Re: Changes to package installer to handle rpm builds

[David Bateman]

Orion Poplawski wrote:
> David Bateman wrote:
>> Orion Poplawski wrote:
>>> This will take a fair bit of work, but I think we need a way to pass the
>>> installer a directory containing a built package and have it install
>>> that without running configure/make.
>>>
>>
>> Orion,
>>
>> This is essentially what the "package" target in the octave-forge
>> packages Makefile does. It makes a binary package that can be installed
>> correctly without rerunning configure or make. Why does this
>> functionality need to be migrated to the Octave pkg command itself?
>>
>> Regards
>> David
> 
> I sent this before realizing this.  However, it is still inefficient to
> tar up a directory just to have the octave installer untar it again. Not
> critical though...
> 
> 


Orion,

Frankly, I agree but it makes pkg simpler. My original (well second
iteration) RPMs for octave-forge proposed in the thread

http://www.nabble.com/Re%3A-OctavePackaging-tf3726123.html#a10427922

Used entirely Octave pkg and the SPEC file itself rather than an
external Makefile. Though Thomas Weber made it pretty clear his
preference for a "./configure; make; make install" process and what is
in the octave-forge package Makefiles is my best attempt to meet that.
Though even that retars the directory.. After that I simplified that
process into the pkg("build","<package>.tar.gz"), and the reason I don't
use that in the octave-forge package Makefile was that Thomas Weber also
insisted on his need to be able to pass flags to make (think make -j2),
and I felt it was easy and cleaner to keep the configure/make out of
Octave to get that effect than to try and allow options to be passed to
configure and make through the pkg command..

D.