Re: RFA: Support Windows extended error numbers in safe_strerror

[Daniel Jacobowitz <drow at false dot org>]

On Thu, Feb 09, 2006 at 03:40:17PM +0100, Mark Kettenis wrote:
> Trust me, I would not be proposing these changes just to make you do extra
> work.  Yes, if things are changed the way I describe, that would make them
> acceptable to me.

Thank you.  I will work on it.

> > We need to decide this now.  I, and I think Mark Mitchell also, are
> > heartily sick of contributing these patches, spending days revising
> > them to satisfy other developers, and then being told the port
> > shouldn't exist at all.  There's a flamewar every time we post one;
> > that's mighty good incentive to leave the port broken.
> 
> Sorry, but from Mark's first batch of patches I got the impression that
> Codesourcery was contributing well-tested code and no further changes would
> be necessary.  I certainly had the impression that we'd get MinGW almost
> for free.  Turns out now that this was too good to be true :(.
> 
> If I had known all of this beforehand, I'd probably put up more resistance
> at that time.  It seems that nearly all global maintainers still think MinGW
> support is worth the additional cost, but I had to make sure we realise that
> there *is* a cost.  I'm sorry if this frustrated you and Mark; I can
> certainly image it would frsutrate me if I were in a similar situation.
> I would certainly appreciate it if you would address my concerns.

I answered this a year ago, at our _last_ flamewar about the subject. 
I distinctly remember apologizing for being unable to predict the
future.

Mark did say that we were done.  I wish he hadn't; all of us on this
list know that maintenance for a platform is rarely done.  I didn't
notice his statement at the time, just reading the archives today.
He was done with what he needed for that project, which didn't include
serial, and predated some bug reports.

You're imposing what I consider a ridiculously high quality threshold
on this code.  I'm willing to oblige you, because you did so much work
cleaning up obsolete host quirks, because it is cleaner, and in hopes
that it will decrease the shouting density on this list.  But I want to
make my opinion perfectly clear - I don't think the changes you've
asked for make a _significant_ improvement in the cleanliness of GDB,
and I think they will be either irrelevant or mildly detrimental to the
future maintainability of the MinGW port, e.g. by leading to code
duplication if we needed to override anything more complex than
safe_strerror.

Anyway, this is the last message I intend to post about the general
issue of Windows support.  We agreed in April 2005 to support it,
we argued it out again just now and agreed again to support it.
I will be a happy man if the question of "should we be doing this
at all" never comes up again in my vicinity.  Thanks.

-- 
Daniel Jacobowitz
CodeSourcery