Re: [xml] missing SAX1 support without giving --without-sax1

[Petr Sumbera <petr sumbera oracle com>]

Dne 29.10.12 07:18, Daniel Veillard napsal(a):
> On Fri, Oct 26, 2012 at 04:51:15PM +0200, Petr Sumbera wrote:
> > Hi,
> >
> > not sure what is happening here. I'm getting following during 2.9.0
> > build on S11.1:
> >
> >    CCLD   xmllint
> > Undefined                       first referenced
> >   symbol                             in file
> > docbSAXParseDoc                     ./.libs/libxml2.so
> > docbParseFile                       ./.libs/libxml2.so
> > docbParseDocument                   ./.libs/libxml2.so
> > xmlSaveNoEmptyTagsThrDef            ./.libs/libxml2.so
> > xmlXPathNZERO                       ./.libs/libxml2.so
> > xmlUnicodeCatTbl                    ./.libs/libxml2.so
> > xmlNanoHTTPFetchContent             ./.libs/libxml2.so
> > docbEncodeEntities                  ./.libs/libxml2.so
> > docbFreeParserCtxt                  ./.libs/libxml2.so
> > xmlUnicodeBlockTbl                  ./.libs/libxml2.so
> > docbCreatePushParserCtxt            ./.libs/libxml2.so
> > docbCreateFileParserCtxt            ./.libs/libxml2.so
> > docbParseChunk                      ./.libs/libxml2.so
> > docbParseDoc                        ./.libs/libxml2.so
> > docbSAXParseFile                    ./.libs/libxml2.so
> > ld: fatal: symbol referencing errors. No output written to .libs/xmllint
> >
> > At least first one seems to be because configure sets
> > WITH_SAX1_SOURCES_TRUE to '#'.
> >
> > And this is because  WITH_TRIO is not '1'. While it seems that it's
> > '1' just for HP something systems.
> >
> > So I'm confused.
>
>    It's weird. I wonder if this isn't a bug in the follwing part of
> configure.in:
>
> if test "$with_sax1" = "no" ; then
>      echo Disabling the older SAX1 interface
>      WITH_SAX1=0
>      TEST_SAX=
> else
>      WITH_SAX1=1
>      TEST_SAX=SAXtests
> fi
> AC_SUBST(WITH_SAX1)
> AM_CONDITIONAL(WITH_SAX1_SOURCES, test "${WITH_TRIO}" = "1")
> AC_SUBST(TEST_SAX)
>
>    I would guess the AM_CONDITIONAL should test WITH_SAX1 there not
> WITH_TRIO.
>
>    Could you try to make the change and see if this fixes the issue on
> S11 ?

I have tried following:

--- libxml2-2.9.0/configure.in.~1~      Mon Sep 10 20:52:46 2012
+++ libxml2-2.9.0/configure.in  Fri Oct 26 08:16:02 2012
@@ -1149,7 +1153,7 @@
     TEST_SAX=SAXtests
 fi
 AC_SUBST(WITH_SAX1)
-AM_CONDITIONAL(WITH_SAX1_SOURCES, test "${WITH_TRIO}" = "1")
+AM_CONDITIONAL(WITH_SAX1_SOURCES, test "${WITH_SAX1}" = "1")
 AC_SUBST(TEST_SAX)

And it worked. But I would expect that the same problem is on Linux then 
too!?

And I have now only following from the above list (which is now other 
not related problem):


  CCLD   libxml2.la
Undefined                       first referenced
 symbol                             in file
xmlSaveNoEmptyTagsThrDef 
/builds/ps156622/userland-libxml29/components/libxml2/libxml2-2.9.0/libxml2.syms
xmlXPathNZERO 
/builds/ps156622/userland-libxml29/components/libxml2/libxml2-2.9.0/libxml2.syms
xmlUnicodeCatTbl 
/builds/ps156622/userland-libxml29/components/libxml2/libxml2-2.9.0/libxml2.syms
xmlNanoHTTPFetchContent 
/builds/ps156622/userland-libxml29/components/libxml2/libxml2-2.9.0/libxml2.syms
xmlUnicodeBlockTbl 
/builds/ps156622/userland-libxml29/components/libxml2/libxml2-2.9.0/libxml2.syms
ld: warning: symbol referencing errors
  CC     testdso.lo
  CCLD   testdso.la
  CC     xmllint.o
  CCLD   xmllint
Undefined                       first referenced
 symbol                             in file
xmlSaveNoEmptyTagsThrDef            ./.libs/libxml2.so
xmlXPathNZERO                       ./.libs/libxml2.so
xmlUnicodeCatTbl                    ./.libs/libxml2.so
xmlNanoHTTPFetchContent             ./.libs/libxml2.so
xmlUnicodeBlockTbl                  ./.libs/libxml2.so

Note that libxml2.syms is our version of this file. So we had these 
exported as public interface. And I removed the patch which was turning 
them from static to public.

I guess that that these might have been as public interface in some 
older libxml2 version. Do you have any comment to these?

Thanks,

Petr