2011/3/8 Ben Walton <[email protected]>:
> Excerpts from Maciej Bliziński's message of Mon Mar 07 03:22:56 -0500 2011:
>
> Hi Maciej,
>
>> It is possible to fit a plugin architecture into the existing code,
>> if communication paths are provided.  I could imagine a bit of
>> Python code which would wrap plugins from a certain directory, and
>> took the responsibility of communicating with them.
>
> Yes, I it could be made to work.  That would more tightly couple
> checkpkg to GAR though (at least in the sense that checkpkg would
> expect to be told where plugins live).

I would need to verify that, but I think that checkpkg can discover
where it's being run from, which would allow to locate a plugin
subdirectory.

> If this were to be
> implemented, the packages themselves should carry the plugin code so
> that it could be verified outside of the GAR/build context.  Does that
> make sense to you?  (Note that I'm not really pushing for this, just
> discussing the idea.)

I wouldn't like checkpkg to execute code contained in the package.  A
broken package with broken code could for example damage the database
and the files I own on the buildfarm.  I'd be more inclined to provide
a way to include data, which could be later used for checking.  The
checking code would be in the repository / on the filesystem, and
would process input data from the package.  As far as code goes, I'm
for strict control of what is executed.

Maciej
_______________________________________________
maintainers mailing list
[email protected]
https://lists.opencsw.org/mailman/listinfo/maintainers
.:: This mailing list's archive is public. ::.

Reply via email to