[Live-demo] OSGeo Live v3 Build 5

[Hamish]

Alex wrote:
> Hmm, so overall the logs looked pretty good on this last
> build. A slight mixup in that postgres got renamed as a file
> but not in the main.sh so it didn't run and pgrouting didn't
> either because postgis was missing.(Fixed)

sorry 'bout that, I thought that install_list was already
activated.
 
> The other really odd one, is that a bunch of R packages
> failed to build because gfortran was missing. I added it
> explicitly to install_R now which I think will fix it but it
> worked fine in Alpha 4 so I'm wondering which other script
> pulled the rug out.(Fixed I think)

It was probably setting install-Recommends to false which lost
it, gfortran was never explicitly installed before AFAICT.
In alpha3 it doesn't get installed until the start of
install_R.sh, as an "extra" package.

fwiw, build-essentials is already installed by setup.sh.


> Maybe we need a better policy about apt-get autoremove,
> should we be doing it and the end of every script and
> ensuring all requirements are in each script.

IMHO all modules that make a mess should clean up after
themselves, even if it means some double downloads.
It's good practice and it makes later cleanup less blunt/risky.


fwiw, I could mirror the gpsdrive .debs locally & it would be
a little bit quicker, but I like having a working example of how
to use "debuild" in the script pool.


> get double downloads (unless it's still in the cache)

we're not running 'apt-get clean' until setdown.sh, so they
should still be around.


regards,
Hamish