5:56 p.m.
Hi,
I attempted to repackage my py4gretl stuff with a fresh cvs build (I
like the redesigned package manager BTW), but got an error claiming
mismatch between the public interface function name and the package
filename. But I never changed anything manually, the mismatch is
entirely due to the (automatically inserted) version number. So it would
be nice if that could be corrected before 1.6.2. And BTW the error
message did not show up in the po-file I think and should thus be added.
Also today I updated de.po in response to your call to translators.
And one trivial thing finally: '(int' is colored differently from ' int'
in a command script, looks like a slight parsing glitch.
Thanks,
Sven
8:15 p.m.
On Wed, 14 Mar 2007, Sven Schreiber wrote:
I attempted to repackage my py4gretl stuff with a fresh cvs
build (I like the redesigned package manager BTW), but got an
error claiming mismatch between the public interface function
name and the package filename. But I never changed anything
manually, the mismatch is entirely due to the (automatically
inserted) version number. So it would be nice if that could be
corrected before 1.6.2. And BTW the error message did not show
up in the po-file I think and should thus be added.
Thanks for the report. I'll look into that shortly.
Also today I updated de.po in response to your call to translators.
Thanks!
And one trivial thing finally: '(int' is colored differently
from ' int' in a command script, looks like a slight parsing
glitch.
Hmm, might be fixable, I'll take a look.
Allin.
4:01 p.m.
On Wed, 14 Mar 2007, Sven Schreiber wrote:
I attempted to repackage my py4gretl stuff with a fresh cvs build (I
like the redesigned package manager BTW), but got an error claiming
mismatch between the public interface function name and the package
filename.
This is now fixed in CVS.
And BTW the error message did not show up in the po-file I think
and should thus be added.
It's in the current de.po around line 8319.
Allin.
6:06 p.m.
Allin Cottrell schrieb:
On Wed, 14 Mar 2007, Sven Schreiber wrote:
> I attempted to repackage my py4gretl stuff with a fresh cvs build (I
> like the redesigned package manager BTW), but got an error claiming
> mismatch between the public interface function name and the package
> filename.
This is now fixed in CVS.
Does it really work for you? I just compiled latest CVS and still get
the error. Actually it also seems to fail when I don't try to have the
version number included.
It would be nice if somebody could check if they can package functions
with current CVS. Maybe my CVS copy is corrupted or something.
> And BTW the error message did not show up in the po-file I think and
> should thus be added.
It's in the current de.po around line 8319.
Indeed, and I can't reproduce that the message shows up in English.
(maybe another indication that something is broken on my side?)
Thanks,
Sven
6:24 p.m.
On Thu, 15 Mar 2007, Sven Schreiber wrote:
Allin Cottrell schrieb:
> This is now fixed in CVS.
Does it really work for you? I just compiled latest CVS and still get
the error. Actually it also seems to fail when I don't try to have the
version number included.
I'll take another look. Maybe it's a CVS issue, you're not truly
getting the latest code? I have gui2/fnsave.c at version 1.43.
>> And BTW the error message did not show up in the po-file I
think and
>> should thus be added.
>
> It's in the current de.po around line 8319.
Indeed, and I can't reproduce that the message shows up in English.
(maybe another indication that something is broken on my side?)
I'll run this under valgrind -- maybe there's some sort of memory
sloppiness going on.
Allin.
6:47 p.m.
Allin Cottrell schrieb:
On Thu, 15 Mar 2007, Sven Schreiber wrote:
> Does it really work for you? I just compiled latest CVS and still
get
> the error. Actually it also seems to fail when I don't try to have the
> version number included.
I'll take another look. Maybe it's a CVS issue, you're not truly
getting the latest code? I have gui2/fnsave.c at version 1.43.
Ahem, I think I compiled alright but forgot to install... Very sorry for
the noise, it works now, and I guess I'll leave the rest for tomorrow
when hopefully I can think straight again.
I'll run this under valgrind -- maybe there's some sort of memory
sloppiness going on.
Again, don't bother, everything is (and probably was) fine...
cheers,
sven
6:55 p.m.
On Thu, 15 Mar 2007, Sven Schreiber wrote:
Again, don't bother, everything is (and probably was) fine...
OK, that's good to hear. However, there are still some issues
with function packages that differ in their version numbers and
internal code details, but have the name of their public interface
in common -- e.g. "foo-1.0" and "foo-1.1", both with public
interface foo().
I'll try to sort these out before 1.6.2.
Allin.
2:31 p.m.
Allin Cottrell schrieb:
.
OK, that's good to hear. However, there are still some issues with
function packages that differ in their version numbers and internal code
details, but have the name of their public interface in common -- e.g.
"foo-1.0" and "foo-1.1", both with public interface foo().
I'll try to sort these out before 1.6.2.
