Re: Guile 1.5.4 beta ready for testing.

[Keith Wright]

> From: Chris Cramer <address@hidden>
> 
> On Tue, Nov 06, 2001 at 12:45:48AM -0500, Keith Wright wrote:
> > I can't believe I am reading this; it's so 1950's.  Which do you
> > hear more often, complaints that somebody's program doesn't work or
> > begging for 50% faster crashes?  I am not sure what ``debugging''
> > does, but how long can it take to print an error message and die?
> > Nobody is asking for logging every procedure call, but it ought
> > to be possible to remember what file you are loading and print
> > the name when you catch the error.
> 
> Maybe I used the wrong argument, but the slowness of Guile startup just
> really, really annoys me. It also annoys me that people are often telling
> me that they don't use Guile because it's slow.

I was probably more harsh than I intended.  Apologies if that seemed
personal.

> But I don't think we're fundamentally disagreeing about this particular
> issue. Getting useful information when a program dies is certainly more
> important than how fast the program runs.

Agree.

> Just remembering the file and printing the name when catching an
> error (or providing other moderately useful information) shouldn't add
> anything significant to the runtime.  But AFAIK turning on debugging does
> a lot more than that. What we should do is improve the error messages
> and I don't think debugging mode is necessary for that.

It could be that too many things are currently wrapped up in one
flag, some of which don't even need to be optional.  I have no
emotional attachment to that particular flag name, but do think
it is important to give error messages that help find the problem.
Often this entails no run-time overhead, but just printing the
information that is already somewhere in core in a well structured
way.

-- 
     -- Keith Wright  <address@hidden>

Programmer in Chief, Free Computer Shop <http://www.free-comp-shop.com>
         ---  Food, Shelter, Source code.  ---