octave-maintainers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Distributing unstable Windows binaries


From: Tatsuro MATSUOKA
Subject: Re: Distributing unstable Windows binaries
Date: Tue, 15 Jul 2014 18:35:16 +0900 (JST)

>From: sergey plotnikov 
>To: octave-maintainers 
>Date: 2014/7/15, Tue 00:38
>Subject: Re: Distributing unstable Windows binaries
> 
>
>On 14/07/14 16:49, Carnë Draug wrote:
>
>On 13 July 2014 14:10, Philip Nienhuis <address@hidden> wrote: 
>>Hi, Markus Bergholz and I were discussing putting up unstable Windows binaries
(3.9.0+ & 4.1.0+) on mxeoctave.osuv.de.
Before proceeding I'd like to have the opinions of other Octave developers
on this. My thoughts: - Using mxe-octave I regularly build binary Windows 
installers anyway from
the gui-release and default branches (currently 3.9.0+ & 4.1.0+ combined
into one installer). Usually soon after a merge from stable -> gui-release
-> default.
I use those binaries at work (always a good test), rather than stable Octave
versions. - The unstable versions offer features (and fixes) not yet in stable 
that
(some/many) users might need. Until now Linux users had a bit of an
advantage here: even with mxe-octave the treshold for building unstable
(Windows, or OSX) Octave versions is still significantly higher than for
Linux users. - I could also distribute "unstable" OF package versions (in my 
case the io
package) this way. - Some increase in bug reports / issues / support requests 
etc can be
expected. Is this a bad thing? - User expectation needs attention. I was 
thinking of an Octave prompt along
the lines of:
"Octave <version> development snapshot <date> - use at own risk!\n>>"
...plus maybe some additions to the readme. Currently my mxe-octave build tree 
is somewhat outdated and a bit messy as I
have many personal mods; so I think my current unstable Windows binaries are
less suited for distribution. But now that mxe-octave is more or less
feature-complete (i.e., Ghostscript has been added) I have a good reason to
upgrade :-)  and it's easier for me to supply the mxe changes to comply with
the GPL. Thought, opinions? 
>>I don't think this is a very good idea and might actually be dangerous
for Octave image. No matter how many warnings you give to the users,
they will just ignore them, and rush to the very last release. Then
they will complain it's crap when it crashes. That's the whole point
of having versions, we know it's not good enough to release and for
general use. So it's not released. There's not even distribution of this for 
Linux, which would be easier
to set up. Even the Octave group in launchpad [1], which used to have
a PPA for the testing "releases", seems to now be down to the stable
versions only [2]. Carnë [1] https://launchpad.net/~octave [2] 
https://launchpad.net/~octave/+archive/ubuntu/stable
>But in this case we may simply limit access to those binaries. So that only 
>those who need it for particular purpose (developers) may download them.
>
>I don't want to say Octave-image-danger is not a problem, but if a
      binary is not distributed officially it's not really an image of
      Octave, but just a state of development version. Almost anyone may
      compile default branch on Linux. And if I see it right this
      doesn't harm Octave image, but just helps to spot potential
      problems. As far as i understand Philip, he just wanted to provide
      something similar to compilation of a default branch, but for
      Windows.
>
>When compiling on our own, some of us may merely overlook a
      critical changeset needed to (cross)compile decently working
      version of default branch. But if the compilation is done
      regularly by experienced person which knows most of pitfalls this
      problem is not so critical. I'm not proposing everyday snapshots,
      but at least more or less regular "correct" builds suitable forOctave 
testing.
>
>Sergey
>


For the case of gnuplot I and Prof. Kakuto regularly make the snapshot of 
development binaries for windows.
I do not hear the blame for the case of gnuplot.  Of course, situation is 
different between gnuplot and octave. 

In the case of LibreOffice, there published Nightly Builds of developer 
snapshots.  

I have been trying mxe native build during a week with help by John Donoghue 
but no success. 
Of course, I could build octave on Ubuntu without so much effort.  
However, such people is not much in those who usually use windows.
Considering the current case, the periodical made the snapshot is worthy, I 
think.

Tatsuro



reply via email to

[Prev in Thread] Current Thread [Next in Thread]