I also just noticed some issues with saving the package after changing
some of the code. Sometimes the help text would not show up anymore in
the editor, and the drop-down menu for the included functions in the
package editor would disappear (after saving, closing the editor, and
opening the editor again). However, it does not happen every time (and
after restarting gretl everything is ok again). Anyway, it seems like a
good idea that any package developer keep her/his source scripts around,
separately from the package itself.
My first attempt to upload a package also failed. First I got an
"unspecified error" (in German), at the second attempt gretl just
crashed (after some network activity).
Unfortunately there was no output on the console, so I don't really have
any idea what was going wrong...
Any hints?
cheers,
Sven
p.s.: the first question for the gretl server login application cannot
be answered, as \sigma_i is never defined ;-)
12:17 a.m.
On Fri, 16 Mar 2007, Sven Schreiber wrote:
My first attempt to upload a package also failed. First I got an
"unspecified error" (in German), at the second attempt gretl just
crashed (after some network activity).
I think this should work now. The issue (or at least, one issue)
was that the py4gretl packages were much bigger than previous
tests, and they overflowed what Apache handles by default for the
http GET method. I have changed the gretl code to use http PUT
when uploading a function package (which makes sense anyway).
There may be some remaining issues relating to the length of the
py4gretl filenames, since this system was originally set up for
gretl databases, which all have small (<16 byte) names. But I
think I've now caught most of those.
Allin.
2:30 a.m.
On Mon, 19 Mar 2007, Allin Cottrell wrote:
I think this should work now. The issue (or at least, one issue) was
that
the py4gretl packages were much bigger than previous tests, and they
overflowed what Apache handles by default for the http GET method. I have
changed the gretl code to use http PUT when uploading a function package
(which makes sense anyway).
s/PUT/POST
Allin
7:33 a.m.
On Tue, March 20, 2007 03:30, Allin Cottrell wrote:
On Mon, 19 Mar 2007, Allin Cottrell wrote:
> I think this should work now. The issue (or at least, one issue) was
> that
> the py4gretl packages were much bigger than previous tests, and they
> overflowed what Apache handles by default for the http GET method. I
> have
> changed the gretl code to use http PUT when uploading a function
> package
> (which makes sense anyway).
s/PUT/POST
Another issue: I was trying to update the code for the HEGY test I wrote
a few months ago. In the new version, I used a "matrix *" parameter,
which is not contemplated by the DTD, so the upload failed. Allin, could
you please take care of this?
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
12:17 p.m.
On Tue, 20 Mar 2007, Riccardo (Jack) Lucchetti wrote:
Another issue: I was trying to update the code for the HEGY test I
wrote
a few months ago. In the new version, I used a "matrix *" parameter,
which is not contemplated by the DTD, so the upload failed. Allin, could
you please take care of this?
Will do.
Allin.
1:03 p.m.
New subject: yet another package saving bug (2)
On Tue, 20 Mar 2007, Riccardo (Jack) Lucchetti wrote:
Another issue: I was trying to update the code for the HEGY test
I wrote a few months ago. In the new version, I used a "matrix
*" parameter, which is not contemplated by the DTD, so the
upload failed.
Fixed, but this is a little tricky. I have given "pointer"
parameter types the following names for the purposes of XML
storage and DTD validation:
scalarref
seriesref
matrixref
If you define a function with the usual form, e.g. "matrix *M",
the type will automatically be written as "matrixref" in the
(XML) .gfn file. The read-back for this has required a small
change in gretl_func.c. This means that new-style .gfn files
using "*" parameters will require gretl 1.6.2.
Allin.
11:01 a.m.
Allin Cottrell schrieb:
There may be some remaining issues relating to the length of the
py4gretl filenames, since this system was originally set up for gretl
databases, which all have small (<16 byte) names. But I think I've now
caught most of those.
Well, do you want me to change the names? If so, now would probably be a
better time than later on. Or to put it differently: Should there be an
official lenght limit for package filenames?
thanks,
sven
12:19 p.m.
On Tue, 20 Mar 2007, Sven Schreiber wrote:
Allin Cottrell schrieb:
>
> There may be some remaining issues relating to the length of the
> py4gretl filenames, since this system was originally set up for gretl
> databases, which all have small (<16 byte) names. But I think I've now
> caught most of those.
Well, do you want me to change the names? If so, now would probably be a
better time than later on. Or to put it differently: Should there be an
official lenght limit for package filenames?
I don't think your filenames are outrageously long, but I think it
would be a good idea to have a limit. I'll do some more checking,
but I think 31 characters might be sensible (which would
accommodate the names you have used).
Allin.
6486
days inactive
6492
days old
14 comments
3 participants
participants (3)
-
Allin Cottrell -
Riccardo (Jack) Lucchetti -
Sven Schreiber