Fixing a XML schema for Cppcheck's analysis results output

[Markus Elfring]

A code review increased the software development attention for the comparison of key attributes in error information which is described to some degree by a XML schema.
I find this data format specification questionable after the discussed details.
Should the attribute “file0” become mandatory?

@@ -41,7 +41,5 @@ adjustment possibility
                            <element name="location">
-                               <optional>
-                                   <attribute name="file0">
-                                       <data type="string"/>
-                                   </attribute>
-                               </optional>
+                               <attribute name="file0">
+                                   <data type="string"/>
+                               </attribute>
                                <attribute name="file">

Will further improvements be needed for the involved data types?

[Markus Elfring]

The Cppcheck software can offer data format descriptions for different files.

• Programming interface information (configurations)
• List of available checks (option “--errorlist”)
• Output of analysis results (option “--xml”)

---

Will these areas be improved any further?

[Daniel Marjamäki]

It is optional by intention. we can avoid writing redundant information.

[Markus Elfring]

• Why do you find such information redundant?
• Is a start point for a data analysis needed?

[Daniel Marjamäki]

file1.cpp

#include "file1.h"
void f(int x)
{
}

file1.h

void f(int data);

Cppcheck will write:

        <error id="funcArgNamesDifferent" severity="style" msg="Function &apos;f&apos; argument 1 names different: declaration &apos;data&apos; definition &apos;x&apos;." verbose="Function &apos;f&apos; argument 1 names different: declaration &apos;data&apos; definition &apos;x&apos;." cwe="628" inconclusive="true">
            <location file="file1.cpp" line="2"/>
            <location file0="file1.cpp" file="file1.h" line="1"/>
            <symbol>f</symbol>
        </error>

We only need to write file0="file1.cpp" once in one <location>. I see no reason to write it on every location because for a single <error> it's always the same.</error></location>

[Markus Elfring]

I see no reason to write it on every location because for a single <error> it's always the same.

I suggest to reconsider your data model design once more here.
Would you like to introduce a case distinction?

• If you know that only a single file was analysed (so that also a single error type was detected for this example), you can avoid a bit of data repetition.
• But the provided data structures need to be consistent for the safe general usage.

[Daniel Marjamäki]

I don't want to waste more time on this debate.

[Markus Elfring]

Why do you think that anything would be wasted by this software design clarification?