Migration of libc++ weirdly blocked by already-fixed bug

[Adam D. Barratt]

On Mon, 2017-08-28 at 18:12 +0800, Boyuan Yang wrote:
> Hello all,
> 
> I was checking the version of libc++ in Debian archive [1] and found that 
> package "libc++" failed to migrate to testing  in last three years.
> 
> While checking for migration excuses, I found that the statement of excuse is 
> kind of weird. Currently it says: [2]
> 
> Excuse for libc++
> 
>     Migration status: BLOCKED: Rejected/introduces a regression (please see 
> below)
>     Too young, only 0 of 5 days old
[...]
>     libc++ has new bugs!
>     Updating libc++ introduces new bugs: #870440
>     Piuparts tested OK - https://piuparts.debian.org/sid/source/libc/libc+
> +.html
>     Not considered 
> 
> Bug #870040 has been fixed for a while [3] yet the excuse still mentions it.

Yes, because it's not fixed on all architectures. You can see this by
looking at the version graph at the top right of
https://bugs.debian.org/870440 - there's a red oval labelled "libc
++/3.9.1-2 (unstable)", indicating the presence of an unfixed version in
unstable. This is due to the armel packages for libc++ still being at
version 3.9.1-2, and therefore *not* being fixed.

> I think it might be a bug in QA scripts. Could you please look into it and 
> make the excuse correct (which is FTBFS on certain architectures, not 
> introducing new bugs)?
[...]
> [1] https://tracker.debian.org/pkg/libc++
> [2] https://qa.debian.org/excuses.php?package=libc%2B%2B
> [3] https://bugs.debian.org/870440

Excuses come from britney, and are therefore a Release Team thing, not a
QA team thing.

In any case, the excuses _are correct_ here. There are binary packages
built from libc++ in unstable that contain the issue discussed in
#870440. There are no binary packages in testing that contain the issue,
at least assuming that the version information for that bug is correct.
Therefore, migrating the current packages from unstable to testing would
introduce the bug in to testing.

Regards,

Adam