Re: stop pushing to master (contextStringTuning is broken)

[Reinhold Kainhofer]

Am Montag, 5. September 2011, 23:05:12 schrieb Janek Warchoł:
> 2011/8/30 Graham Percival <address@hidden>:
> > Guys,
> >  ----------------------------------------------
> >  ***  STOP PUSHING BROKEN STUFF TO MASTER.  ***
> >  ----------------------------------------------
> > 
> > At least until we've had two more unstable releases (.9 and .10)
> > If you haven't tested your patch from a COMPLETELY EMPTY build
> > directory, done make and make docs FROM SCRATCH, then don't push
> > at all.
> 
> Is it possible that a patch which doesn't touch Documentation/ at all
> could break make doc?  I usually don't try making doc because it's
> quite time-consuming and my patches are rarely about docs (and if so,
> they have complexity of a typo-fix)...

Yes, it is possible as this incident shows. However, I see this as a bug in 
our regtests, in so far as a regtest was missing for a feature that was used 
in the docs.

So, I don't build the docs myself, either, before submitting or pushing a 
patch. I run make and make check (and I have one or two small test cases 
usually, too); everything else simply takes too long. A normal docs build 
takes ~2.5 hours here, so if I were required to do a clean build of the 
binaries, the regtest and the docs, my development would come to a standstill.

And to be honest, I usually don't even do a clean binaries build. The build 
system should be smart enough to get the dependencies right (if it doesn't I 
can always blame it on the build system ;-))

Cheers,
Reinhold

-- 
------------------------------------------------------------------
Reinhold Kainhofer, address@hidden, http://reinhold.kainhofer.com/
 * Financial & Actuarial Math., Vienna Univ. of Technology, Austria
 * http://www.fam.tuwien.ac.at/, DVR: 0005886
 * LilyPond, Music typesetting, http://www.lilypond.org