[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#66023: PROPOSAL] Re: Shared libs vs. plugins.



On Sat, Apr 28, 2001 at 11:36:41PM -0500, Manoj Srivastava wrote:
>  >> +        to by third party executables (binaries of other packages),
>  >> +        should be installed in the subdirectories of the
>  Richard>                           ^^^
> 
>  Richard> I would drop that "the", to make clear that packages can create
>  Richard> their own subdirectory for the plugins.
> 
> 	Umm. Perhaps one3 should specify ``the subdirectories of
>  "/usr/lib/<package>/: directory''?

Wouldn't that imply that one has to make /usr/lib/<package>/<something>/
subdirectory, or several of them?

I agree with what Richard said, just drop the "the".

>  Richard> Anyway, do you really mean _all_ the rules?  I would expect
>  Richard> that they should still be compiled with -fPIC, for the same
>  Richard> reasons as shared libraries -- memory pages with relocatable
>  Richard> code can otherwise not be shared between processes.  (Please
>  Richard> correct me if plugins are not normally relocatable.)
> 
>  Richard> Also, stripping with --strip-unneeded still seems like a good idea.
> 
> 	Hmm. I assumed that since these were internal details of the
>  package, one need not make policy about them, since internal detail
>  should be left to the maintainers to implement. How about adding what
>  you said above as an informative footnote? That way, every maintainer
>  that reads policy shall no it is a good idea, and why, but policy
>  shall not intrude into internal matters of the package.
> 
> 	I definitely agree that what you say is a darned good idea,
>  but I am not convinced that we need policy about that. 

I think we should amend the proposal with a note that the plugins aren't
exempt from stripping -- this is my oversight. Also, that is a "should"
rule, so if there are exceptions, this won't cause serious bugs. I'm don't
know about -fPIC, though, that is a "must" rule. A footnote would be fine,
I suppose.

-- 
Digital Electronic Being Intended for Assassination and Nullification



Reply to: