Paul F. Kunz wrote:
>>>>>>On Fri, 30 Jun 2006 09:53:53 -0500, Troy Dawson <[log in to unmask]> said:
>
>
>>Andy Buckley wrote:
>>
>>>Paul F. Kunz wrote:
>>>
>>>
>>>>>>>>>On Fri, 16 Jun 2006 14:42:11 +0100, Andy Buckley
>>>>>>>>><[log in to unmask]> said:
>>>>
>>>>- Minuit (C++ version). There's no RPM file for this, altho I
>>>>could contribute one for their next replease. I could also
>>>>include Minuit sources in the source RPM file and make it a
>>>>subpackage hippodraw-minuit.
>>>>
>
>
>>If he could provide the rpm, that would be best. He probrubly knows
>>more about what should be in it.
>
>
> Providing a source RPM for Minuit would be easy. I would go with
> the ROOTified version as the non ROOT version is at its last release.
> In spite of being in namesapce ROOT, it has no other dependencies on
> any part of ROOT (yet).
>
>
>>>>- ROOT. Is there an RPM for ROOT? When I log into lxplus.cern.ch
>>>>(which is SLC 3), there doesn't appear to be any ROOT installed
>>>>and no ROOT in the RPM database. And if there were, it would be
>>>>compiled with gcc 3.2.3, which I can't use if HippoDraw gets
>>>>compiled with gcc 3.4 or later. Of course, I could do like
>>>>Minuit.
>>>>
>
>
>>I tested compiling the Fedora Core 3 Extra's rpm. It recompiled
>>without any problems. Looks do-able.
>
>
>>>>- wcslib. Word Coordinate System, interesting for astrophysics.
>>>>No RPM available.
>>>
>
>>I have a bit of a problem with it's installation. It recompiled
>>just fine. But partway through it's installation it tried to change
>>the permissions of /usr/lib. That's a big no no in my book. I
>>don't have time to fix this. In the end, someone needs to bundle
>>this into an rpm, and make it so that when you build the rpm, it
>>isn't messing with permissions on the build machine.
>
>
> I see the problem. It would be easy to provide a patch in the
> source RPM to fix it. WCSlib uses GNU make, it that allowed in a
> Linux distribution?
>
Yes, and is usually installed on machines that have compilers on them.
>
>
>>> > - numarray. Not part of the standard Python install, but RPM is
>>>
>>>>available at least for Fedora Core.
>>>
>
>>In our DAG repository, could be done easily.
>
>
> Fedora Core 5 also has an RPM for it.
>
>
>>Did you mean graphviz or graphwiz. We have graphviz already in the
>>release, I can't find graphwiz
>
>
> Yes, I meant graphviz. But as Andy pointed out, it is only a
> build requirement to generate documenation, it is no a install
> requirement.
>
Well, graphviz is in the distribution, so there's no problem with it.
>
>>Commenting on below. There is a python-numeric that is available in
>>the dag repository. I have also created a numpy rpm, that is
>>currently in our contrib area. I've seen some packages want
>>python-numberic, and some want numpy. I personally have not used
>>either. is python-numeric the same as numpy? Should I include one
>>or the other, or both?
>
>
> This is indeed very confusing. There are three numeric packages
> for Python which mostly share the same API.
>
> - Numeric is the original and mostly comes from LLNL.
>
> - Numarray is a rewrite tuned for large arrays and comes from the
> people who manage Hubble Telescipe
>
> - Numpy is another re-write being done by company under contract.
>
So it sounds like, I could put in all three, and whichever ones users
want, they can install.
> As far as HippoDraw is concerned, it has been tested with both Numeric
> and numarray and con be configured to compile for either. There's
> only one #ifdef in one file that makes the switch.
>
> The work I would need to do before the HippoDraw RPM spec file is
> ready for a Linux distribution is for its configure script to look for
> external software in /usr/lib as well as /usr/local/lib. Then the
> external packages boost, cfitsio, Numeric/numarray, etc would be found
> correctly. Of course this could be put into the RPM spec file in the
> configure command of the build step.
>
> This still leaves the question for people who want to use HippoDraw
> with the ROOT files. I only need 5 libraries from ROOT and there
> corresponding header files. Perhaps if there was a bare bones ROOT
> rpm available? It might be that 90% of ROOT users use only 10% of
> ROOT anyway?
That might be true, but I really have no idea. I'll have to ask some of
it's users and developers if we'd be able to make a trimmed down version.
Troy
--
__________________________________________________
Troy Dawson [log in to unmask] (630)840-6468
Fermilab ComputingDivision/CSS CSI Group
__________________________________________________
|