Re: Undocumented test cases

I
Karl wrote:
 >It seems the following test-cases are not mentioned in the documentation 
(didn't check for more):
 >xmltest/not-wf/not-sa/010.xml,
 >xmltest/not-wf/not-sa/011.xml and sun/not-wf/not-sa03.xml.

You are right,  it was an oversight on my part when doing a reclassification.


Karl wrote:
 >Question: xmltest/not-wf/not-sa/010.xml and xmltest/not-wf/not-sa/011.xml 
seem to check "Validity Constraint: Proper Declaration/PE Nesting" (P29). 
So, why are they >found in the directory for test cases that check 
violation of *wellformedness* constraints?

These test were reclassified from invalid to not well-formed during the WG 
review prior to publication. The tests are considered invalid.  However, a 
conformant processor would indicate that they fail "WFC: PE Between 
Declarations" before they reach the end of the declarations that fail "VC: 
Proper Declaration/PE Nesting".

Given the following production:

  [31] extSubsetDecl ::= ( markupdecl | conditionalSect | DeclSep)*

The replacement text of these entities clearly does not match this 
production, failing the WFC check. The moment that the end of the 
replacement text of each of these PEs is reached it clearly is the case 
that the replacement text is not well-formed per the WFC. A fatal error 
must be reported at this point.

Karl wrote:
 >An analog question also applies to xmltest/not-wf/not-sa/005.xml, where 
the >docs actually state explicitly that it checks for violation of 
validity constraint "Entity Declared >VC".

This test was also addressed during the review and it was determined that 
it should be classified as type "error", see 
(http://www.w3.org/XML/Group/2002/02/xml10-test-suite-issues ) issue TS11 
for more detail on the issue  and a list additional tests that fall into 
the same category .  These tests contain references to entities that have 
not been declared, a validity constraint failure, that are also not 
recoverable because the references occur in attribute values. For a test 
harness to distinguish this case, there needs to be such an indication in 
the test description.  This mechanism will be added in a future revision of 
the TS,  for now it was determined to set the type of
these tests to "error" so that inappropriate conformance failures are not 
reported.

Thank you for your comments.

Sandra I. Martinez


Sandra I. Martinez
National Institute of Standards and Technology
100 Bureau Drive, Stop 8970,
Gaithersburg, Md. 20899

(301) 975-3579
sandra.martinez@nist.gov

Received on Wednesday, 26 June 2002 09:20:27 UTC