[util/macros] factorize tests for xmlto

[Gaetan Nadon]

On Tue, 2010-01-05 at 09:13 +0100, Matthieu Herrb wrote:

> On Mon, Jan 04, 2010 at 07:29:51PM -0500, Gaetan Nadon wrote:
> > 
> > Let's see if we can communicate what this macro does.
> > 
> > # XORG_WITH_XMLTO
> > # -------------------
> > # Minimum version: 1.5.0
> > #
> > # Check for the availability of xmlto, a front-end to the xsl toolchain.
> > # Set output variable XMLTO to the full path of the xmlto program found.
> > # Allows user to override the value for XMLTO 
> > # Define an Automake conditional HAVE_XMLTO
> > # Provide a feature "--with-xmlto" where values can be:
> > # auto:	set HAVE_XMLTO=1 if xmlto program is present
> > # yes:	set HAVE_XMLTO=1 if xmlto program is present
> > # no:	set HAVE_XMLTO=0
> > #
> > ]) # XORG_WITH_XMLTO
> > 
> > 
> > I am wondering if we need the AC_ARG_VAR to supply a path to xmlto. I
> > noticed only document kind of tools have this (e.g. asciidic, groff).
> > All other tools in the toolchain (a total of about 32) are expected to
> > be properly installed and in the PATH.
> 
> I have no idea on this. Iirc the autoconf documentation recommends that
> AC_ARG_VAR is used with any variable used by AC_PATH_PROG, and this 
> is how it is currently done in exiting uses cmlt.

I wasn't aware of this recommendation. 

> > 
> > The value of the XMLTO variable should not be unset when user says "no".
> > The path to the tool and the decision of not using are separate issues.
> > It's normal to have xmlto and not wanting to build the documentation.
> > The Automake HAVE_XMLTO conditional is the variable to use for this
> > decision, not a blank tool path. In addition, automake caches the value
> > of XMLTO, so there could be trouble ahead.
> 
> If I understand things correclty, unsetting XMLTO is only local to the
> configure script. It may be a bad coding trick to tell that we want
> to ignore its value, but it won't affect the value seen by make later on.
> 

This will be something to watch and test for. From what I read, the
value of XMLTO is cached and and saved during a reconfiguration. This is
an added behaviour when declared in AC_ARG_VAR. Aside from
implementation, do you see any problem in leaving XMLTO with a correct
value when builder said 'no'? The HAVE_XMLTO should be used as a
switch. 
Also there could be another line somewhere in the configure.ac file that
sets XMLTO. 


> > 
> > Next I'll check the candidate users of this macro to test if it's really
> > "generic" and useful. There is also the asciidoc in libXi which seems
> > very similar to xmlto in usage.
> > 
> 
> Thanks a lot for your help on this. I'm really not an expert in autotools,
> So my initial patch was really just a Draft to start the discussion.
> 

It looks like we can team-up on this. Care to submit the version of the
patch? I have a Linux box only. I never could build the xfs pdf doc. My
Linux distro doesn't have it either, I guess they couldn't build it
either.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.x.org/archives/xorg-devel/attachments/20100105/345aab1e/attachment-0001.htm