-Description: Point Cloud Library
  large consortium of researchers and engineers around the world. It is
  written in C++ and released under the BSD license.
 
-Depends: gazebo5
+Depends: gazebo7
 
 Depends: openrave
 Homepage: http://openrave.org/
diff --git a/tasks/robotics-dev b/tasks/robotics-dev
index ec6200c..b0cf433 100644
--- a/tasks/robotics-dev
+++ b/tasks/robotics-dev
@@ -29,7 +29,7 @@ Depends: liburdfdom-dev, liburdfdom-headers-dev
 
 Depends: libslicot-dev
 
-Depends: libsdformat-dev
+Depends: libsdformat4-dev
 
 Depends: libconsole-bridge-dev
 
@@ -37,7 +37,7 @@ Depends: libcomedi-dev, python-comedilib
 
 Depends: libccd-dev
 
-Depends: libgazebo5-dev
+Depends: libgazebo7-dev
 
 Depends: libsimbody-dev
 
diff --git a/tasks/simulations b/tasks/simulations
index 58201f4..d05aad9 100644
--- a/tasks/simulations
+++ b/tasks/simulations
@@ -33,6 +33,8 @@ Depends: python3-woo | python-woo
 
 Depends: libceres-dev
 
+Depends: gazebo7
+
 Suggests:  ceres-solver-doc
 
 Depends: python-escript | python3-escript | python-escript-mpi | python3-escript-mpi
-- 
2.7.4

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi Julien,

On Thu, Jan 05, 2017 at 09:03:41AM +0100, Julien Puydt wrote:
> Ah, thanks for the heads-up and taking care of the blends!

You are welcome.
 
> Those should clearly be mathematics :
> - edge-addition-planarity-suite

I added planarity since the tasks files contain *binary* packages.

> - flint-arb

Hmmm, this is a library.  Are you sure that it is a user oriented
package.  For the moment I've added libflint-arb-dev to mathematics-dev.

> - flintqs

Added.
 
> and that one should be mathematics-dev:
> - rw

Added.

> The case of rubiks isn't clear-cut: it could be seen as mathematics,
> geometry, combinatorics (non-existing) or "brainy game" (non-existing) -- it
> contains Rubiks' cubes solvers, after all. What do you think about it?

Its t long ago that I was quite good with Rubiks' cube. :-P

Honestly, tasks are not either or.  I've added rubiks to mathematics and
added also a Suggests in geometry.  Any reader feel free to suggest
something additional or different.
 
Here are all changes I did:

diff --git a/tasks/geometry b/tasks/geometry
index bf8aae4..d154bf6 100644
--- a/tasks/geometry
+++ b/tasks/geometry
@@ -22,3 +22,4 @@ Suggests: libmath-geometry-voronoi-perl
 
 Suggests: octave-geometry
 
+Suggests: rubiks
diff --git a/tasks/mathematics b/tasks/mathematics
index 3e81e34..726074b 100644
--- a/tasks/mathematics
+++ b/tasks/mathematics
@@ -162,6 +162,10 @@ Depends: frobby
 
 Depends: geneagrapher
 
+Depends: planarity, flintqs
+
+Depends: rubiks
+
 X-Removed: Packages removed from Debian
 
 Depends: octaviz
diff --git a/tasks/mathematics-dev b/tasks/mathematics-dev
index 5569cb7..d53f967 100644
--- a/tasks/mathematics-dev
+++ b/tasks/mathematics-dev
@@ -238,3 +238,5 @@ WNPP: 832659
 Depends: python3-gimmik
 
 Depends: cppad
+
+Depends: libflint-arb-dev, librw-dev


Kind regards

  Andreas.


-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi Thibaut

On Thu, Jan 05, 2017 at 07:07:16AM +0100, Thibaut Paumard wrote:
> All my packages should really go to astronomy. Two sets of packages I
> would hide in any case, the last one would need to fit in the right task
> in the astronomy blend.
> 
> (python|python3|yorick)-pyorick hide
> yorick-mira should go to astronomy
> (python|python3|yorick)-svipc hide

I realised that the astronomy task fully vanished from Debian Science
since now there is a Debian Astro Blend.  IMHO it makes sense to
consider at least a task that sumarises all Debian Astro metapackages.
Currently it looks as if Debian does not have any astronomy packages if
somebody would have an uneducated look at Debian Science. 

