Re: [MoM] mmseqs2
Hi Shayan,
On Mon, Jul 22, 2019 at 10:56:29PM +0100, Shayan Doust wrote:
> I have received an email from another MMseqs2 developer with some
> information and proposed changes.
>
> Firstly, it came to my attention that MMseqs2 was developed with 64-bit
> systems in mind and may break when using x86, as within the code, there
> is the assumption that sizeof(size_t) == 8. Would changing arch=any to
> strictly match that of amd64 or x86_64 be an issue?
I'd recommend something like
Architecture: any-amd64 any-ppc64 any-ia64 arm64 mips64el alpha sparc64
(if there is no simple way to enable 32bit).
> I was also informed of a regression / test suite[1] which seems to be
> nice. Maybe I will end up completely replacing the current test unit to
> use this instead of my patched one.
I'm fine with any enhancement you intend to implement.
> There were a few licensing issues which I will rectify. MMseqs2 uses
> source files that are adapted from these projects[2] and are spread all
> around the codebase. Should I include these into the copyright file or
> not, as I am not sure as to how much these have been adapted or look
> different.
I'm not sure whether I understand the question. Any copyright information
belongs to debian/copyright in the given format.
Thanks again for your repeated work it
Andreas.
> [1]: https://bitbucket.org/martin_steinegger/mmseqs-benchmark/src/master/
> [2]:
> https://github.com/soedinglab/mmseqs2/wiki#external-libraries-used-in-mmseqs2
>
> On 22/07/2019 13:02, Shayan Doust wrote:
> > Hello again,
> >
> > I Should have combined this with the previous email.
> >
> >> In any case we can just backport what is available in Debian testing.
> >> If a package has landed there (we will see when ftpmaster will accept
> >> the current package and what upstream will be released at that time)
> >> we can ask whether this is a sensible target for backporting.
> >
> > Sounds good. I guess the best way to tell is to keep an eye on this [1].
> >
> >> I admit I'm very positively impressed by your energy. You are the
> >> most productive MoM student and you have obviously a lot of fun to
> >> challenge the learning curve with an extreme speed. I like this a
> >> lot!
> >
> > Thanks :) ! I hate giving a bad impression and not getting anything
> > done, so this is great to hear. Once I'm into something, I don't really
> > lose interest nor sustainability.
> >
> >> I'm *very* lazy with life chat techniques. I think most of the time
> >> these drain more time than they give you. I'm fine with appointments
> >> to meet in IRC and I will do my best. When beeing at DebConf I even
> >> have a IRC window open. But I prefer mail for serious information
> >> since its archived and you can seek in it. But if others are fine
> >> with IRC this channel is not restricted to automatic messages at all.
> >
> > Ahh alright. Sometimes a realtime communication can be good for any
> > virtual meetings where lots of communication are to be thrown or
> > anything that doesn't really need a long term archive.
> >
> > Also Andrius:
> >
> >> Great! The upstream should be very grafetul to you for hunting these
> >> problems down. It would be best if they could incorporate your
> >> patches.
> >
> > :). I assume their tests were quickly made and rough - not intended for
> > any test suites. I don't remember if they even use any CI. I doubt it,
> > but it's good that it's all pretty much fixed.
> >
> >> This should not be a problem. You may look around to find someone in
> >> your region via db.debian.org. Other option would be to check the
> >> database each time you have a conference or vacation somewhere. This
> >> is how I got my signatures :)
> >
> > Ahh that's great, thanks! I'll keep an eye on anything that is
> > relatively close once I move out in under a month.
> >
> > Best regards,
> > Shayan Doust
> >
> > [1]: https://ftp-master.debian.org/new.html
> >
> > On 22/07/2019 10:27, Andreas Tille wrote:
> >> Hi again,
> >>
> >> On Mon, Jul 22, 2019 at 01:50:44AM +0100, Shayan Doust wrote:
> >>> Just more of a closing email.
> >>>
> >>>> I backport on request only.
