Mlmmj 1.2.18a1 released
An alpha release for the upcoming Mlmmj 1.2.18 has been published. Find it on the downloads page[1]. Please help test it if you can and report bugs using our bug tracker[2] or here on the discussion mailing list. Also please report successes to help gain confidence in the quality of the code. Please also contact me (the maintainer, Ben Schmidt) if you can help with code review.
This version contains a number of important new features:
- Richer list texts (including a new naming scheme, conditionals, automatic wrapping and true UTF-8 support).
- Notifying posters when their posts are being moderated.
- Not-me-too feature to avoid having the poster receiving their own messages.
- Ability to explicitly reject posts.
- Ability to explicitly obstruct subscriptions.
- Unsubscribe from all versions of a list at once.
- Ability to switch between different versions of a list by using +subscribe[-{digest|nomail|both}] ('both' means normal and digest, and is a 'hidden feature').
- A number of bug fixes (including fixing some bugs which could cause mail to be lost).
- Improved documentation.
(See the ChangeLog[3] for a more detailed list of changes.)
It is almost entirely backward-compatible. For instance, old list texts should continue to work. The main backward-incompatible change is that unsubscribe will now remove a user from all versions of the list, not just one (no matter which unsubscribe address is used).
It has only been minimally tested and not all changes have been reviewed yet, so there will probably be some bugs, possibly some nasty ones, waiting to be found. Any and all help with code review and testing will be greatly appreciated.
Following further code review and testing, a beta version and/or release candidate will be published.
Ben.
[1] http://mlmmj.org/downloads
[2] http://mlmmj.org/bugs
[3] mlmmj.org/hg/mlmmj/file/RELEASE_1_2_18a1/ChangeLog
Quoted 1 line · Ben Schmidt
An alpha release for the upcoming Mlmmj 1.2.18 has been published.
Hi!
Why suddenly changing versionning scheme? Does the "a" means alpha?
Later, if you publish 1.2.18.0, it's going to be considered smaller than 1.2.18a1, so I think it's not a nice numbering scheme (it might break upgrading in Debian, for example). Something like:
1.2.18.0.rc1
would have been better. Then, you would release 1.2.18.1 ...
Cheers,
Thomas
Quoted 6 lines · Thomas Goirand
Quoted 1 line · Ben Schmidt
An alpha release for the upcoming Mlmmj 1.2.18 has been published.
Hi!
Why suddenly changing versioning scheme?
I didn't realise I changed anything.
Quoted 1 line
Does the "a" means alpha?
Yes, A is for alpha. Followed by B for beta, and RC for release candidate. That's the system in my head, anyway.
Quoted 2 lines
Later, if you publish 1.2.18.0, it's going to be considered smaller than 1.2.18a1, so I think it's not a nice numbering scheme
Well, that depends how you sort, doesn't it? What I've used is a fairly common numbering system that a lot of packages use, so it wouldn't surprise me if many sorting algorithms for version numbers dealt with it OK. A simple ASCII sort won't though, it's true.
Quoted 1 line
(it might break upgrading in Debian, for example).
I'm not sure that an alpha release should be included in a distribution anyway, to be honest.
Quoted 5 lines
Something like:
1.2.18.0.rc1
would have been better. Then, you would release 1.2.18.1 ...
I guess I probably should have had a hyphen in there, that's all. 1.2.18-a1 would be better, right? I'll do that for beta and beyond.
Cheers,
Ben.
Quoted 3 lines · Ben Schmidt
Quoted 2 lines · Thomas Goirand
Quoted 1 line · Ben Schmidt
An alpha release for the upcoming Mlmmj 1.2.18 has been published.
Works fine so far. Didn't notice any regression.
Quoted 12 lines
Quoted 4 lines
Hi!
Why suddenly changing versioning scheme?
I didn't realise I changed anything.
Quoted 1 line
Does the "a" means alpha?
Yes, A is for alpha. Followed by B for beta, and RC for release candidate. That's the system in my head, anyway.
Quoted 7 lines
Quoted 2 lines
Later, if you publish 1.2.18.0, it's going to be considered smaller than 1.2.18a1, so I think it's not a nice numbering scheme
Well, that depends how you sort, doesn't it? What I've used is a fairly common numbering system that a lot of packages use, so it wouldn't surprise me if many sorting algorithms for version numbers dealt with it OK. A simple ASCII sort won't though, it's true.
Yeah, this is used by several projects and shouldn't cause any trouble... It works under Arch Linux (pacman, which uses a version comparison algorithm ripped off from RPM):
$ vercmp 1.2.17.0 1.2.18.0
-1
$ vercmp 1.2.18.0a1 1.2.18.0
-1Quoted 11 lines
Quoted 2 lines
(it might break upgrading in Debian, for example).
