On Mon, Feb 24, 2003 at 04:42:26PM +0100, Jérôme Marant wrote: > > Apparently gcc 3.2 now knows how to access 64 bit integers unaligned, > > while older gcc didn't know how to do this, or maybe it is just the > > configure script testing which went bad. > > This is good to hear that this has been worked out. Is Xavier going > to fix this a more clean way?
Yes, altough not immediately, and i will not upload a fixed version until 3.06-15 has reached testing, i think. > BTW, I noticed that ocaml 3.06-16 has been rejected by ftp-masters > and that ocaml-3.06-source was empty in this version. Could you confirm > that this was a mistake? Yes, i did wrongly rename the ocaml-source.files, and thus nothing was copied to the ocaml-source package. This was the reason the package was rejected, and i have already fixed it. I am waiting to see if 3.06-15 will enter testing without problems, and try to fix some other things (like the tex stuff i asked some time ago here, but nobody cares about it). I will also fix the policy stuff Remi submitted, and the unix sleep patch Goswin sent, mmm, i will probably get the patch xavier did in the cvs version instead. I think i will have a look at the ocaml_packaging_policy document to bring it to par with the changes that we have been doing lately. Also executables package will be separated in 3 groups : o bytecode only, like ledit. o native & bytecode versions, like spamoracle. o native & custom built bytecode versions, like advi. A bit of explanation about this latest one. It contains C bindings, and as thus it doesn't make much sense making it non-custom bytecode, since we will have to separate the stublibs into a tiny arch: any package anyway. What would be nice would be if the ocaml runtime could be put in a shared library or something for such packages, so that it will not be copied for every custom executable. Also, i will write in the policy the decision we did take about the library dependencies, or maybe use the same tactic as with the ocaml package. Friendly, Sven Luther
