[gentoo-user] sci-physics/root slotting?

[gentoo-user] sci-physics/root slotting?

[daid kahl]

Hello,

For anyone who uses the data analysis framework ROOT developed mainly
at CERN (sorry, I didn't name it 'root'), I can imagine that slotting
would be an extremely useful feature.

Anyone who doesn't use or know root, but has experience or opinions on
how or when slots should be used, your feedback would be appreciated,
too.

All kinds of macros and libraries will go bonkers on the wrong
version, and there are even binary linux executables out there that
want certain versions installed.  It's a pity I didn't think of this a
few years ago.

It occurred to me tonight that adding slotting should be easy and very
useful.  I've never added slotting, but I'm already running root as a
local overlay because I need root-5.20 (with patches from 5.22+ to
keep my system otherwise current!), and so I'm going to give it a go
at least for myself.  Actually, few of the people I know in physics
run Gentoo, but they also complain about root versions.  Maybe I at
least have an argument that would compel anyone to switch to Gentoo if
we get slots running.

If this seems like a good feature request, I'll put a modified ebuild
on bugzilla for all present root versions after I can test it (may
take a few days, since root isn't a quick compile and I have physics
to do).

Obviously, if this happens, I need to consider a bugzilla feature
request on eselect as well, or make eselect-root.  Never touched that
source either, but I will, if nothing else, be hacking together an
eselect-root shell script for myself.  If I manage to modify the
eselect source, then at least I might be more deserving of the
ChangeLog credit, since I don't think making the number for slot
non-zero in a few ebuilds really qualifies as real work.  Might also
need the multislot use flag.

Anyone out there interested in this or have some feedback for me?  If
it's only me, I'm hesitant to submit it to bugzilla (no sooner than
next week), but I'll be running it as local ebuilds as slots from here
on out.

Regards,
daid

[gentoo-user] Re: sci-physics/root slotting?

[daid kahl]

> Hello,
>
> For anyone who uses the data analysis framework ROOT developed mainly
> at CERN (sorry, I didn't name it 'root'), I can imagine that slotting
> would be an extremely useful feature.

> It occurred to me tonight that adding slotting should be easy and very
> useful.

Not true!  The slotting isn't supported upstream, and it requires a
lot of modification to all the ebuilds.  As I have no experience, it's
also not clear to me if I slot everything including things in etc and
man pages.  I can see arguments both ways.

> If this seems like a good feature request, I'll put a modified ebuild
> on bugzilla for all present root versions after I can test it (may
> take a few days, since root isn't a quick compile and I have physics
> to do).

I will consider this a serious project, but progress will not be
quick, since I have a lot to learn.  Honestly it might make more sense
to hit the upstream mailing lists first and see what they think about
slotting, since the implementation is probably a lot easier from the
source than doing crazy crap in the ebuilds like moving things in,
say, usr/include/root to usr/include/root-${SLOT} and every other
directory, not to mention recursively hitting the same thing on bin/
and setting up symlinks and a module to handle switching all them.
For example, suppose a new version gives me a new binary.  Now the
eselect module needs to know it can't switch that symlink on the old
version; I'm sure this has been done for other packages, but like I
said, I need to learn about it first.

> Obviously, if this happens, I need to consider a bugzilla feature
> request on eselect as well, or make eselect-root.

This will need some learning as well on my part, as above.

~daid

Re: [gentoo-user] Re: sci-physics/root slotting?

[Roger Mason]

Hello Daid,

daid kahl <daidxor@gmail.com> writes:

>> Hello,
>>
>> For anyone who uses the data analysis framework ROOT developed mainly
>> at CERN (sorry, I didn't name it 'root'), I can imagine that slotting
>> would be an extremely useful feature.
>
>> It occurred to me tonight that adding slotting should be easy and very
>> useful.

I use ROOT a fair amount.  I have not personally run into issues with
sensitivity of macros to ROOT version, at least not that I know of.  I
am running 5.22-r2 on an X86_64 system and 5.25/02 on a ppc.  The latter
was compiled outside portage because ROOT is masked with missing keyword
on that system.

If there is something I can do to help out then please let me know.  I
am not much of a programmer, but at the very least I should be able to
run tests.  I have a few old machines around that could be fired up and
used as test beds.

I am tied up with end of term stuff for the next week or so, but then my
time will be a bit more flexible.

Cheers,
Roger