Michael Tautschnig wrote:

I'd be willing to package PPL for Debian, even more so as most of the work has already been done by the developers themselves.

Dear Michael,

this is great news! The project "packaging PPL for Debian" was frozen because our group does not currently have an expert in Debian packaging. If you can guide us, we could finally complete the job.

One question remains though: Which version to package? 0.9 seems to be pretty old already, but neither the latest CVS snapshot nor the latest release package succeed in building (primiarily because of problems when building the docs). Roberto, could you gives me some hints on that one?

PPL 0.10 is not yet ready for release. One of the problems is due to the fact that PPL 0.10 is not currently suitable for binary distribution. In fact, at library startup the library needs to know, for IA32, whether the CPU supports SSE, whether it supports SSE2, whether it supports 387. Currently this selection is only possible at configure time with the option

--enable-fpmath

which can take the following values:

sse, sse2, 387, sse+387, sse2+387.

We have still not decided how best to implement this test at runtime. Until that is done, a binary release of what will become PPL 0.10 will not be possible for IA32. Other things are not yet ready in PPL 0.10 (the implementation of intervals and boxes are missing the approximated set difference operator, for instance).

So, I think that packaging PPL 0.9 is currently the right thing to do. Most applications we know of are indeed using that version. When PPL 0.10 is officially released, with your help (I hope) we will sort out the additional packaging issues (PPL 0.10 has new foreign language interfaces for OCaml and Java, for instance).

We look forward to hearing from you. All the best,

Roberto