Bug#296863: libgnomedb FTBFS (#296863)

[Loïc Minier]

        Hi,

On sam, fév 26, 2005, Gustavo R. Montesino wrote:
> >  Just wanted to know whether you had time to investigate a bit on the
> >  libgnomedb FTBFS (#296863), I've heard you had a stuck package in NE=
W,
> >  but I don't know if it's related.
> I've taken a look in the report, and thought it's probably because of
> some change in gtk-doc-tools or related softwares (Jade?). 
> 
> As you said, I'm just waiting the new version of libgda (1.2.0) get out
> of the NEW queue (actually, for the one that is sitting there to be
> rejected, then upload a better one and wait...) to upload a newer
> libgnomedb, where things are working well.
> The "debian/ dir" for building the 1.2 packages can be found at
> svn.debian.org

 I wonder why a newer libgnomedb would be doing better, I'll have to
 rebuild libgda 1.2 and libgnomedb to understand that!


> If you're really wish to do so go ahead, and I'll try to make an upload
> if you send me a patch. However, I don't think this is really
> necessary.

 Here's a short summary of what I understand of this problem:
 - with libgnomedb-1.0.4/doc/C/tmpl and the source code, the gtk-doc
   makefile snipset builds a sgml directory,
 - the sgml directory should serve as input to jade for the generation
   of the html doc, but jade fails on this because of:
   /usr/bin/jade:../sgml/gnome-db-config.sgml:68:69:E: character ":" is
   not allowed in the value of attribute "LINKEND"

 The ":CAPS" snipsets is added to ids in SGML files by "gtkdoc-mkdb"
 when no lowercase letter is found, this is to avoid namespace collision
 in ids.  I've checked upstream gtk-doc, and "gtkdoc-mkdb" still acts
 like this.

 What I can't tell is whether jade is bugged or doesn't respect the SGML
 standard (jade only had an upload, but seems to have a long history of
 misbuilds), or whether it's gtk-doc assuming too much of the standard.

 I called openjade1.4devel on the SGML files, and openjade complained in
 the very same way, so I was first tempted to blame gtk-doc for using
 ":CAPS" in ids, but nsgmls didn't complain when ran on the same
 document.

 Finally, I checked ISO/IEC 10179, and found in 7.3.2 that ids are
 allowed to use ":", definitely a jade bug it seems, since the "linkend"
 attribute of the "link" element is often defined as of IDREF type.
 /usr/share/sgml/docbook/stylesheet/dsssl/modular/print/dblink.dsl
 confirms this definition of linkend in docbook-dsssl, which
 gtk-doc-tools uses.

 My personal conclusion is that the breakage shouldn't come up if
 upstream doesn't use all uppercased id refs, such as macros, but it is
 unlikely, or if jade is fixed.

 Now I'll have a look at newer upstream packages, and see how they fix
 the problem.

   Bye,

-- 
Loïc Minier <lool@dooz.org>
"Neutral President: I have no strong feelings one way or the other."