Re: gnuplot 4.8?

[sfeam <sf...@us...>]

On Sunday, June 29, 2014 03:44:23 PM Allin Cottrell wrote:
> I understand that 5.0 is supposed to be the release where the
> gnuplot developers are free to make some backward-incompatible
> changes. I've no problem with that, but I wonder if there's any
> thought to produce a gnuplot 4.8?

The 4.6 development source is a separate branch in CVS.
The 4.7 (later 5.0) branch split off from it and gradually diverged.
Most bugfixes have been applied to both branches, but they have 
diverged enough that in some cases the "fix" for a problem in 4.6
is limited to the existence of a better implementation in 5.0.

There will be at least one more patchlevel release 4.6.6.
My crystal ball does not tell me at what rate fixable bugs will continue
to be reported againt 4.6 as opposed to 5.0.  Other incremental
releases could eventually appear, but I don't see a good rationale for
starting a 4.8  series.   If a fix/change is compatible then it could go into
4.6; if not then it belongs in 5.0. 

A brief listing of changes accumulated since the last release is
kept at the top of the NEWS files:

NEWS
New features, changes and fixes in gnuplot version 4.6.6
========================================================

* NEW linetype keyword "nodraw" can be used to draw only the points in "with lp"
* NEW plot option to "skip N" lines at start of an ascii data file
* NEW 'set fit prescale' normalized fit parameters before M-L refinement
* NEW update svg terminal to grey out the key entry when a plot is toggled off
* CHANGE Accept "with image pixels" as a synonym for "with image failsafe"
* CHANGE return NaN if a requested numerical data value fins a string instead
* CHANGE Consume only one space following the font name in an enhanced test string
* FIX Faster recovery from outboard server gnuplot_qt being killed
* FIX get rid of O(N^2) memory allocation for string data in long input lines
* FIX large integers in iteration spec could cause overflow end condition check
* FIX object fillcolors should be consistent with the color of current linetypes
* FIX LFS support on 64bit platforms (not backported for 32bit platforms)
* FIX timecolumn() applied to non-axis data reports an error rather than faulting
* FIX clipping could fail on integer overflow
* FIX segfault resulting from strcol(N) applied to empty field in a *.csv file
* FIX adjustment of key size to accommodate long key title
* FIX treat data value read as "NaN" the same as we would "1/0"
* FIX handling of events triggered by closing the qt plot window
* FIX iteration failure due to integer overflow
* FIX clip r axis tics to current plot boundary
* FIX logscale cb axis with volatile data
 
	Ethan


> I get the impression -- admittedly, without studying the code base
> in detail -- that CVS gnuplot has accumulated a number of helpful
> fixes along with some nice new options. My notion of a possible 4.8
> is basically 5.0 with the backward-incompatible elements reverted. I
> guess it would be trivial to revert some things (the default color
> scheme, the default for text in respect of enhanced versus
> non-enhanced) but maybe less trivial to revert others.

> Anyway, just wondering. The "use case" is that some of us who employ
> gnuplot as ancillary to data-analysis software would find it very
> nice to have a version that's both backward-compatible -- the
> important thing here is that scripts designed for 4.4 and 4.6 will
> still work -- and up-to-date with all bug-fixes, and buildable for
> win64 out of the box.