So I did *not* added yorick-mira to astronomy since this task does
not exists.  I rather excluded all packages via

diff --git a/misc/sql/0-uncategorised.sh b/misc/sql/0-uncategorised.sh
index be73f1a..da02cd6 100755
--- a/misc/sql/0-uncategorised.sh
+++ b/misc/sql/0-uncategorised.sh
@@ -17,7 +17,7 @@ case "$1" in
 
'lua-torch-cwrap','lua-torch-dok','lua-torch-paths','lua-torch-sundown',
 'kwwidgets',
 'mpi-defaults',
-'polyml',
+'polyml','python-pyorick',
 
'r-cran-formatr','r-cran-gsl','r-cran-jsonlite','r-cran-learnbayes','r-cran-lubridate',
 'r-cran-maxlik','r-cran-mime','r-cran-misctools',
 
'r-cran-polyclip','r-cran-polycub','r-cran-randomfields','r-cran-repr','r-cran-rjson','r-cran-rprotobuf',
@@ -34,6 +34,7 @@ case "$1" in
 'ros-wstool',
 'vmtk',-- is mentioned in Debian Med; there is 
no really good fitting task in Debian Science
 'xmds-doc',
+'yp-svipc',
 'yorick-av',
 'yorick-curses',
 'yorick-gl',
@@ -42,6 +43,7 @@ case "$1" in
 'yorick-imutil',
 'yorick-ml4',
 'yorick-mpeg',
+'yorick-mira', -- belongs to Debian Astro and should not be 
listed in Debian Science
 'yorick-optimpack',
 'yorick-soy',
 'yorick-ygsl',

Since this script has the logic to check whether a package that is
maintained by the Debian Science team is mentioned in Debian Science
tasks the best way to do the exclusion would be rather if you move your
packages into Debian Astro team.

I have also *not* included yorick-mira into any Debian Astro which you
definitely should if you think it should be part of some metapackage.

Kind regards

   Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Johannes Ring]

On Thu, Jan 5, 2017 at 6:04 AM, Drew Parsons  wrote:
> dolfin is the front-end of the FENiCS system, and I think the fenics
> package is now deprecated.  Johannes, did you intend to update the
> fenics Debian metapackage, or should we remove it from the Debian
> archive?

I have already updated the fenics meta package, but was waiting for
dolfin to be available in unstable before uploading. Now that dolfin
has been uploaded (thanks Drew!), I will go ahead and upload the
fenics meta package.

>We could handle it in the dolfin source package to save
> maintenance efforts if you want to keep it.

I prefer to keep it as it is if that is okay.

Johannes

> Whether we want the fenics package removed or not, I guess there's no
> need to have the two entries in the task list for fenics/dolfin. Once
> we're decided I can log in to the tasks server to make the edits to
> keep one entry.
>
> Drew
>
> On Wed, 2017-01-04 at 15:50 +0100, Andreas Tille wrote:
>> Hi,
>>
>> as in every release cycle I'm trying to verify that every package
>> maintained in Debian Science team is properly categorised in our
>> Blends
>> tasks.  If a package is just a predependency for some other
>> scientific
>> software I can add it to a blacklist of packages that should not show
>> up in the Debian Science metapackages explicitly.  I have uploaded
>> the
>> list of not yet categorised (not yet blacklisted) source packages to
>>
>>http://blends.debian.net/tmp/debian-science.txt
>>
>> The list also contains the latest uploader - so simply seek for your
>> name - everybody in CC is mentioned at least once in the list and
>> should
>> definitely have a look.  If other readers here feel competent to
>> classify a package for one or more (!) tasks in our task list
>>
>>https://blends.debian.org/science/tasks/
>>
>> evary suggestion is welcome.
>>
>> To add a binary (!) package to a task you can simply
>>
>> debcheckout -u your_alioth_login debian-science
>> cd debian-science/tasks
>>
>> and edit the task in question.  Members of the Debian Pure Blends
>> team
>> as well as any DD (ACLs are set but I have heard this does not work
>> reliably) have commit permissions.  I'm also fine if you debcheckout
>> anonymously and send me `git format-patch` formated changes or just
>> answer here to this mail.
>>
>> If you are not sure whether a package belongs to a task or not feel
>> free to discuss this here.
>>
>> Kind regards and thanks for your cooperation
>>
>>   Andreas.
>>

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Andreas Tille]

Hi Drew,

I'm not sure whether I fully understand your suggestion.

