We are reluctant to do this for a number of reasons, the most important of which is that making all import statements explicit severely clutters up the source code.
Thank you for the offer nonetheless. What I would recommend is that you pressure the GCJ people into fixing this issue (which IMO _really_ limits their usefulness as a serious java compiler), or switch to a different compiler (Sun or IBM both have decent compilers).
I accept your decision, it's your project after all, but I am a little disappointed. GCJ clearly is at fault here, but I would say it is bad programming style to use wildcard import statements. Explicit imports would only add 4 or 5 lines to the source in most cases, for the primary model interfaces.
I am working on another GPL project that incorporates Sesame and I am trying to avoid the so-called "Java trap" by building in parallel with Sun and GCJ, see below. Sesame certainly does compile successfully with the Sun compiler, but it would really help if Sesame was "clean" coded upstream so my project can keep up to date with the Sesame code base.
No hard feelings, just one last appeal to you.
I've started a discussion about this on the forum, at http://www.openrdf.org/forum/mvnforum/viewthread?thread=472
, to solicit the opinions of other developers.
A possible compromise for your project could be to use a tool like Eclipse to make the imports explicit every time you update your local source copy (simply select the project and press Crtl+Shift+o).
After further discussion (see the forum link), we've decided to reformat the code base. The current CVS trunk now only uses explicit imports.