Indeed! My idea was to call fenics-install.sh from the web interface
on fenics-project.org using curl.

Or, perhaps to ensure repeatability of virtual builds, I would
curl/wget it from the raw file interface on bitbucket where we can ask
for a specific commit of fenics-install.sh.

Infact, might I suggest that instead of placing fenics-install-all.sh
directly on the webserver, it might make sense for fenics-project.org
to return a HTTP 303 to the raw file on bitbucket. Then there is no
disconnect between the repository and the website version. You could
also specify a known working commit or branch this way. So that will
be the 'stable' version.

Jack
-----
Dr. Jack S. Hale

Marie Skłodowska-Curie Postdoctoral Fellow

University of Luxembourg
Campus Kirchberg G005
Phone +352 44 66 44 5236
[email protected]

Latest publications and conferences: http://goo.gl/rNiISG
ORCID: http://orcid.org/0000-0001-7216-861X
Google Scholar: http://scholar.google.com/citations?user=Fx9lQ7MAAAAJ&hl=de




On 21 January 2015 at 11:42, Anders Logg <[email protected]> wrote:
> But that argument also goes for fenics-install.sh. That is also not a
> developer tool. The developer tools in that repository are:
>
> fenics-install-component.sh (previously cmake.local)
> fenics-install-all.sh (calls fenics-install.sh to build deps and then
> fenics-install-component.sh for each FEniCS component)
>
> Wouldn't it be impractical for the virtual containers to depend on a script
> inside another repository (fenics-install-all.sh)?
>
> Or would it be calling fenics-install-all.sh via the web interface (wget |
> bash)? Then it might work.
>
> I think the most important points here are:
>
> 1. That the tools (HashDist scripts + virtual containers) are easy to find
> by users
> 2. That the tools can be used with minimal hassle (requiring on the order of
> one single command to run)
> 3. That if the tools depend intimately on each other, they be put in the
> same repository to make fixes simple
>
> Another possibility would be to find a new and better name for
> fenics-developer-tools or a new name for a completely new repository and
> throw everything in there.
>
> --
> Anders
>
>
> Wed Jan 21 2015 at 11:36:10 AM skrev Garth N. Wells <[email protected]>:
>>
>> I would favour (1) because the virtual environments/containers are not
>> just developer tools.
>>
>> A nice things is that containers/virtual environments and Hashdist are
>> complementary; Hashdist will always be a challenge because it cannot know
>> all the details of every system it's deployed on. Using Hashdist inside a
>> container means Hashdist can operate inside a well-defined and controlled
>> environment, and one which can be moved across platforms. I can see
>> containers + Hashdist being a great way for developers to test different
>> platforms (and debug buildbot failures).
>>
>> Garth
>>
>> On 21 January 2015 at 08:10, Anders Logg <[email protected]> wrote:
>>>
>>> Good proposal. I would vote for (2) to:
>>>
>>> (i) tie the two efforts more closely together
>>> (ii) make development easier (I imagine a small change in one of them
>>> might require a small change in the other)
>>>
>>> --
>>> Anders
>>>
>>>
>>> Wed Jan 21 2015 at 9:00:33 AM skrev Jack HALE <[email protected]>:
>>>
>>>> As some of you might know, Garth Wells, Lizao Li and I have been
>>>> working on virtual environments for portable and reusable distribution
>>>> of FEniCS. This work is in garth-wells/fenics-virtual mainly under the
>>>> docker branch.
>>>>
>>>> The hashdist effort provides an excellent, simple and consistent
>>>> cross-platform way of building FEniCS. Nonetheless, I do not think it
>>>> provides:
>>>>
>>>> a) a really, really easy environment for absolute beginners on Windows.
>>>> b) a completely consistent environment for; teaching, repeatability of
>>>> results, cross-platform use within a research group.
>>>> c) a method for quickly moving the same environment from the users
>>>> computer to a cluster environment.
>>>>
>>>> However, I think that together the two projects should complement each
>>>> other nicely.
>>>>
>>>> Within the fenics-virtual project we essentially have our own set of
>>>> build scripts, but it seems sensible to me to re-write at least some
>>>> of our virtual environments to use the new Hashdist scripts. More
>>>> specifically, Docker stable-ppa and vagrant stable-ppa would continue
>>>> to use the PPA archives, and Docker developer and stable-src would
>>>> move to using Hashdist.
>>>>
>>>> The two options are:
>>>>
>>>> 1) Bring the re-written garth-wells/fenics-virtual under
>>>> fenics-project and keep fenics-developer-tools separate. Simple!
>>>> 2) Bring the functionality of fenics-virtual directly into
>>>> fenics-developer-tools. The advantage of this is that users and
>>>> developers can immediately see all of the ways we offer for using
>>>> FEniCS. The downside is it introduces complexity.
>>>>
>>>> My personal opinion is to go for option 1) for simplicity and
>>>> separability of the two efforts.
>>>>
>>>> Let me know what you think!
>>>>
>>>> Cheers,
>>>>
>>>> Jack
>>>> _______________________________________________
>>>> fenics mailing list
>>>> [email protected]
>>>> http://fenicsproject.org/mailman/listinfo/fenics
>>>
>>>
>>> _______________________________________________
>>> fenics mailing list
>>> [email protected]
>>> http://fenicsproject.org/mailman/listinfo/fenics
>>>
>
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to