On Thu, Jan 05, 2017 at 01:26:46PM +0800, Drew Parsons wrote:
> Hi Andreas, there are a couple more fenics components in the 
> mathematics-dev task:
> 
>   python-dolfin (and dolfin-dev)
>   python-ffc
>   python-ufl
> 
> I'm thinking it's redundant to list these separately under 
> mathematics-dev if there's already an entry in the mathematics task.

Do you want these three packages to be under mathematics-dev or
mathematics in other words are these more developer oriented or user
oriented packages.  In principle it is perfectly fine to list packages
in more than one task if this makes sense.
 
> Likewise in your not-yet categorised list:
>   dijitso
> 
> python-dijitso is another of the FENiCs components that I think doesn't
> need to be listed separately.

$ git diff HEAD^
diff --git a/misc/sql/0-uncategorised.sh b/misc/sql/0-uncategorised.sh
index c97012c..be73f1a 100755
--- a/misc/sql/0-uncategorised.sh
+++ b/misc/sql/0-uncategorised.sh
@@ -10,7 +10,7 @@ case "$1" in
 debian-science)
 
team="'debian-science-maintain...@lists.alioth.debian.org','pkg-scicomp-de...@lists.alioth.debian.org'"
 ignore="'cbp2make','coinor-cgl','coinor-osi','coinutils',
-'dh-r',
+'dh-r','dijitso',
 'dune-common', 'dune-geometry', -- Dependencies of dune-grid
 'gmp',  -- very basic, used in lots of 
other packages as dependency
 'imview-doc', 'instant',

 
> How is "blacklisting" done? Is it just a matter of removing the entries
> in the files in debian-science/tasks ?

Its a bit hidden "tool" - or rather a quickly hacked script available
here:


https://anonscm.debian.org/cgit/blends/website.git/tree/misc/sql/0-uncategorised.sh

To make a proper tool the excluded source packages should be either in a
config file or a database table - but for the moment it worked for me
and so far nobody else showed interest to run it. :-)

Thanks for the contribution

  Andreas.

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Andreas Tille]

Hi,

On Thu, Jan 05, 2017 at 02:46:55AM +, lumin wrote:
> I'm holding 6 uncategorized d-science packages.
> 
> * caffe and caffe-contrib are categorized into machine-learning task.
>   See the patch attached.

Thanks for the patch.  Please note that for packages that are either
existing in Debian (like caffe-cpu and caffe-cuda) or at least have some
packaging in Debian Science Git or SVN (that's the reason why I insist
using {git|svn}.debian.org) there is no need to specify anything else
than just the package name.  The information is in UDD and will be taken
from there.  So I striped your patch to only contain

   Depends: caffe-cpu | caffe-cuda
 
> * the remaining 4 packages are core components of the torch7 framework,
>   and the torch7 metapackage (meta-torch-core-free) is still in experimental.
>   Torch7 is also a machine-learning framework.

OK, I did:

$ git diff HEAD^
diff --git a/tasks/machine-learning b/tasks/machine-learning
index e0d3456..8c52063 100644
--- a/tasks/machine-learning
+++ b/tasks/machine-learning
@@ -147,3 +147,5 @@ Suggests: libdlib-dev
 Depends: r-cran-mlbench
 
 Depends: caffe-cpu | caffe-cuda
+
+Depends: torch-core-free


Explanation: The package will show up on the relevant tasks page to
inform users that they will find the package in experimental.  When
creating Debian Science metapackages it will only get a

Suggests: torch-core-free

since the metapackage creation process verifies whether a package is in
testing or not.  So there is no harm done in specifying this package in
the tasks files.  In other words:  Please add your packages early (be it
in VCS, new or experimental) to the tasks files.

