-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Barry Jaspan writes:
> I have recently upgraded my SpamAssassin-based product to use SA
> 3.0-rc2. Generally, I think 3.0 is excellent; congratulations to the
> developers. I have a few comments:
>
> - A bug was introduced between rc1 and rc2. In BayesStore/DBM.pm, the
> calls to tie were changed to store the return value. For example,
>
> $res = tie %new_toks, "DB_File", "${name}.new", O_RDWR|O_CREAT|O_EXCL,
> (oct ($main->{conf}->{bayes_file_mode}) & 0666);
> umask $umask;
> return 0 unless $res;
>
> Later, %new_toks is untied and files are renamed. Storing $res, however,
> keeps a handle to the underlying file open, so the rename fails (at least
> on Windows). The fix is simple: after "return 0...", just add:
>
> undef $res;
Hi Barry!
gack. that definitely needs a BZ entry, a big bug on windows :(
> - The plugin interface is a great idea; it allowed me to move code I
> previously had to integrate into each new version of SA out of the SA
> tree. As currently designed, however, plugin methods are effectively
> restricted to returning integers due to the comparison with
> INHIBIT_CALLBACKS. I suggest that as few assumptions as possible be made
> about plugin method return values, and we can just require methods to call
> inhibit_callbacks() if they want to instead of returning INHIBIT_CALLBACKS.
yep, good point. I hadn't spotted that. could you open a separate
bug on that?
> - I suggest adding a plugin callback to PerMsgStatus::_get_tag, so plugins
> can define their own report tags.
sure! another bug ;)
> - I also suggest adding plugin callbacks (a) to get_report(), so plugins
> can change the report format, and (b) to report_safe() etc., so plugins can
> modify the rewritten message. The implementation of these is less clear
> than the _get_tag callback, though.
hmm. yep.
I'd suggest that this and the previous one could share a "bug" on the
BZ, since they're quite related. Also, since you're keen to see
them in the code and working, I suggest that you go for it, come up
with some kind of implementation (even if just a stub "demo" one)
so we have something to think about, and I'm sure we'll have some
constructive comments ;). It's unlikely we'd get around to
implementing these ourselves otherwise, unfortunately...
as Theo said, we use the BZ to track pretty much all patches, feature
reqs, work items, and discussion thereof ;)
thanks!
- --j.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Exmh CVS
iD8DBQFBNhVmQTcbUG5Y7woRAjZYAJ9Q1m5rnYi2fNwwTGwxKfhlk8W30QCcDjCp
HC/4ukY4xcTawbhLbYj9ej8=
=cBys
-----END PGP SIGNATURE-----
