Re: Latest commit breaks master

[David Kastrup]

James <address@hidden> writes:

> Hello
>
> On 19 December 2012 08:11, James <address@hidden> wrote:
>> Hello,
>>
>> author  Jean-Charles Malahieude <address@hidden>
>> Mon, 17 Dec 2012 18:23:13 +0000 (19:23 +0100)
>> committer       Jean-Charles Malahieude <address@hidden>
>> Mon, 17 Dec 2012 18:23:13 +0000 (19:23 +0100)
>> commit  12c6693055728e69dce5c4e5a4a2b5f71180a5e2
>>
>> Any specific reason this wasn't
>>
>> 1. Put up for reivew
>> 2. Pushed to staging first
>>
>> This breaks master because, I am guessing as I haven't done much
>> analysis, the file name contains commas?
>>
>> I'm running patchy merge again (just in case it was something
>> anomolous - I have time, no problem retesting) but could someone
>> revert this or make sure it isn't something else?
>>
>
> Yep this is broken now.

Ok, in case this is still not clear to everybody: NEVER PUSH TO
MASTER!!!!  Nobody, ever.

It messes up our testing and branch organization even when we are not
talking about untested changes breaking the compilation.

The only one permitted to push directly to master is the
lilypond-patchy-staging script since it gives a thorough beating to what
it is going to push, making sure that it passes all reasonable tests.

In case of manual emergency pushes to master (which should be done by
experts only), the state must _first_ be established in staging, _then_
pushed to master.  However, pushing from staging to master again is
preferably done by lilypond-patchy-staging _after_ appropriately
manupulating staging.

I'll merge master into staging, revert the problematic commit there, and
then the automatisms should be able to take over from there.

-- 
David Kastrup