[foreign] RFR 8210860 Tree visitors should be used to handle empty names, to do symbol filtering and to do typedef name propagation

Maurizio Cimadamore maurizio.cimadamore at oracle.com
Tue Sep 18 16:22:49 UTC 2018


Looks really good,
the tree-based approach is starting to shine here, as we can clearly see 
the kind of manipulation that were already taking place in the code, but 
were previously scattered all over the place; now they are dealt with in 
a 50 LoC tree pass; nice!

A bit unfortunate that typedefs are handled so poorly by the clang API, 
so much that we need to build a dictionary of replacements, but as we 
have discussed, not much can be done about that.

Moving forward, I'd like to see those main methods turned into real 
tests: if we added ways to 'dump the tree after step XYZ', we could 
easily write unit tests which compare the output of jextract against a 
golden file.

Maurizio


On 18/09/18 16:35, Sundararajan Athijegannathan wrote:
> Please review.
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8210860
> Webrev: http://cr.openjdk.java.net/~sundar/8210860/webrev.00/
>
> Thanks,
> -Sundar



More information about the panama-dev mailing list