Re: deferred deallocation of local objects

[Chris B. Vetter]

On Tue, 14 Oct 2003 13:46:22 -0600
Adam Fedor <fedor@doc.com> wrote:
> On Tuesday, October 14, 2003, at 10:48 AM, Chris B. Vetter wrote:
[...]
> > FWIW, I can vouch for everything Alex said, especially the "It works
> > on my box"_TM and "your BSD must be broken" (with respect to the 
> > networking
> > stack for crying out loud!!! yea, very likely...) parts.
> I don't really like to respond to rants such as this, but I think that
> people on the sidelines of GNUstep might get the wrong impression.
> Free Software projects, by their very nature, depend a certain level
> of user responsibility, and it's not just because they are free.

I agree and that is _exactly_ why his occasional "What do you want? It's
working on my box"_TM approach is NOT the way to go.

> - It's not easy doing remote debugging without a lot of help from the 
> person who is having the problem.

No one is talking or even expecting anyone to do remote debugging. Alex
(and I) are talking about, for example, the now (in)famous commit of a
certain class with a missing /* at the head of the file ...

Checking that code works BEFORE actually commiting it to CVS is the
issue here.

> - The libraries (particularly gnustep-base) are in some places very 
> complex and it's very often that even minor changes break things in 
> other places.

True. Yet, no one can be experienced in everything. While someone may
have experience in security, another has experience in networking. Not
taking the advice or recommendation of someone who HAS experience in a
certain field is, pardon my French again, stupid.

> I'm very reluctant to change anything of importance in the base
> library because of this and I think it's amazing that Richard has kept
> up so well with the intricacies of the library. No one else has
> invested the knowledge and time understanding as much as is needed to
> keep the base library maintained.

Absolutely, and no one disputes that fact. Yet, see above. His responses
along the line of "it works on my box"_TM and "then your system must be
broken" annoy not only me.

You are right in saying that the occasional visitor to this list will
get a wrong impression by reading this thread, however, talk to people
who actually develop for or are actually trying to use GNUstep and they
can and will tell you that there are lots of problems being left
untouched (although pointed out) or are fixed in a ... weird way. There
is one application in particular that comes to mind that had to be
rewritten (at least partially) to work around bugs that were introduced
while fixing the very same problem.

I don't want to step on Richard's toes, far from it. But I have to agree
with Alex, that his quality control leaves a lot wanting...

-- 
Chris