Re: [bug-mailutils] mail ignores last line in stdin if it not followed by \n

[Sergey Poznyakoff]

Jordi Mallach <address@hidden> ha escrit:

> I figure it's hard, at this point, to track down the commit in an attempt
> at a 2.2-branch backport?

It's downright impossible. Version 3.x differs drastically from 2.x,
that's why it is assigned a new major number. 

> The alternative is to bite the bullet and start working on my promised
> experimental 3.0 packages.

That's what I'd urge you to do!

> However, since the last 3.0pre release, I've
> seen a number of big changes, some of which appeared to be interface
> changes.

None of them is critical.  The only reason why 3.0 is not yet released
is that I'd hate to make a release with such a *defective* documentation
as we have now.  However, given my sky-rocketed load average, I suppose
I'd finally cease and release it, with at least minor improvements in
the docs.  (Not willing to divulge anything, I'd only say that I'm going
to release two GNU packages this weekend ...)

> If I upload a libmailutils3 package to experimental, the API needs to be
> stable, or I'll have to bump the soname to something ugly like
> libmailutils3debian$num myself, and I'd really want to avoid that.

Generally speaking, I do not expect any major change in the
API. At most some minor bugfixes, but I don't expect many of these.
I've got plenty of major changes in my mind indeed, and there *will be*
a lot of major changes after 3.0, but not before.

> My plan is to start using a symbols list to track API additions to the
> libs, and if symbols disappear (or the interface changes) between releases
> with no soname bump, I'll have a problem.

I can assure you nothing essential is going to change without being
reflected in the library version.

> How far are we from an API/ABI freeze?

We're already there.  I'm not going to introduce any principal changes
before the major release.

> PS: off-topic, but with last night's upload, mailutils is finally
> available for the hurd-i386 architecture on the unstable suite, after a
> way too long series of build failures, all unrelated to mailutils'
> code.

Ah, that's a good news.  It's been long since I've had a possibility to
try it on Hurd. 

> https://buildd.debian.org/status/logs.php?pkg=mailutils&arch=hurd-i386
> OTOH, a new Debian architecture (s390x) has testsuite failures, but I
> still don't have an idea if this same error is still present in the
> current git master code:

> https://buildd.debian.org/status/fetch.php?pkg=mailutils&arch=s390x&ver=1%3A2.2%2Bdfsg1-5&stamp=1330644291

It's been more than a year ago, so it's hard to tell.  Since then a lot
of things has changed.  In particular, I've succeeded in getting rid
of DejaGNU as the testing framework, since it was too cumbersome to
use.  It still persists for interactive utilities, such as mail (given
no feasible alternative), but again, given the amount of changes and
insufficient information in this log, I cannot say anything definite...
Can you try this with the recent snapshot?

Regards,
Sergey