 
>   lua-torch-cwrap
>   lua-torch-dok
>   lua-torch-paths
>   lua-torch-sundown
> 
>   The 4 source packages should be hidden.

I excluded these four packages in the helper script[1] which I'm using.

Thanks for your cooperation

 Andreas.


[1] 
https://anonscm.debian.org/cgit/blends/website.git/tree/misc/sql/0-uncategorised.sh

-- 
http://fam-tille.de

Re: Please categorise your packages for the Debian Science metapackages

[Julien Puydt]

Hi,

On 04/01/2017 15:50, Andreas Tille wrote:


as in every release cycle I'm trying to verify that every package
maintained in Debian Science team is properly categorised in our Blends
tasks.  If a package is just a predependency for some other scientific
software I can add it to a blacklist of packages that should not show
up in the Debian Science metapackages explicitly.  I have uploaded the
list of not yet categorised (not yet blacklisted) source packages to

   http://blends.debian.net/tmp/debian-science.txt

The list also contains the latest uploader - so simply seek for your
name - everybody in CC is mentioned at least once in the list and should
definitely have a look.  If other readers here feel competent to
classify a package for one or more (!) tasks in our task list

   https://blends.debian.org/science/tasks/

evary suggestion is welcome.

To add a binary (!) package to a task you can simply

debcheckout -u your_alioth_login debian-science
cd debian-science/tasks

and edit the task in question.  Members of the Debian Pure Blends team
as well as any DD (ACLs are set but I have heard this does not work
reliably) have commit permissions.  I'm also fine if you debcheckout
anonymously and send me `git format-patch` formated changes or just
answer here to this mail.

If you are not sure whether a package belongs to a task or not feel
free to discuss this here.

Kind regards and thanks for your cooperation


Ah, thanks for the heads-up and taking care of the blends!

Those should clearly be mathematics :
- edge-addition-planarity-suite
- flint-arb
- flintqs

and that one should be mathematics-dev:
- rw

The case of rubiks isn't clear-cut: it could be seen as mathematics, 
geometry, combinatorics (non-existing) or "brainy game" (non-existing) 
-- it contains Rubiks' cubes solvers, after all. What do you think about it?


Cheers,

Snark on #debian-science

Re: Please categorise your packages for the Debian Science metapackages

[Thibaut Paumard]

Dear Andreas,

All my packages should really go to astronomy. Two sets of packages I
would hide in any case, the last one would need to fit in the right task
in the astronomy blend.

(python|python3|yorick)-pyorick hide
yorick-mira should go to astronomy
(python|python3|yorick)-svipc hide

Kind regards, Thibaut.

Le 04/01/2017 à 15:50, Andreas Tille a écrit :
> Hi,
> 
> as in every release cycle I'm trying to verify that every package
> maintained in Debian Science team is properly categorised in our Blends
> tasks.  If a package is just a predependency for some other scientific
> software I can add it to a blacklist of packages that should not show
> up in the Debian Science metapackages explicitly.  I have uploaded the
> list of not yet categorised (not yet blacklisted) source packages to
> 
>http://blends.debian.net/tmp/debian-science.txt
> 
> The list also contains the latest uploader - so simply seek for your
> name - everybody in CC is mentioned at least once in the list and should
> definitely have a look.  If other readers here feel competent to
> classify a package for one or more (!) tasks in our task list
> 
>https://blends.debian.org/science/tasks/
> 
> evary suggestion is welcome.
> 
> To add a binary (!) package to a task you can simply
> 
> debcheckout -u your_alioth_login debian-science
> cd debian-science/tasks
> 
> and edit the task in question.  Members of the Debian Pure Blends team
> as well as any DD (ACLs are set but I have heard this does not work
> reliably) have commit permissions.  I'm also fine if you debcheckout
> anonymously and send me `git format-patch` formated changes or just
> answer here to this mail.
> 
> If you are not sure whether a package belongs to a task or not feel
> free to discuss this here.
> 
> Kind regards and thanks for your cooperation
> 
>   Andreas.
> 

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Drew Parsons]

Hi Andreas, there are a couple more fenics components in the 
mathematics-dev task:

  Python-dolfin (and dolfin-dev)
  python-ffc
  python-ufl

I'm thinking it's redundant to list these separately under 
mathematics-dev if there's already an entry in the mathematics task.

