Now, I am going to ask you that question once more and if you say “no” I am going to shoot you through the head. (MPFC 33)
Why the question always is “what it means” and never “does it mean anything”?
It can mean for instance Yet Another Gtk DOCcumenter. Or Yeti's Alternative Gtk-DOC. Or Yeast Assault Group's Dance Of Contradiction.
Is my pronunciation really relevant? I pronounce it yuck-dots [jŭk'dŏʦ], if you insist.
2.4 or newer.
SimpleParse, the Python parser. Not many distributions include it, I'm considering bundling it with yagdoc.
For GObject inspection, GLib is of course required, including working GModule – and at this moment GThread too – plus compiler, linker, the usual lot. If your package builds a library with GObjects it requires them anyway.
I wonder where you got the “instead” part, definitely not by looking at who contributes to gtk-doc.
gtk-doc codebase has been spaghettizing by various special cases and afterthought features for a long time and it badly needs a rewrite. What's worse, it has to support all the crap in ancient Gnome documentation. Writing a new program works around the compatibility problem nicely.
This question does not make any sense.
inspect
does not work with --as-needed
Does it still surprise you --as-needed
breaks things? You'd
better get used to it. And what question is this anyway?
On the positive side, I plan to compensate by using
--no-as-needed
when linking inspect
. If libtool is
used, with_gnu_ld
is set accordingly in configure. It can be
harder when the package does not use libtool.
This does not sound much like a question either.
Given my approach to the design, it's amazingly fast. For instance it
could be considerably sped up just by using plain tuple
s instead
of Token
s. But I prefer to be able to read the code. Unless
the time taken by gtkdoc-scan
starts being comparable with the
DocBook → HTML step, there's no need to worry about its speed.
Isn't index.sgml
and foo.devhelp2
enough?
Unfortunately, both are crap – or their design is shortsighted if you like euphemisms:
index.sgml
maps symbols names after conversion to
SGML identifiers. The SGMLized id is already present in the
href
attribute after #
, making the id
attribute completely redundant. On the other hand the actual symbol name is
not anywhere. So instead of
<ANCHOR symbol="GtkMenu:tearoff-state" href="gtk/GtkMenu.html#GtkMenu--tearoff-state">
it contains
<ANCHOR id="GtkMenu--tearoff-state" href="gtk/GtkMenu.html#GtkMenu--tearoff-state">
foo.devhelp2
maps titles to ids,
therefore instead of
<keyword type="property" name="GtkMenu:tearoff-state" link="GtkMenu.html#GtkMenu--tearoff-state"/>
it contains
<keyword type="" name="The :tearoff-state property" link="GtkMenu.html#GtkMenu--tearoff-state"/>
Ah, did I mention it does not index enum
values at all?