Re: docbook: fix fatal error in linux/input.h

[Randy Dunlap <rdunlap@xenotime.net>]

On Tue, 26 Oct 2010 14:07:26 -0700 Randy Dunlap wrote:

> On Tue, 26 Oct 2010 13:59:57 -0700 Randy Dunlap wrote:
> 
> > On Tue, 26 Oct 2010 18:15:10 -0200 Mauro Carvalho Chehab wrote:
> > 
> > > Em 26-10-2010 17:17, Randy Dunlap escreveu:
> > > > On Tue, 26 Oct 2010 17:13:01 -0200 Mauro Carvalho Chehab wrote:
> > > > 
> > > >> Em 26-10-2010 16:45, Randy Dunlap escreveu:
> > > >>> Hi,
> > > >>>
> > > >>> Recent commit 8613e4c2872a87cc309a42de2c7091744dc54d0e:
> > > >>> Author: Mauro Carvalho Chehab <mchehab@redhat.com>
> > > >>> Date:   Thu Sep 9 21:54:22 2010 -0700
> > > >>>     Input: add support for large scancodes
> > > >>>
> > > >>> causes a fatal error in docbook generation:
> > > >>>
> > > >>> linux-2.6.36-git8/DOC1/Documentation/DocBook/device-drivers.xml:40690: parser error : StartTag: invalid element name
> > > >>> #define INPUT_KEYMAP_BY_INDEX	(1 << 0)
> > > >>>                              	    ^
> > > >>>
> > > >>> I think that this message comes from xsltproc, but I'm not positive about that.
> > > >>> I also don't know of a really good fix for it.  However, I did find 2 ways to
> > > >>> work around the error:
> > > >>>
> > > >>> a/ move the #define value to after the end of the struct input_keymap_entry, like:
> > > >>>
> > > >>> /* flags bits: */
> > > >>> #define INPUT_KEYMAP_BY_INDEX	(1 << 0)
> > > >>>
> > > >>> or
> > > >>> b/ don't use the "<< 0" (can leave the #define where it is in this case):
> > > >>> #define INPUT_KEYMAP_BY_INDEX	(1)
> > > >>>
> > > >>> I have tested both of these patches and they work OK.
> > > >>> Are you OK with either of them?  or want to choose one?
> > > >>
> > > >>
> > > >> Hmm... probably, the first one is not correct.
> > > > 
> > > > Why not?  All it does is move the line to after the end of the struct.
> > > > 
> > > >> I suspect that the right way is to use something like:
> > > >>
> > > >> 	#define INPUT_KEYMAP_BY_INDEX	(1 &lt;&lt; 0)
> > > >>
> > > >> Could you please test if this would fix the issue?
> > > > 
> > > > You are suggesting putting that into include/linux/input.h ??
> > > > 
> > > > That would make sense if some tool converted "<<" to "&lt;&lt;".
> > > > And that may be where the problem is, but I don't know those tools.
> > > 
> > > 
> > > Hmm... such rule is already at DocBook/Makefile:
> > > 
> > > %.xml: %.c
> > >         @$($(quiet)gen_xml)
> > >         @(                            \
> > > 	   echo "<programlisting>";   \
> > >            expand --tabs=8 < $< |     \
> > >            sed -e "s/&/\\&amp;/g"     \
> > >                -e "s/</\\&lt;/g"      \
> > >                -e "s/>/\\&gt;/g";     \
> > >            echo "</programlisting>")  > $@
> > > 
> > > Not sure why this didn't work. It should have replaced the < from the c file.
> > > Maybe some patch broke Makefile?
> > 
> > Maybe that rule needs to apply to .h files also.
> > 
> > Would
> > 
> > %.xml: %.[hc]
> > 
> > work?
> 
> Doesn't work like that or as a separate rule for
> %.xml: %.h
> 
> The gen_xml rule isn't firing at all AFAICT.

That rule was used for procfs_exmaple.c in 2.6.33 and earlier.
It's no longer used/needed (but I'm in no hurry to delete it either).

It appears that scripts/kernel-doc is the problem with not converting
<,> to &lt;, &gt; etc.  I'll look into that.


---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***