[csw-maintainers] /usr/local references, and packages

[Ben Walton]

Excerpts from Philip Brown's message of Mon Jan 24 17:51:01 -0500 2011:

> Ben, what you said is incorrect. This is not some "new policy".  It
> has **always** been an issue. From day one. And I have always
> enforced it, whenever it came to my attention.  With the understood
> exception of "oh this is just documentation examples, so can be
> ignored".

Ok, it at least _feels_ new.  I don't recall seeing packages rejected
for this until very recently.  More strictly enforced?  

I don't disagree that in some cases references to things like
/usr/local are bad.  I do think that there are cases where it doesn't
detract from package quality.

> The only thing that has changed, is that now I guess the pre-release
> examination tools have gotten better at picking it up. Or perhaps
> more at discriminating between [this is in a doc file, but That is
> not].

Did you change this recently then?  Are you using new tools?

> The principle is so glaringly obvious, that it has never needed much
> explicit "writing up" until now.  "When reconfiguring software that
> defaults to /usr/local, to instead run under /opt/csw, replace all
> occurrences of '/usr/local' with '/opt/csw' " Claiming that I am
> enforcing a "non-existent standard", is unfathomable. This is not
> "new".

Again, I don't think the point of this is that what you're saying is
wrong, although I think you could argue that anything in share/doc is
fair game for an exclusion at the maintainers discretion.  What I'm
hearing is that people perceive this as a 'new rule' whether or not
you feel that it's new.

In the interest of turning this into productive discussion, lets lay
down the ground rules for this.  How do people feel about:

Files that contain references to /usr/local and are _not_ under
/opt/csw/share/doc must have a valid rationale for why they're ok or
the package must be altered to remove it.

Using the recent perl release as an example, it turned out to be a
harmless reference.  A note to this effect is, imo, sufficient to make
the exception.  If we later see /usr/local turn up in other .h files
the package could have a packaging bug filed against it and it would
be barred from batch until it's fixed or explained.

We can vote on this if people disagree with it.

Maciej, what is checkpkg doing in this regard currently?  Does it
check for this string at all?  If so, does it check every file or only
certain areas?

Thanks
-Ben
--
Ben Walton
Systems Programmer - CHASS
University of Toronto
C:416.407.5610 | W:416.978.4302