Building RPMs for Fedora with circular "suggest" dependencies

[William Emmanuel S. Yu]

Hello,

I am new to this list and in the process of building some RPMs for R packages. I have been
rolling my own packages for a few years now and post them for public use. I am a heavy
user of R2spec with spec file tweaks of course. Thanks for the author for making life
easier.

Every so often, I encounter the circular dependency issue with "suggest". For
example, I am building R-reshape that depends on a bunch of packages including R-iterator
and R-foreach. Unfortunately, R-iterator and R-foreach both suggest each other. 

I used to be able to force ignoring suggests but this causes problems with check and test
and skipping check and test isn't exactly what I want to do either.

I would like to know what the "proper" way of building RPMs for the both of
these R packages. 

Thank you!
"Sent via BlackBerry from Smart"

[Allen S. Rout]

On 06/15/2012 08:27 AM, William Emmanuel S. Yu wrote:
>  Hello,
>
>  I am new to this list and in the process of building some RPMs for R packages. I have
> been rolling my own packages for a few years now and post them for public use. I am a
> heavy user of R2spec with spec file tweaks of course. Thanks for the author for making
> life easier.
>
>  Every so often, I encounter the circular dependency issue with "suggest". For
> example, I am building R-reshape that depends on a bunch of packages including R-iterator
> and R-foreach. Unfortunately, R-iterator and R-foreach both suggest each other.
>
>  I used to be able to force ignoring suggests but this causes problems with check and test
> and skipping check and test isn't exactly what I want to do either.
>
>  I would like to know what the "proper" way of building RPMs for the both of
> these R packages. 
Hah.

 From the perspective of a scientist or engineer, the 'proper' solution 
would be to have a dependency _TREE_ instead of a dependency 
plate-of-pasta.

In short, your problem looks insane because the underlying reality is 
insane.


http://article.gmane.org/gmane.comp.lang.r.devel/21496

includes my statement of the problem, ca. 2009.

http://article.gmane.org/gmane.comp.lang.r.devel/21597

for a stab at a solution.  It's not elegant, but I think it'd work.


I've frankly been out of the loop on where Pierre-Yves is going on this 
issue for some time.   I'm sure he's got some comments.



- Allen S. Rout

[William Emmanuel S. Yu]

Hi Allen,

Thanks for the quick reply.

I actually encountered this point of yours and that is why I found this mailing list.

On your solution, I did try putting 'export _R_CHECK_FORCE_SUGGESTS_ = false; R CMD
check' in the spec file. But now, it fails again again doing a test that requires
foreach while building iterator. 

Thanks again.
"Sent via BlackBerry from Smart"

-----Original Message-----
From: "Allen S. Rout" <asr(a)ufl.edu&gt;
Sender: r-devel-bounces(a)lists.fedoraproject.org
Date: Fri, 15 Jun 2012 09:09:24 
To: <r-devel(a)lists.fedoraproject.org&gt;
Reply-To: r-devel(a)lists.fedoraproject.org
Subject: Re: [Fedora-r-devel-list] Building RPMs for Fedora with circular
 "suggest" dependencies

On 06/15/2012 08:27 AM, William Emmanuel S. Yu wrote:
>  Hello,
>
>  I am new to this list and in the process of building some RPMs for R packages. I have
> been rolling my own packages for a few years now and post them for public use. I am a
> heavy user of R2spec with spec file tweaks of course. Thanks for the author for making
> life easier.
>
>  Every so often, I encounter the circular dependency issue with "suggest". For
> example, I am building R-reshape that depends on a bunch of packages including R-iterator
> and R-foreach. Unfortunately, R-iterator and R-foreach both suggest each other.
>
>  I used to be able to force ignoring suggests but this causes problems with check and test
> and skipping check and test isn't exactly what I want to do either.
>
>  I would like to know what the "proper" way of building RPMs for the both of
> these R packages. 
Hah.

 From the perspective of a scientist or engineer, the 'proper' solution 
would be to have a dependency _TREE_ instead of a dependency 
plate-of-pasta.

In short, your problem looks insane because the underlying reality is 
insane.


http://article.gmane.org/gmane.comp.lang.r.devel/21496

includes my statement of the problem, ca. 2009.

http://article.gmane.org/gmane.comp.lang.r.devel/21597

for a stab at a solution.  It's not elegant, but I think it'd work.


I've frankly been out of the loop on where Pierre-Yves is going on this 
issue for some time.   I'm sure he's got some comments.



- Allen S. Rout
_______________________________________________
r-devel mailing list
r-devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/r-devel

[José Abílio Matos]

On 06/15/2012 02:25 PM, William Emmanuel S. Yu wrote:
>  Hi Allen,
>
>  Thanks for the quick reply.
>
>  I actually encountered this point of yours and that is why I found this mailing list.
>
>  On your solution, I did try putting 'export _R_CHECK_FORCE_SUGGESTS_ = false; R CMD
> check' in the spec file. But now, it fails again again doing a test that requires
> foreach while building iterator. 
>
>  Thanks again.
>  "Sent via BlackBerry from Smart" 
Hi,
    have you been able to proceed here?

I suppose that the only real option is to proceed in several steps in a
kind of bootstrapping setup, first ignoring the suggests and the checks
and on a second round adding those back. Probably this is easier said
then done. :-)

Regards,

-- 
José Matos

[William Emmanuel S. Yu]

Well. I just built the RPMs to ignore the depedencies and installed them manually (force).
Not elegant but it works and still uses RPM.
"Sent on the move."

-----Original Message-----
From: José Matos <jamatos(a)fc.up.pt&gt;
Sender: r-devel-bounces(a)lists.fedoraproject.org
Date: Wed, 17 Oct 2012 16:30:36 
To: <r-devel(a)lists.fedoraproject.org&gt;
Reply-To: r-devel(a)lists.fedoraproject.org
Subject: Re: [Fedora-r-devel-list] Building RPMs for Fedora with circular
 "suggest" dependencies

On 06/15/2012 02:25 PM, William Emmanuel S. Yu wrote:
>  Hi Allen,
>
>  Thanks for the quick reply.
>
>  I actually encountered this point of yours and that is why I found this mailing list.
>
>  On your solution, I did try putting 'export _R_CHECK_FORCE_SUGGESTS_ = false; R CMD
> check' in the spec file. But now, it fails again again doing a test that requires
> foreach while building iterator. 
>
>  Thanks again.
>  "Sent via BlackBerry from Smart" 
Hi,
    have you been able to proceed here?

I suppose that the only real option is to proceed in several steps in a
kind of bootstrapping setup, first ignoring the suggests and the checks
and on a second round adding those back. Probably this is easier said
then done. :-)

Regards,

-- 
José Matos

_______________________________________________
r-devel mailing list
r-devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/r-devel