Likewise in your not-yet categorised list:
  dijitso

python-dijitso is another of the FENiCs components that I think doesn't
need to be listed separately.

How is "blacklisting" done? Is it just a matter of removing the entries
in the files in debian-science/tasks ?

Drew

On Thu, 2017-01-05 at 13:04 +0800, Drew Parsons wrote:
> dolfin-bin and fenics are currently listed under the mathematics
> task.
> 
> In a sense python-dolfin is more appropriate for listing than dolfin-
> bin.  But dolfin-bin depends on it and also provides a couple of
> small
> utility programs. So I'm inclined to leave dolfin-bin as listed
> 
> dolfin is the front-end of the FENiCS system, and I think the fenics
> package is now deprecated.  Johannes, did you intend to update the
> fenics Debian metapackage, or should we remove it from the Debian
> archive?  We could handle it in the dolfin source package to save
> maintenance efforts if you want to keep it.
> 
> Whether we want the fenics package removed or not, I guess there's no
> need to have the two entries in the task list for fenics/dolfin. Once
> we're decided I can log in to the tasks server to make the edits to
> keep one entry.
> 
> Drew
> 
> On Wed, 2017-01-04 at 15:50 +0100, Andreas Tille wrote:
> > Hi,
> > 
> > as in every release cycle I'm trying to verify that every package
> > maintained in Debian Science team is properly categorised in our
> > Blends
> > tasks.  If a package is just a predependency for some other
> > scientific
> > software I can add it to a blacklist of packages that should not
> > show
> > up in the Debian Science metapackages explicitly.  I have uploaded
> > the
> > list of not yet categorised (not yet blacklisted) source packages
> > to
> > 
> >    http://blends.debian.net/tmp/debian-science.txt
> > 
> > The list also contains the latest uploader - so simply seek for
> > your
> > name - everybody in CC is mentioned at least once in the list and
> > should
> > definitely have a look.  If other readers here feel competent to
> > classify a package for one or more (!) tasks in our task list
> > 
> >    https://blends.debian.org/science/tasks/
> > 
> > evary suggestion is welcome.
> > 
> > To add a binary (!) package to a task you can simply
> > 
> > debcheckout -u your_alioth_login debian-science
> > cd debian-science/tasks
> > 
> > and edit the task in question.  Members of the Debian Pure Blends
> > team
> > as well as any DD (ACLs are set but I have heard this does not work
> > reliably) have commit permissions.  I'm also fine if you
> > debcheckout
> > anonymously and send me `git format-patch` formated changes or just
> > answer here to this mail.
> > 
> > If you are not sure whether a package belongs to a task or not feel
> > free to discuss this here.
> > 
> > Kind regards and thanks for your cooperation
> > 
> >   Andreas.
> > 
> 
> 

Re: Please categorise your packages for the Debian Science metapackages: fenics/dolfin

[Drew Parsons]

dolfin-bin and fenics are currently listed under the mathematics task.

In a sense python-dolfin is more appropriate for listing than dolfin-
bin.  But dolfin-bin depends on it and also provides a couple of small
utility programs. So I'm inclined to leave dolfin-bin as listed

dolfin is the front-end of the FENiCS system, and I think the fenics
package is now deprecated.  Johannes, did you intend to update the
fenics Debian metapackage, or should we remove it from the Debian
archive?  We could handle it in the dolfin source package to save
maintenance efforts if you want to keep it.

Whether we want the fenics package removed or not, I guess there's no
need to have the two entries in the task list for fenics/dolfin. Once
we're decided I can log in to the tasks server to make the edits to
keep one entry.

