osr_xml test flawed

[warmerdam]

In a build of GDAL with the JP2KAK driver, a full run of the autotest fails in osr_xml. This is because previous tests (jp2kak presumably) also does xml encoding and the xml encoder uses a process life long counter to allocate new unique gml:id's for components of the srs. So we get a failure where the first gml:id is ogrcrs13 instead of ogrcrs1 as expected.

The problem can also be reproduced by just copying osr/osr_xml.py to osr/osr_xml2.py and reruning just the osr series (run_test.py osr).

I imagine the osr_xml.py test will need to make a pass through the XML stripping out the values of the gml:id's to avoid unnecessary test failures based on different ids being produced.

[Even Rouault]

Frank,

Wouldn't it be better and simpler to avoid the static variable in ogr_srs_xml.cpp instead ? That's the purpose of the attached patch. The only thing that prevents me to commit it right now is that I'm not sure what the expected extent of the unicity of the gmlid is (haven't read the OGC XML spec...) . If it must only be unique in the context of the elements generated by a call to exportToXML(), then my patch should do it.

[Even Rouault]

Remove use of static variable

[warmerdam]

Even, the uniqueness must at the very least extend over one complete XML document, and the exportToXML() is expected to produce only a fragment of a document. There are cases, particularly the production of coordinate system dictionaries, in which one GML document might include several coordinate systems.

So I don't think removing the static variable is an appropriate solution.

[Even Rouault]

Chaintanya,

you can forget the patch and follow Frank's initial suggestion

[chaitanya]

Modified the test in trunk (r20245).