[rkward-devel] [rkward-users] Call for testing: RKWard 0.5.3-pre1 / Release schedule

[Thomas Friedrichsmeier]

Hi,

thanks for the good testing. I did not look at everything, yet, but to make a 
start:

On Saturday 17 April 2010, Prasenjit Kapat wrote:
> 1. File browser (Alt+1)
> 
> 1a. Single click and double click: shouldn't it honor the default KDE
> settings? Double clicking on a filename opens the files thrice (when
> KDE default=single click) or twice (when KDE default=double click),
> neither of which is correct.

Indeed. It's a bug in kdelibs, though, and it would be very hard to work 
around this. Does anybody have a KDElibs 4.4.x installation at hand? Could you 
check, whether the double-click behaviour is still the same, there?

> 1b. Can we save the view mode (Short view / Detailed view)? The "short
> view" seems almost unusable with large number of files and/or large
> filenames.

Implemented in SVN. Please test (another preview release will also follow some 
time next week).

> (I miss the List view from KDE3!)  (1b can be targeted after 0.5.3)

Added an icon for tree view (which is pretty similar, IIRC).

> 2. Shortcuts
> 
> 2a. (Although this is specific to my setup, but can occur to others as
> well.) Some of my codes are named blah.0.blahblah.r,
> blah.1.blahblah.r, ..... As a reproducible case, create two (or 6)
> script files: a0.R, a1.R (..,.a6.R) and open them. Now, Alt+0, Alt+1 (
> ..., Alt+6) become ambiguous ;) If this is tied to katepart, then at
> least having an option to disable script tab shortcuts will be useful!
> (For some reason this does not happen on KDE3's katepart) Sometime, at
> this stage, a crash may occur - I haven't been able to
> deterministically reproduce a crash.

Indeed. I see the problem, but don't know how to fix this, yet (other than to 
assign different shortcuts). BTW, the katepart is not to blame. It's the tab-
bar that assigns shortcuts to the tabs, apparently without taking existing 
shortcuts into account.

> 2b. In the absence of Previous (or Next) Window, hitting Alt+< (or
> Alt+>) inputs  "," (or ".") character into the script - which is
> obviously dangerous. Ideally, the key combination should be disabled
> in such situations.

I can see the problem, but don't know what to do about this, either. I suppose 
the problem could be side-stepped, if "previous" would wrap around to the most 
recent, and "next" to the first window, so that there is always something 
reaction to the shortcut. How does that sound?

> 2c. Window Left and Right are always disabled for some reason!

Indeed. Will investigate.

> 2d. "Run all" shortcut:
> Has anyone done the mistake of hitting Shift+F9 instead of Shift+F8
> (F9 and F8 are close enough)? My suggestion would be to not have any
> default shortcut for "Run all" and let the user set one, if he/she
> needs.

Hm, I use "Run all" a lot, personally, and that makes me reluctant to remove 
the shortcut. What does everybody else think / use?

> 3. make plugintests
> 
> 3a. Existence of R2HTML should be checked before starting any tests.
> Otherwise it is really not any useful - all tests are skipped!

You have a point, there...

> 3b. The default size of output buffers (Command log / R Console) are
> too small for plugintests' output. Can we call sink(file = ...) before
> and after running the tests? (Or will every plugin need to be modified
> for this?)

Should be possible, somehow. BTW, the tests are just a bunch of R scripts 
(make plugintests basically sources rkward/tests/all_plugins.R; see also 
http://sourceforge.net/apps/mediawiki/rkward/index.php?title=Automated_Plugin_Testing), 
so maybe you would like to take a look, yourself? 

> 3c. Anyway, ignoring the --skipped-- tests, here are the FAILed ones:
> 
> shapiro_wilk_test             match          MISMATCH       match
> (empty)  no             FAIL
> ad_test                       match          MISMATCH       match
> (empty)  no             FAIL
> 
> import_spss                   match          MISMATCH       MISMATCH
>     ERROR          FAIL
> import_stata                  match          MISMATCH       MISMATCH
>     ERROR          FAIL
> 
> linear_regression             match          MISMATCH       match
> (empty)  no             FAIL
> shapiro_wilk_test             match          MISMATCH       match
> (empty)  no             FAIL
> ad_test                       match          MISMATCH       match
> (empty)  no             FAIL
> 
> As Mat asked, what other "FAIL" output would be useful?

Depends on what you can make of the output yourself. ATM, a number of the 
failures have reasons that do not really relate to what is being tested, but 
rather to the test framework. For instance, the snippet Mat posted seems to be 
just a matter of a different type of quotes being printed, probably due to some 
locale-setting that we need to normalize before running the tests. So 
basically many failures are simply false alarams, and we'll just need to do 
some more tweaking like this here and there to shut them off.

If you can spot such patterns, there's no need to post the full diffs (but do 
report the failure). However, if unsure, post the summary of the failed tests 
(as you did, above), and also the diffs that are shown above the summary. 
Better to post too much info, than too little.

-

On another note: I meant to write about this, earlier, but forgot: I modified 
your solution for the CRAN mirror settings, as discussed some months ago. This 
removes the hard-coded list of mirrors, but also replaces the tcl-dialog with 
an "rkward native" dialog. What do you think of the current state?

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20100418/f3d80715/attachment.sig>