What's been happening with Perl-RPM
"Randy J. Ray" <[email protected]> Wed, 10 Apr 2002 15:36:04 -0700
| Newsgroups | gmane.linux.redhat.rpm.perl |
|---|---|
| Message-ID | <[email protected]> |
Unfortunately, the short answer (at least with regards to the past 3-4 months) is "not much". Due to other demands on my time, I haven't been giving much attention to the Perl-RPM bindings of late, with the exception of reviewing the changes that get made by the maintainer of rpm itself. I hope to get some more work on this done in the near future, though. Right now, my plans for the near future are (in rough order of priority): * Test and validate the current source base against perl 5.6.0 and 5.6.1 rpm's from redhat.com. * Remove existing support for RPM 3.0.X packages. The rpm software is proceeding at a pace that makes it no longer reasonable to try and maintain compatibility with the 3.X line. Too many things have changed too thoroughly. To try a and maintain both lines only makes it that much harder to get *anything* done on the module at all. * Design invocation-logic into the Makefile.PL so that it can discern between a build that is within the context of a larger build of the overall rpm system, versus a separate build that is on a different track (an update that cannot wait for the next rpm release, for example). * Likewise, enable the testing system so that when the build is part of an overall rpm suite the local libraries are linked against at test-time. * Attempt to lighten up the overhead in RPM::Database both in terms of memory usage and processing time, when iterating over a large number of headers. Part of the limitation here is how Perl handles hashes for operations such as "keys", in particular tied hashes. Perl wants to iterate once through to get the full list of keys (because it is a tied hash, the keys are not as readily-available as with ordinary hashes) before doing the actual iteration itself. Doing this may require some re-engineering of RPM::Header to make their initial creation faster and lighter. None of this gets me any closer to being ready to handle package-level or transaction-level operations (install, uninstall, verify, etc.), and there are even more things under the hood than that (GPG signing, integrating a newer logging model). But, one step at a time, as they say. Randy -- ------------------------------------------------------------------------------- Randy J. Ray | Men occasionally stumble over the truth, but most of them [email protected] | pick themselves up and hurry off as if nothing had happened. +1 408 543-9482 | -- Sir Winston Churchill