I think I may have time, once I recover from lack of sleep through being
oncall, to start the 4.75 cut next week. I expect this to be a much
more leisurely affair, with plenty of time for people to comment upon
RCs, etc.
My current goal is to have as stable a release as possible, prioritising
that above new features, so that people have a trusted based
post-lockdown to be able to fall back to. I think that at this point,
bug-fixes only into the master repo going forward until after the
release? Exim 4.76 starts picking up new features again.
If there's something not in the below, or a deferral you wish to argue
against, please speak up now, in the thread, on the list, so that I can
track TODOs.
Note that this is purely the stuff *I* intend to do and what my
expectations are, but I'm only one of the maintainers. Other
maintainers might argue against one of these, submit new features over
the weekend before I actually cut 4.75_RC1, etc etc. So I'm posting
this because it's an opportunity for people whose patches have been
missed to state "you missed this bug-fix" or "it's a really small
feature and well isolated, and I've provided the documentation fix too,
so please apply it".
Items already in the ChangeLog:
http://git.exim.org/exim.git/blob/HEAD:/doc/doc-txt/ChangeLog
(plus other fixes to the supporting test/release infrastructure which
don't affect mail admins, only exim devs, so I didn't include mention of
them, since we have the git changelog for those).
Things I expect to commit (fixes in bug) once more rested:
* bug 1083: logging of expansion-failed-dnsdb-newlinejoin truncates log
message, misleading
* bug 1042: freeze_signal
* bug 1061: sensitive data disclosure leaks
Things that we should also chase for consideration for 4.75:
* bug 1019: DKIM changes (Uwe Doering's patch for multiple signature
generation)
* HP-UX: check current status of building? There was a report of
needing LIBS=-lelf which we might need to put into the OS-specific
Makefile, but how portable is this?
* bug 927 had some robustness fixes suggested in the comments (tracking
down what turned out to be a Linux kernel bug)
* anything not mentioned elsewhere in this mail but in bugzilla with a
4.75 milestone; if this applies to you, speak up if you don't want a
push back to 4.76.
Things not in for this one, IMO, but should go into 4.76:
* bug 1080, tail invocation: risks breakage to other platforms
* Eli Sand's $env_VARNAME patch
* Perhaps interrupted log_write() fix, even though log_write() should
be fine on Unix systems?
* moving the SSL versioning display into the { -d -bV } output for
consistency, and out of the non-debug { -bV } output.
-Phil