Drew

On Wed, 2017-01-04 at 15:50 +0100, Andreas Tille wrote:
> Hi,
> 
> as in every release cycle I'm trying to verify that every package
> maintained in Debian Science team is properly categorised in our
> Blends
> tasks.  If a package is just a predependency for some other
> scientific
> software I can add it to a blacklist of packages that should not show
> up in the Debian Science metapackages explicitly.  I have uploaded
> the
> list of not yet categorised (not yet blacklisted) source packages to
> 
>    http://blends.debian.net/tmp/debian-science.txt
> 
> The list also contains the latest uploader - so simply seek for your
> name - everybody in CC is mentioned at least once in the list and
> should
> definitely have a look.  If other readers here feel competent to
> classify a package for one or more (!) tasks in our task list
> 
>    https://blends.debian.org/science/tasks/
> 
> evary suggestion is welcome.
> 
> To add a binary (!) package to a task you can simply
> 
> debcheckout -u your_alioth_login debian-science
> cd debian-science/tasks
> 
> and edit the task in question.  Members of the Debian Pure Blends
> team
> as well as any DD (ACLs are set but I have heard this does not work
> reliably) have commit permissions.  I'm also fine if you debcheckout
> anonymously and send me `git format-patch` formated changes or just
> answer here to this mail.
> 
> If you are not sure whether a package belongs to a task or not feel
> free to discuss this here.
> 
> Kind regards and thanks for your cooperation
> 
>   Andreas.
> 

Re: Please categorise your packages for the Debian Science metapackages

[lumin]

(Dropping the CC list)

Hi Andreas,

I'm holding 6 uncategorized d-science packages.

* caffe and caffe-contrib are categorized into machine-learning task.
  See the patch attached.

* the remaining 4 packages are core components of the torch7 framework,
  and the torch7 metapackage (meta-torch-core-free) is still in experimental.
  Torch7 is also a machine-learning framework.

  lua-torch-cwrap
  lua-torch-dok
  lua-torch-paths
  lua-torch-sundown

  The 4 source packages should be hidden.

Thanks.

On Wed, 2017-01-04 at 15:50 +0100, Andreas Tille wrote:
> I'm also fine if you debcheckout
> anonymously and send me `git format-patch` formated changes or just
> answer here to this mail.
From e0811df11c582cd53eccea3feb02b50a8c0b6a9c Mon Sep 17 00:00:00 2001
From: Zhou Mo 
Date: Thu, 5 Jan 2017 02:39:44 +
Subject: [PATCH] categorize caffe and caffe-contrib to machine-learning

---
 tasks/machine-learning | 8 
 1 file changed, 8 insertions(+)

diff --git a/tasks/machine-learning b/tasks/machine-learning
index c69534b..a27eecc 100644
--- a/tasks/machine-learning
+++ b/tasks/machine-learning
@@ -145,3 +145,11 @@ Suggests: ask
 Suggests: libdlib-dev
 
 Depends: r-cran-mlbench
+
+Depends: caffe-cpu | caffe-cuda
+License: BSD-2-Clause
+Homepage: http://caffe.berkeleyvision.org/
+Pkg-Description: Fast, open framework for Deep Learning
+ Caffe is a deep learning framework made with expression, speed,
+ and modularity in mind. It is developed by the Berkeley Vision
+ and Learning Center (BVLC) and community contributors.
-- 
2.11.0

Re: Please categorise your packages for the Debian Science metapackages

[Maarten van Gompel]

Hi Andreas,

Quoting Andreas Tille (2017-01-04 15:50:17)
> as in every release cycle I'm trying to verify that every package
> maintained in Debian Science team is properly categorised in our Blends
> tasks.  

> I'm also fine if you debcheckout
> anonymously and send me `git format-patch` formated changes or just
> answer here to this mail.

python-pynlpl and uctodata should go to 'linguistics'

Regards,

--

Maarten van Gompel
Centre for Language Studies
Radboud Universiteit Nijmegen

proy...@anaproy.nl
http://proycon.anaproy.nl
http://github.com/proycon

GnuPG key:  0x1A31555C  
XMPP: proy...@anaproy.nl  Matrix: @proycon:anaproy.nl
Telegram:   proycon   IRC: proycon (freenode)
Twitter:https://twitter.com/proycon
Bitcoin:1BRptZsKQtqRGSZ5qKbX2azbfiygHxJPsd