[csw-maintainers] /usr/{local,share} references

[Dagobert Michelsen]

Hi Phil,

Am 08.02.2011 um 23:01 schrieb Philip Brown:
> On Tue, Feb 8, 2011 at 1:20 PM, Dagobert Michelsen <dam at opencsw.org> wrote:
>> Am 08.02.2011 um 21:40 schrieb Philip Brown:
>>> A silly little example being
>>> 
>>> for f in `find . -type f | xargs grep -l /usr/local` ; do
>>> sed 's:/usr/local:/opt/csw:g' $f >$f.cswfix; mv $f.cswfix $f
>>> done
>>> 
>>> Feel free to rewrite in your preferred language of perl, python, or
>>> whatever floats your boat :)
>> 
>> You are not serious about this, right?
> 
> Well, I would certainly prefer that people manually inspect and replace.
> But for those people whining that this takes too much effort, the
> above provides an almost effortless alternative.

As I outlined below it is worth than doing nothing.

>> I inspected a couple of my package
>> and every occurence needs to be verified manually. There are even
>> scripts looking for configuration stuff which should keep the include
>> to /usr/local but must have /opt/csw manually added and the check
>> overridden.
> 
> I would be interested to hear more details, about specific cases where
> the above script did something "bad". I think that all maintainers
> would benefit from this information.

Think of "The default location for package installation is /usr/local if
you don't specify --prefix". Changing this to /opt/csw would be plain
wrong. There are lots of other cases.

>> This is even more bad than keeping /usr/local because it says "I have
>> looked at it and judged to replace it" instead of "well, I kept the
>> defaults, if they are wrong I look again".
> 
> But if they dont even look, how are they going to come to the point of
> examining the defaults, knowing they are wrong, and "looking again"?

Because there is a bug report. And because noone has tainted the
defauls they are clearly visible.

> In some cases, having wrong defaults, does not "break" the app. but
> you silently lose functionality.
> The maintainer may  never notice that unless they explicitly examine
> all /usr/local occurrences.
> Which is what I'm in favor of, but others are opposed to.

Not at all. I doubted the use of the /usr/local | /usr/share checks,
but after inspecting yaz I came to the opinion that the checks are
useful in general and really increase quality. But at the same time
I came to the opinion the blindly replacing occurrences makes the
situation worse by inadvertently changing valid occurrences.

>> BTW, I found in yaz a real positive where tons of stuff is correctly
>> auto-configured with the exception of one path to /usr/local/share/...
>> and a comment /* Don't know how to make this with autoconf */ :-(
>> And, yes, this is important to 100% functionality.
> 
> That's nice that yaz "does the right thing" for 99%.
> Nothing wrong with csw-specific patching for that "one path" that they
> dont know how to autoconf either though.

Exactly my point. The check was good as it revealed the issue.
But if you change pathes do it consciously or not at all and delay
inspection for some later time, but not a half-hearted job.


Best regards

  -- Dago