[pkg-go] Plans for uploading Go 1.5 to unstable?

[Michael Hudson-Doyle]

On 16 October 2015 at 06:40, Michael Stapelberg <stapelberg at debian.org> wrote:
>
>
> On Thu, Oct 15, 2015 at 6:44 PM, Tianon Gravi <admwiggin at gmail.com> wrote:
>>
>> On 13 October 2015 at 16:43, Potter, Tim (Converged Cloud)
>> <timothy.potter at hpe.com> wrote:
>> > Hi tianon.  I was curious what your plans were for uploading Go 1.5 to
>> > unstable.  Is there anything you can share?
>>
>> Go 1.5 has some minor breaking changes, so I've been waiting until I
>> (or some other kind soul) has had the time to do a full ratt run on Go
>> 1.5 and make sure all our packages either build/test successfully or
>> that we have a plan/patches ready to fix the issues that are bound to
>> come up.
>
>
> FWIW, I don’t think that’s necessary for the compiler itself. Go adheres to
> its Go 1 stability guarantee, and any package which breaks likely did
> something wrong in the first place :). If this is the only thing that’s
> holding back the upload, I’d recommend to just go ahead with the upload.

That's a nice idea, and for pure go things it might even be close to
true, but it's definitely not true universally. Here's a few reasons
for failures I saw when prepping 1.5 for ubuntu:

1) cgo is a particularly rich source of incompatibilities (see docker,
and also the tightening over time rules about passing Go pointers to
C)
2) the go tool now explodes if you're careless and set GOPATH to have
an empty component
3) packages that have expectations of what tracebacks look like in
their testsuite
4) testing/quick can produce nil pointers now
5) golang-x-text was broken by a new unicode version
6) a few ftbfs on arm builders with low ram

Some of these are definitely bugs in the package or packaging but not all.

Cheers,
mwh