Re: FYI: Ship and install bison.xsl

[Paolo Bonzini]

Joel E. Denny wrote:
> On Sun, 18 Nov 2007, Joel E. Denny wrote:
>
> > On Sat, 17 Nov 2007, Joel E. Denny wrote:
> >
> > > On Wed, 14 Nov 2007, Akim Demaille wrote:
> > >
> > > > It might be time to write some (simple) tests of the XML output.  If I
> > > > have some time I will, but...
> > > We might consider adding a maintainer-xml-check similar to 
> > > maintainer-push-check.  That is, we might test the XSLT output against the 
> > > --report and --graph output for all grammars in the test suite.  Of 
> > > course, this would not help with xml2xhtml.xsl, so it would definitely 
> > > need its own test cases.
> > >
> > > I'm sure we could come up with something in M4 to make this happen.  
> > > Unfortunately, the "Big triangle" and "Big horizontal" torture tests never 
> > > complete on my system when testing xml2dot.xsl, so we would want to be 
> > > able to specify that certain test cases be skipped.
> > Using the following patch, here's what I've been doing to test 
> > xml2text.xsl and xml2dot.xsl:
> >
> >   BISON_TEST_XML=1 ./testsuite 1-156 159-
> >
> > M4 seems like a cleaner approach, but maybe we can salvage some of this 
> > patch.
>
> Here's an uncommitted patch that implements maintainer-xml-check with a 
> new macro named AT_BISON_CHECK.  It also uses AT_BISON_CHECK to replace 
> the ugly AT_CHECK rewrite from testsuite.at.
>
> What should we do about xsltproc detection?  Should we just mention 
> xsltproc in README-hacking and assume every maintainer will have it in his 
> PATH?  Should we check for it in configure.ac using AC_PATH_PROG?  The 
> patch below does not do the latter.

Yes, and IMO the test should be done unconditionally if xsltproc is 
available.

Also, this:

  AT_CHECK([[xsltproc \
             `]]AT_QUELL_VALGRIND[[ bison 
-print-datadir`/xslt/xml2text.xsl \
             xml-tests/test.xml > xml-tests/test.xml.output]])
  AT_CHECK([[diff -u xml-tests/test.output xml-tests/test.xml.output]])

should be done like this since Autotest knows how to do diffs:

  AT_CHECK([xsltproc \
            `AT_QUELL_VALGRIND bison -print-datadir`/xslt/xml2text.xsl \
            xml-tests/test.xml], 0, [xml-tests/test.output])])

(i.e. first it should tell autotest to do the diff itself, and second I 
don't see the point in double-quoting if we know macros like xsltproc 
and bison are not going to be defined).

Paolo