On Wed, 23 Jul 2014, Ignacio Diaz-Emparanza wrote:
On 22/07/14 19:04, Logan Kelly wrote:
>> This is a follow-up to Logan Kelly's question at
>>
http://lists.wfu.edu/pipermail/gretl-devel/2014-July/005194.html . While
>> installing and running more than one version of gretl on a given machine
>> is
>> not something we'd generally expect people to do, it's possible if you
>> take
>> some care, and I thought it might be useful to spell out what's what.
>>
>> The specific case Logan mentioned has the last gretl release (1.9.90)
>> installed
>> alongside a snapshot. To get this working you could first install today's
>> snapshot, then either use regedit to change the gretldir registry entry to
>> point back to the gretl 1.9.90 installation, or re-install 1.9.90 to
>> overwrite the
>> registry. Then 1.9.90 will use the registry entry, and the snapshot will
>> (or
>> should) ignore the registry and find its own path.
>>
>> Some testing of this would be nice!
>>
> OK so I have both the snapshot and the 1.9.90 installed. Installed the
> snapshot first.
>
> I have two scripts A and B. A works in 1.9.90 but not the snapshot (gnuplot
> 5rc thing) and B works in snapshot but not in 1.9.90 (new feature thing).
> Both A and B run and fail as expected. Hurray!
I think you should change the Gnuplot path in
Tools->Preferences->Programs->Command to launch gnuplot
Did you do it?
There is no way to set the gnuplot path via the GUI in the gretl package
for Windows: it's assumed that we'll always be using the wgnuplot.exe in
the gretl installation directory (and the ability to choose the gnuplot
path would only invite problems).
But if I'm reading Logan correctly, he's saying that the new "gretldir"
mechanism is working, such that each installed version of gretl is calling
the appropriate, matching version of gnuplot.
However, I should also mention that the gnuplot 4-versus-5 incompatibiity
which both Logan and Artur have noted, and which I explained in
http://lists.wfu.edu/pipermail/gretl-users/2014-July/010226.html , has now
been softened in gnuplot CVS. Specifically, the command "set style
increment user" (which we've been using in gnuplot scripts generated by
gretl) now works again. So I'm thinking of rebuilding gnuplot for the
Windows packages before release. This will help a bit, in that older
gnuplot scripts should still work with the new gretl/new gnuplot
combination.
Allin