On Tue, Mar 12, 2013 at 9:25 AM, Chris Nighswonger < cnighswon...@foundations.edu> wrote:
> I double checked and .Inline and friends have correct perms. Clearing > everything below .Inline shows that the module is rebuilt as expected on > the next call to the CGI. > > I've tried BUILD_NOISY and the other Inline switches which would seem to > be helpful for debug, but with little success at obtaining useful > information. I'm guessing some of that may be largely dependent upon how > the author of Inline::Python has implemented them. > > It seems there should be some way to convince inline::Python to fess up to > any errors it is encountering along the way. > Perhaps I should mention this as well: The CGI scripts executed by miniserv.pl via an eval of a code block which sets up the environment and calls up the actual CGI via 'do.' See https://github.com/webmin/webmin/blob/master/miniserv.pl#L2337 for the exact code. I suspected the forced package change on line 2339 might be causing some scoping problems or some such, so I did this: root@biblios:/usr/share/webmin/pykota (0.02_pre_alpha *)# perl -e 'package pyktoa; do "PyKotaConfig.cgi"; die $@ if ($@);' And get this: Can't locate object method "new" via package "PykotaConfig" (perhaps you forgot to load "PykotaConfig"?) at PyKotaConfig.cgi line 33. This is exactly the same error I receive when running the same script as CGI. BUT... commenting out line 2339 in miniserve.pl did not fix the problem. So I suspect there is more to it than that. Kind Regards, Chris
