On 2007-09-25 18:11+0100 Andrew Ross wrote:
Sorry for my silence over the last couple of weeks, but I've been snowed
under at work.
No problem. I have a lot of calls on my time as well at the moment which is
why my current PLplot contributions are limited to extremely short projects
(such as testing changes done by others) that I can just squeeze in.
Plplot works fine with all 2.1.xx versions of octave I have tried. It
certainly works fine with 2.1.73 on Ubuntu feisty.
According to http://www.gnu.org/software/octave/download.html, 2.1.73 is now
considered to be the stable version of octave with no further changes
contemplated in the 2.1.x series so I am really glad that you have confirmed
our current octave bindings work for that case.
It does not work on
octave 2.9.x. I've tried 2.9.9 on Ubuntu feisty. This is a different
version to that tested by other users. More worryingly I seem to have
different errors to other reports on the list. Various things seem to
have changed which will require rewriting some of the octave parts of
the bindings. I have been pondering the best way of doing this in a
compatible manner. It might be cleanest to have different versions of the
bindings depending on the octave version. I am sure I can sort this, but
I'm not sure of the timeframe. It's personally important to me, so I
will try to devote some time to it.
Well, from the URL above the octave testing version has the remark (you
probably want this). So I suggest you designate our current interface
(which we know works for 2.1.73) as the octave stable interface and create
a modified octave interface called our octave testing interface that works
for the octave testing version (currently 2.9.14). Hopefully, that version
won't change too often, and there will be no changes required to our octave
testing interface when there is a change in the version number designated
as octave testing.
Of course this strategy means our octave testing interface _might_ not work
for early 2.9.x versions that are in distributions (e.g., 2.9.9 in Ubuntu
feisty). If that turns out to be the case, you can cover that off by
disabling the octave testing interface unless exactly the right version that
you have tested has been installed (usually by specifically downloading and
building it for those who want to be on the absolute cutting edge). This
strategy should take care of cutting edge octave users but not interfere
with access to PLplot for octave since presumably every recent distro will
have packages for octave stable = 2.1.73. However, an octave testing
interface would be nice to have as well since it will gradually lead to an
octave 3.0 interface for us when the 2.9.x series finally converges to 3.0.
There is also a psttf.c issue Alan reported to me recently which I would
like to look into before the release.
Apparently from your commit message remarks you have now fixed the
positioning issue. Thanks. Also, you do not confirm the segfault issue I
found (which may simply mean my pango stack made from source downloads and
builds does not include some key fixes that are in the distro versions or
has some incompatible libraries in the stack). So I think we should ignore
the segfault issue unless it shows up again when I update my system
(probably still at least another month from now) from Debian oldstable to
testing.
[out of order] How long are we looking at until the
release?
That's the big question. This morning I finished up the desired changes in
the files that create the website examples, but those rather extensive
changes still have to be tested by Hazen to make sure good-looking results
are produced on his platform.
All the other issues I mentioned are fixed, put off until after 5.8.0, or in
the would be nice category (e.g., more Ada examples). I believe the
octave testing interface is probably in the would be nice category as well
because of the current octave stable interface that will work as a good
alternative.
Therefore, here is what I suggest with regard to the release timing. Hazen,
once you have confirmed my recent changes produce good-looking web site
examples why don't you go ahead and set a release date for 5.8.0-RC1 that is
the earliest that is convenient for you (say this weekend or the next). Then
if Jerry can come through with more Ada examples or Andrew can put together
the proposed octave testing interface before that release date, then fine,
but otherwise those issues can be dealt with after the release of 5.8.0. (I
am assuming here that only minimal differences will be allowed between
5.8.0-RC1 and the 5.8.0. Ideally, those differences will be just the
version number changes. However, if some showstopper bug was found in the
week or so of testing from our users for 5.8.0-RC1, then obviously that fix
should go into 5.8.0 as well.)
Alan
__
Alan W. Irwin
Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria
