Re: NSPathUtilities Patch - 5 - NSPathUtilities.m

[Kazunobu Kuriyama]

Sheldon Gill wrote:

> On Wed, 21 Apr 2004 17:56, Kazunobu Kuriyama wrote:
>
> > Sheldon Gill wrote:
> > IIRC, some people were not in favor of setting
> > GNUSTEP_CONFIGURATION_FILE to
> > the fixed path /etc/GNUstep/GNUstep.conf. Did you reached some
> > agreement with
> > them?
>
>
> The definition of the fixed path is to be a compile time configuration 
> option.

NSPathUtilities.m *is* not written so. (Your code is not arranged so that
config.h or can overwrite it; it's hard coded, anyway.)

> > > Provision for "Platform Support" which causes
> > > NSSearchPathForDirectoriesInDomains() to return additional paths for
> > > different platforms. This makes it easier to find applications or other
> > > resources.
> >
> > If the order of search paths has been "corrected", that is likely to 
> > break
> > other places in both GNUstep and third-parties' applications, where that
> > function is called. Perhaps, that is a right thing to do. Nonetheless, I
> > strongly feel you should state this fact more loudly and consult with 
> > those
> > who are in charge of version control on making a transition period for
> > that. No one is pleased with this sort of change without any warning.
>
>
> How is it likely to break other places? All "Platform Support" does is 
> add
> additional paths at the end of the array. It doesn't change things in a
> damaging way. Things that call it get a couple of additional places to 
> search
> in for whatever they're looking for and will only go looking there if it
> hasn't already been found. Nothing breaks.

For instance, look at the case where -reverseObjectEnumerator is used.  Can
you still guarantee nothing is broken?

> The order of search paths of GNUstep is currently incorrect. There is 
> a bug
> which is noted in the source.

I'd like to make sure of this: Did you fix it or leave it as is?

> > Why didn't you pay any attention to backwards compatibility at all?
>
>
> If I didn't pay any attention to backwards compatibility, why did I 
> bring up
> the issue?
>
> I believe I've payed considerable consideration to backwards 
> compatibility.
> Only where there are reasonable advantages have made changes which effect
> compatibility. These proposed changes require only a recompile. No
> modifications need to be made to application source at all.

Only a recompile...  I feel you need to explain why other people have to 
spend
their time for your cause.  It's all about the issue on backwards 
compatibility!
(Or do I use the term "backwards compatiblity" differently from the 
meaning in
your mind?)

> If the enumeration order isn't changed then even recompiling isn't 
> needed.
> I've put forward the change now because there are other changes 
> elsewhere in
> GNUstep which are likely to require it anyway.

Could you think about the incremental approach that some people already
suggested you?

Regards,
- Kazunobu Kuriyama