I'm not sure that an alpha release should be included in a distribution anyway, to be honest.
Quoted 5 lines
Something like:
1.2.18.0.rc1
would have been better. Then, you would release 1.2.18.1 ...
Works as well but looks kind of weird, imho:
$ vercmp 1.2.18.0.rc1 1.2.18.1
-1Quoted 3 lines
I guess I probably should have had a hyphen in there, that's all. 1.2.18-a1 would be better, right? I'll do that for beta and beyond.
Many distributions use hyphens for pkgrels/patches:
$ vercmp 1.2.18.0-a1 1.2.18.0
0Feel free to use that anyway, it's your decision :)
Quoted 7 lines
Cheers,
Ben.
Quoted 3 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
An alpha release for the upcoming Mlmmj 1.2.18 has been published.
Works fine so far. Didn't notice any regression.
Thanks a lot for the feedback.
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Later, if you publish 1.2.18.0, it's going to be considered smaller than 1.2.18a1
In trying to figure out a better scheme, I disproved it for Debian:
$ dpkg --compare-versions 1.2.18a1 lt 1.2.18.0 && echo YES || echo NO YES
It's a problem if the final .0 is omitted:
$ dpkg --compare-versions 1.2.18a1 lt 1.2.18 && echo YES || echo NO NO
That is what I would have done, but due to this issue, I'm happy enough to include the final .0 as occasionally we do have three-point releases anyway, and this may be the best solution.
Quoted 8 lines
Yeah, this is used by several projects and shouldn't cause any trouble... It works under Arch Linux (pacman, which uses a version comparison algorithm ripped off from RPM):
$ vercmp 1.2.17.0 1.2.18.0 -1 $ vercmp 1.2.18.0a1 1.2.18.0 -1
What about
$ vercmp 1.2.18a1 1.2.18.0
?
Smiles,
Ben.
Quoted 14 lines · Ben Schmidt
Quoted 3 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
An alpha release for the upcoming Mlmmj 1.2.18 has been published.
Works fine so far. Didn't notice any regression.
Thanks a lot for the feedback.
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Later, if you publish 1.2.18.0, it's going to be considered smaller than 1.2.18a1
In trying to figure out a better scheme, I disproved it for Debian:
$ dpkg --compare-versions 1.2.18a1 lt 1.2.18.0 && echo YES || echo NO YES
Quoted 8 lines
It's a problem if the final .0 is omitted:
$ dpkg --compare-versions 1.2.18a1 lt 1.2.18 && echo YES || echo NO NOThat is what I would have done, but due to this issue, I'm happy enough to include the final .0 as occasionally we do have three-point releases anyway, and this may be the best solution.
Oh, I didn't even notice the missing revision/build part. Yeah, should should include that...
Quoted 5 lines
Quoted 5 lines
Yeah, this is used by several projects and shouldn't cause any trouble... It works under Arch Linux (pacman, which uses a version comparison algorithm ripped off from RPM):
Quoted 8 lines
Quoted 4 lines
$ vercmp 1.2.17.0 1.2.18.0 -1 $ vercmp 1.2.18.0a1 1.2.18.0 -1What about
$ vercmp 1.2.18a1 1.2.18.0
It is pretty clever:
$ vercmp 1.2.18a1 1.2.18.0
-1... :)
Quoted 9 lines
?
Smiles,
Ben.
Quoted 5 lines
Quoted 2 lines
$ dpkg --compare-versions 1.2.18a1 lt 1.2.18.0 && echo YES || echo NO YES$ vercmp 1.2.18a1 1.2.18.0 -1
OK. This scheme wins.
Thanks for raising and clarifying the issue, guys.
Ben.
Hi everyone,
I know this comes a little late, and I don't really ave the time to introduce myself right now, but I would like to back on the 1.2.18-a1 / 1.2.18 proposal. It seems to me that it is the most intuitive and widely used.
If your only concern is to be compatible with debian's scheme, the tilde was introduced for this reason: it is the only character smaller than the empty string, meaning that
1.2.18~ < 1.2.18 < 1.2.18-
and therefore that
1.2.18~a3 < 1.2.18~b2 < 1.2.18~rc1 < 1.2.18 < 1.2.18-p1
You can even formally call the package 1.2.18-a1 and give the deb the number 1.2.18~a1, I don't think anyone would get confused.
As soon as I have more time I will try to do more fundamental contributions.
Signature · 1 line
Quentin