crash
by Marcin Błażejowski
Hi,
simple script:
----------------------
set verbose off
open australia.gdt
series Y = lpau
list X = e2 lpus le
ols Y X --simple
COEFF = $model.coeff
STDER = $model.stderr
# This works:
eval COEFF ./ STDER
#... but this does not:
eval $model.coeff ./ $model.stderr
----------------------
GDB sayes: "double free or corruption (fasttop)" and Valgrind said:
Invalid free() / delete / delete[] / realloc()
==16926== at 0x48369AB: free (vg_replace_malloc.c:530)
==16926== by 0x5553129: free_node.part.48 (geneval.c:366)
==16926== by 0x55532E7: free_tree (geneval.c:420)
==16926== by 0x5553305: free_tree (geneval.c:422)
==16926== by 0x55532E7: free_tree (geneval.c:420)
==16926== by 0x557E7A4: gen_cleanup (geneval.c:18278)
==16926== by 0x559BCD2: generate (genmain.c:816)
==16926== by 0x56691DE: gretl_cmd_exec (interact.c:2930)
==16926== by 0x1EFB63: gui_exec_line (library.c:10315)
==16926== by 0x1F1667: execute_script (library.c:9437)
==16926== by 0x1F268F: run_native_script (library.c:8250)
==16926== by 0x1F268F: do_run_script (library.c:8457)
Marcin
--
Marcin Błażejowski
6 years
modprint not accepting string array in bundle
by Sven Schreiber
Hi,
not sure if it's a bug:
<hansl>
strings sarr = defarray("a", "b")
bundle b = defbundle("names", sarr)
matrix coeffmat = {0, 1; 0, 0.5}
modprint coeffmat sarr # works
modprint coeffmat b.names # fails
</hansl>
cheers
sven
6 years
obsolete "const matrix *" occurrences in the guide
by Sven Schreiber
Hi,
following up on a very old message (3 ys ago) I found about 7 cases
where in the guide a function header has an argument construct like
"const matrix *X". I believe this is obsolete, because if you want to
change X inside the function it mustn't be "const", and if you want to
avoid the copying then "const" is already doing that.
Should I change those occurrences?
(Haven't checked the hansl guide.)
cheers
sven
6 years, 1 month
new series appears as child
by Artur Tarassow
Hi,
I am not sure the following is as intended:
<hansl>
clear
set verbose off
nulldata 10
setobs 1 1 --time-series
series y = seq(1,10)
series fnaive = y(-1)
</hansl>
The series "fnaive" appears as a separate series but as a "child" of y
(as would all leads/ lags of y automatically named y* or y_*) rather
than a separate entry at the highest level in the main window. I thought
that only leads/lags of "y" would be shown as "child" series.
This happens when using latest git.
Artur
6 years, 1 month
GTK broadway support in gretl?
by Sven Schreiber
Hi all,
yet another "funny" question from me about other possibilities to run
gretl: I've read that there is a module called "broadway" in the GTK
framework which in principle allows the graphical interface of a program
(based on GTK) to appear in a HTML5-enabled (thus: any modern) browser.
Is it possible to build gretl with this broadway support? (Or is it
already activated?) On Linux? On Windows?
thanks
sven
6 years, 1 month
string-valued series
by Riccardo (Jack) Lucchetti
Hi all,
I'm working with a panel dataset that includes a few string-valued series,
and I'm finding a few things annoying, so I'm sending this message to the
list with a two-fold purpose: (a) to hear you guys' opinions and see of
there's antything I'm missing and (b) as a reminder to myself to work on
these things as soon as possible.
- when you compute lags of a discrete variable, the "discrete" flag is not
propagated (I think it should); can anyone think of similar cases?
- similar to the above: lagging a string-valued variable does not preserve
labels. Example:
<hansl>
nulldata 6
setobs 1 1 --special-time-series
l = defarray("foo", "bar", "baz")
series x = ceil(uniform()*3)
stringify(x, l)
lags 1 x
print x x_1 -o
</hansl>
produces
<output>
x x_1
1 bar
2 baz 2
3 baz 3
4 foo 3
5 bar 1
6 bar 2
</output>
of course you can do "stringify(x_1, l)", but it's sort of annoying.
- printing out an array of strings requires a loop if it has more than 9
elements. In some cases, this is VERY inconvenient. I see two way to go
around this: either we introduce a "set" variable, which replaces the
hard-wired limit at 10 we have now (something like "set arrayprint 20") or
introduce an option to the print command ("--full" or similar). What do
you guys prefer?
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
6 years, 1 month
no equation display in "wls" command ref
by Sven Schreiber
Hi,
I'm not seeing the intended equation in the 'wls' reference, i.e. I
believe something must be wrong in doc/commands/gretl_commands_en.xml in
the part:
<equation status="display"
tex="\[R^2 = 1 - \frac{\rm ESS}{\rm WTSS}\]"
ascii="R^2 = 1 - ESS / WTSS"
graphic="wlsr2"/>
(line 14688...)
This is 2019a.
cheers
sven
6 years, 1 month
