[rkward-devel] Observations on the rkward Windows port

[Thomas Friedrichsmeier]

Hi Ben,

thanks for testing. I have not looked at your script in much detail, yet, but 
I'll write about some observations I've made, so I don't forget about them, 
and perhaps others can build on them.

On Friday 14 August 2009, Ben Goodrich wrote:
> This just seems to be a
> dependency problem. I was able to get around it by putting a separate
> copy of pcre3.dll into KDE/bin and renaming it libpcre.dll but there is
> probably a way to do that automatically.

This sounds like an issue in the kde win installer. I'll take a look, later.

> In using it, rkward does not crash that much. I did observe an odd
> problem that if you run many lines of code that take a long time to
> execute, when it is finished the main rkward window disappears but
> rkward consumes 100% of the CPU resources and I had to kill the process.

Confirmed on a similar setup. Did not yet attach a debugger. Possibly a problem 
in handling the large amounts of output that were generated, or possibly a 
side-effect of some earlier bugs triggered by your script.

> It can crash when certain examples are run, but most of the time they
> are not 100% reproducible. In particular, I sometimes had crashes with
> the examples from "anova.mlm", "check.options", "close.screen", and
> "erase.screen" but only when run with all the other examples in the
> script. If you run those examples individually, they seem to be okay.

I noted that anova.mlm, and check.options both produce an error (also in a 
plain R 2.9.0 session on windows). Possibly the crash is not so much "caused" 
by these examples, but rather it's that the script stops here (so it might be 
the same bug as above).

For close.screen and erase.screen, I've observed, that it does not produce any 
error when run individually. However, after I interrupted your script at some 
point, and tried those, then, they caused errors (but no crashes). That makes 
it look, like something is going wrong earlier, somewhere. The crash *might* 
once again be the same issue about long-running commands.

> Maybe in the future a script like this can be run before each release on
> all supported platforms to check for problems? The script is pretty easy
> to expand or customize for different platforms.

As Meik has already written, we are currently working on similar features. 
Your script looks like a nice addition to that (although it probably makes 
sense to keep this somewhat separate due to the large amount of time this 
requires).

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20090814/06d52216/attachment.sig>