[csw-maintainers] Runpath menagerie for review

[Philip Brown]

On Thu, Mar 25, 2010 at 3:09 AM, Maciej (Matchek) Blizinski
<maciej at opencsw.org> wrote:
> 2010/3/23 Philip Brown <phil at bolthole.com>:
>> stuff starting with /optcsw is mostly acceptible.
>
> What about rpath orthography?  Would you like checkpkg to throw errors
> on the following ones?
>
> '/opt/csw/lib/.'  (as opposed to '/opt/csw/lib')
> '/opt/csw/X11/lib/' (as opposed to '/opt/csw/X11/lib')

I'd say yes

>
> Next question: specific architecture entries, for instance:
>
> '/opt/csw/postgresql/lib/sparcv9'
>
> Shouldn't this be '/opt/csw/postgresql/lib/$ISALIST' instead?

Hmmm... in the more general case, yes. however, since there is one and
only one "64bit" arch for sparc... not really.



> Next question: What about $ORIGIN?  In some cases it makes, sense, for
> instance if a 32-bit binary looks into '$ORIGIN/../lib', I guess
> that's fine.  But if it's a 64-bit binary in /opt/csw/bin/sparcv9 and
> the corresponding library is in /opt/csw/lib/sparcv9, the
> '$ORIGIN/../lib' entry won't help much.

Oooh, good point.
except what happens if it was an isaexec'd binary? that owuld be the
only exception I might think of, and its probably an invalid
exception.


>
> Bad rpath: '$ORIGIN'

actualy that can be quite okay, if its an RPATH in a library.


> Bad rpath: '/opt/csw/lib/sparcv8plus+vis'

If it is in an executable specifically compiled for that, then it
might not strictly speaking be "wrong".

ie: if it was in /opt/csw/bin/sparcv8plus+vis/executable


> Bad rpath: '/opt/csw/mysql4//lib/mysql'

meh.. mysql is a very individual case, so that could be let slip. it
will probably never be recopmiled anyway :)

> Bad rpath: '/usr/ccs/lib'

this is not bad.


> Bad rpath: '/usr/ccs/lib/sparcv9'

ditto.


> Bad rpath: '/usr/dt/lib'

not bad.


> Bad rpath: '/usr/lib'
> Bad rpath: '/usr/lib/sparcv9'

ditto.

> Bad rpath: '/usr/openwin/lib'


not bad.