Re: Patchy email

[James]

Hello,

On 27/07/2019 10:10, David Kastrup wrote:
> James <address@hidden> writes:
>
> > On 26/07/2019 19:36, David Kastrup wrote:
> > > ...
> > > I run Patchy when I notice something went to staging.  Due to its cost,
> > > I tend to abort it when I discover someone else pushing before me.
> > >
> > > So it would appear that your repository (and probably that of Knut) have
> > > a local master branch which would mask that the patch in question does
> > > not produce output relative to the origin repository and thus produces
> > > stuff that is not reducible.  A local master branch tends to be a bad
> > > idea (though not as bad as a local staging branch) since you don't want
> > > to collect changes of your own on it.
> > However the patchy scripts set up a local master
> >
> > e.g. If I manually delete my local master and then run the patchy scripts:
> >
> > > Branch 'master' set up to track remote branch 'master' from 'origin'.
> > > Switched to a new branch 'master'
> > (or is that not what you are talking about?)
> It is, but that does not happen in my repository when running
> lilypond-patchy-staging.py .  Since there is no point in maintaining a
> local master potentially differing from upstream in the testing scripts,
> I wonder what script would be responsible here.

I don't think it is any script per se, I used to use Lily-git (which 
fetches master and staging and sets up dev/local_working.

So I've always had a local master.

Also, I test patches against current master (not staging) so I'd need a 
local master then too right?

i.e

checkout master, run make, make test-basline, apply patch etc etc.

> > My workflow is that I always make sure that dev/local_working (where I
> > do my own changes before creating patches), local staging and local
> > master are always 'in sync' before I run patchy and that staging is
> > cleaned.
> >
> > It's no different than what I have always done.
> >
> > So I wonder why the tests passed and my own patchy merge passed but
> > yours failed?
> My patchy-staging test does not create a local master and I don't
> maintain one of my own.  I wonder why it would be different from yours.

I don't run a script to test patches - the script 'broke' when we moved 
from Google to Sourceforge, so I just test patches 'manually' (out of 
tree) but I do make sure my local master is reset 'hard' (so to speak) 
before I test things.

Patchy-merge scripts are the ones I have always used

and it creates a 'lock' branch test-master-lock (I think as a kind of 
simple failsafe if a user ends up running two or more instances or a 
previous instance failed to complete or clean up properly).

I use the script from lilypond-extra/patches/(compile_lilypond_test).


James