> >>>
> >>> Ahh. That's great to know. I think the upstream developer(s) *might* be
> >>> interested, but it's too early to even tell what release they'd want.
> >>
> >> In any case we can just backport what is available in Debian testing.
> >> If a package has landed there (we will see when ftpmaster will accept
> >> the current package and what upstream will be released at that time) we
> >> can ask whether this is a sensible target for backporting.
> >>
> >>> I've rectified the test binaries by patching. I have excluded taxonomy
> >>> for the sole reason that taxonomy requires the software to download
> >>> roughly 20 GB of compressed data from the internet, and as a result use
> >>> a lot more space when it self decompresses as well as processing power.
> >>
> >> Everything can be included into the next upload once the package in new
> >> was accepted. Its hard to predict when this will be. We have seen
> >> periods of days, weeks and monthes. Waiting for your work to become
> >> effective is some of the patience draining things in Debian sometimes.
> >>
> >>> Seems like with this package completion, I'm on a hunt to find something
> >>> to do now.
> >>
> >> I admit I'm very positively impressed by your energy. You are the most
> >> productive MoM student and you have obviously a lot of fun to challenge
> >> the learning curve with an extreme speed. I like this a lot!
> >>
> >>> Additionally, I have been lurking in the debian med irc with znc and
> >>> haven't seen any activity. Does this irc channel get used, or is it more
> >>> of an informational channel of monitoring med-team commits?
> >>
> >> I'm *very* lazy with life chat techniques. I think most of the time
> >> these drain more time than they give you. I'm fine with appointments
> >> to meet in IRC and I will do my best. When beeing at DebConf I even
> >> have a IRC window open. But I prefer mail for serious information
> >> since its archived and you can seek in it. But if others are fine
> >> with IRC this channel is not restricted to automatic messages at all.
> >>
> >> Thanks again
> >>
> >> Andreas.
> >>
> >>> Best regards,
> >>> Shayan Doust
> >>>
> >>> On 21/07/2019 00:00, Andreas Tille wrote:
> >>>> Hi Shayan,
> >>>>
> >>>> On Sat, Jul 20, 2019 at 11:14:24PM +0100, Shayan Doust wrote:
> >>>>> Thank you for the upload! :-)
> >>>>
> >>>> Ahhhhh! Great!
> >>>>
> >>>>> Quick question. Would it be worth the effort backporting, or is that
> >>>>> usually based on factors like demand and interest from users,
> >>>>> developers, etc?
> >>>>
> >>>> I backport on request only.
> >>>>
> >>>>> Additionally, I've managed to fix one of the few tests which I was
> >>>>> initially unable to include within the unit test. It takes a bit of
> >>>>> effort to run valgrind and to try figure out where SIGSEGV was tripped.
> >>>>> My aim is to have maximum functionality testing coverage with this very
> >>>>> soon.
> >>>>
> >>>> Very cool!
> >>>>
> >>>>> This package was hugely beneficial in learning. I've been in touch with
> >>>>> the upstream developer and have learned about their periodic releases,
> >>>>> etc... which means that I can focus on keeping upstream happy by
> >>>>> incorporating future releases within the package.
> >>>>>
> >>>>>> I am amazed by your speed and quality - two packages in twenty days! I
> >>>>>> hope to see you becoming independent Debian developer soon.
> >>>>>
> >>>>> Sadly, I'm right before entering university so I'm still young and
> >>>>> travelling for key signing / etc is not a viable option yet. I won't be
> >>>>> surprised if I was the youngest in the team :-)
> >>>>
> >>>> The younger the better. ;-)
> >>>>
> >>>>> Many thanks to you and Andreas for your efforts,
> >>>>> Shayan Doust
> >>>>
> >>>> I love to support newcomers as best as possible.
> >>>>
> >>>> Kind regards
> >>>>
> >>>> Andreas.
> >>>>
> >>>
> >>
> >>
> >>
> >>
> >
>
--
http://fam-tille.de
Reply to: