Re: docstrings in Guile!

["Greg J. Badros" <gjb at cs dot washington dot edu>]

Valentin Kamyshenko <val@kamysh.materials.kiev.ua> writes:

> >>>>> "Greg" == Greg J Badros <gjb@cs.washington.edu> writes:
> 
> Greg> Michael Livshin <mlivshin@bigfoot.com> writes:
> >> "Greg J. Badros" <gjb@cs.washington.edu> writes:
> >> 
> >> hmmm...  isn't it backwards-compatible?
> 
> Greg> What, guile-doc-snarf?  I suppose it is, though that is an
> Greg> implementation detail.  The calling sequence from the makefile
> Greg> is a little different, though (guile-doc-snarf demultiplexes its
> Greg> two output streams by opening the files itself, guile-snarf just
> Greg> writes its output to stdout and the makefile is responsible for
> Greg> redirecting that into the .x file).
> Pleeease, do not forget to include into the distribution both
> versions. I mean, guile-snarf (without "-doc-") was used by authors
> of many packages, based on guile. We _must_ have the "exact" backward
> compatibility here.

Well, once I document that guile-doc-snarf is backward compatible as
part of its interface, then it's not just an implementation detail;  As
mentioned, the execution sequence is different so Makefiles would have
to be changed to use guile-doc-snarf as it stands now.... but it'd be
easy enough to make guile-doc-snarf write .x output to stdout and the
.doc output directly into the file.  I'll probably switch to that
technique and print the errors to stderr where they should be going
anyway.

Greg