[foreign-jextract] RFR: 8262733: jextract generates clashing names which results in compilation error with javac [v2]

[Maurizio Cimadamore]

On Mon, 1 Mar 2021 12:44:29 GMT, Athijegannathan Sundararajan <sundar at openjdk.org> wrote:

>> name clash is avoided by checking surrounding class names
>
> Athijegannathan Sundararajan has updated the pull request incrementally with one additional commit since the last revision:
> 
>   incorporating changes suggested in review.

Added more comments

src/jdk.incubator.jextract/share/classes/jdk/internal/jextract/impl/StructBuilder.java line 58:

> 56:         this.structType = structType;
> 57:         prefixElementNames = new ArrayDeque<>();
> 58:         setterParameterName = safeParameterName("x");

Can we avoid the caching and just call `safeParameterName` where required? I think caching like this makes code generation more brittle than it has to, as it introduces an implicit dependency. In other words, in a way it is an "accident" that all methods generated in struct builder accept a parameter named "x" - I'd like not to rely on this assumption too much, and make sure that _all_ parameters of _all_ methods in any of the nested builders (StructBuilder, FunctionalInterfaceBuilder) are validated accordingly.

For instance, in `emitIndexedFieldSetter` the problem is not just `x` - the problem is also the `seg` and the `index` parameter (as the struct could be called `seg` or `index`). In other words, I think the parameter name should be mangled in a lot more places.

-------------

PR: https://git.openjdk.java.net/panama-foreign/pull/462