[Cin] LV2_blacklist.txt & Debian packaging improvements contributions

Wed May 29 16:16:59 CEST 2019

Hi Phyllis,

Le 2019-05-29 01:50, Phyllis Smith a écrit :
>> Now that things as settled around, and now that I'm good with the
>> website French translation, I'm taking back the packaging of Cin-GG
>> for
>> my audio-linux debian-based distribution (and maybe later, a
>> potential
>> inclusion into Debian mainstream, but... one step after another).
> 
> When you do get this packaged with audio-linux LibraZiK, it would be
> good to add this to cinelerra-gg.org [1] same as AV Linux and Bodhi
> Linux.

OK, I'll let you know when it'll leave the testing phase and enter the 
users' repo.

>> Doing that, I found a bunch of LV2 plugins which are avoided Cin to
>> be launched and then, which needs to be added to the LV2_blacklist.txt
>> file (after I did that locally, Cin is launching fine). Those 
>> following
>> lines are the culprits:
>> 
>> file:///usr/lib/lv2/MonoEffect.ingen/MonoEffect.ttl
>> file:///usr/lib/lv2/MonoInstrument.ingen/MonoInstrument.ttl
>> file:///usr/lib/lv2/StereoEffect.ingen/StereoEffect.ttl
>> file:///usr/lib/lv2/StereoInstrument.ingen/StereoInstrument.ttl
>> http://example.org/raffo
>> http://www.wodgod.com/newtonator/1.0
>> https://sami.boukortt.com/plugins/intersect#Intersect
>> https://sami.boukortt.com/plugins/intersect#SymmetricDifference
>> https://sami.boukortt.com/plugins/intersect#Upmix
> 
> The latest GIT checkin a few minutes ago, includes the updated
> lv2_blacklist.txt with the first 4 lines BUT I could not test it.  You
> already did so it should be fine.

OK.

> HOWEVER, I do not understand the last 5 URL lines. None of them
> worked for me and I am confused as to why they were included.

I'm unsure what you mean here while writing they didn't worked ?
I'm assuming you're meaning that you tried to enter the url in a 
browser.
If this assumption is wrong, please correct me.

Whether or not they are opening a web page when entered in a browser 
isn't very relevant for our purpose.
To the LV2 stack, they are used as URIs, meaning lilv is using those URI 
(starting with http, https, file,...) as an unique identifier for a 
plugin. They are identifier for the following LV2 plugins:
- https://github.com/nicoroulet/moog/
- http://newtonator.sourceforge.net/
- and https://github.com/sboukortt/intersect-lv2

I've got them installed here on my LibraZiK-2, and if I'm launching cin 
without blacklisting those URIs, then the LV2 check step fails which 
lead to cin not being launched.

That's a big issue since I'm can still patch the LV2_blacklist.txt for 
my LibraZiK-2 package and I'll be good.
That said I'd strongly recommend them being added to the cin repo 
LV2_blacklist.txt because it's likely to be a blocker (read: unable to 
launch cin) for anyone on any system with one of those plugins 
installed.

>> Also, please, find attached an improved version for the
>> ./blds/debian/control file which contains those
>> improvements/changes:
> 
> At the end of the month, GG is really pressed for time doing all of
> the distros.  He is not very good at Debian system
> packaging/maintenance -- we work on Fedora.  Today he is trying to jam
> in some changes that got put off while working on the update of
> OpenGL.  We will save your "control" file and work on looking more
> carefully at your suggested improvements.  He did take the time to
> evaluate it, found the duplicate libraries that your control file
> pointed out and the libusb missing one and alphabetized all of the
> dependencies to make it easier to find a specific one.

OK, nice.

> The reason for
> not listing them one per line despite the fact that it is more
> human-readable is that just like cpu-s, memory, and disks, the SCREEN
> IS A RESOURCE -- most people do not think of that as such, but when
> you are working with several windows on several program routines
> simultaneously, you need space. So he really needs to see fewer lines
> to get to the one he needs to see.  Also, time is a big factor and
> when he does all of the monthly builds, he uses _xen_ and he has to
> edit a bunch of individual files before starting to get the builds to
> work and wants to quickly find the lines he has to modify in each.

OK, fair enough.

> Their are some real experts on package control for all of the distros
> -- in the commercial world where he used to work, there was a
> "builder" who took care of all of this type of stuff.  We are using
> bare bones / old stuff just to manage.  Maybe some day, a Cinelerra
> builder will show up and take care of all of this!

I don't want to be looking arrogant enough to consider myself as a 
"debian packaging expert", but just letting you know that I've got >200 
debian packages that I'm working on for my LibraZiK-2 project, plus I'm 
contributing to a few Debian packages in the official Debian repo 
where/when I can.

>> -- why is inkscape needed for building ? It sounds like a mistake at
>> a first glance.
> 
> There is a plugin called SVG for Inkscape so that is why it is needed.

I'm not 100% sure of this here. Let me rephrase it my thought:
I'm questionning myself if: 1) inkscape is need to build Cin, or 2)
if inkscape is need to run Cin.

If you tell me that Inkscape is needed to *build* the cin SVG plugin, 
then OK we're fine, I get it.
But if you're telling me that inkscape is needed to run the cin SVG 
plugin when using cin, then there is a modification to do in the cin 
debian packaging script since *build-dependencies* aren't considered the 
same from the debian/control point of view.

Please let me know.

>> -- why is e2fsprogs needed ?
> 
> We did not find the exact reason -- i think it is for mount bluray - I
> will see if I can find out for sure

OK, let me know when you find out (with the same as above regarding: 
*build* dependency or *runtime* dependency?)

> but at any rate he does not want to take it out until we are sure.

Neither I do.

>> -- same for linux-firmware, why is that needed for building ?
> 
> The main reason for this is supposed to be OpenGL + the TV tuner _
> hardware.  I will look at this again too since you say it is not even
> in the Debian repository anymore.

OK, thanks. Same as above (*build* or *runtime* dependency? )

>> - The "Standards-Version" number is currently defined to
>> "5.1.20190430".
>> It is not supposed to be the program version number (which is
>> defined in
>> the debian/changelog file), but a number matching a Debian standard.
>> If
>> you're OK with that, I would set it up to 3.9.8 which is the
>> standard
>> for Debian Stretch (current stable - Debian 9) and which will work
>> as
>> well for the next stable (Buster - Debian 10).
> 
> Every month when gg does a build for Debian, it errors out UNLESS he
> change the Standards-Version in the control file -- so we are
> surprised that he could put this at 3.9.8 and it would work, month
> after month.  Again, this may be because of xen.  Also, this is why he
> leaves this line at the beginning of the file -- to make it easy to
> find to change it.

Please, not that this isn't a big deal from my side since I'm changing 
it when packaging it for LZK-2.

Something sounds to be strange here. You can find more info here:
https://www.debian.org/doc/debian-policy/ch-controlfields.html#s-f-standards-version

> Thanks for all of your input.  If I find out anything more, I will let
> you know.

Thank you too,
Olivier

-- 
Site web : https://librazik.tuxfamily.org/
Donation : https://liberapay.com/LibraZiK/
Diaspora : 
https://framasphere.org/people/8c184af0c9450134f6682a0000053625
Mastodon : https://mastodon.xyz/@LibraZiK