Re: make test-baseline / make check problem

2017-10-20 Thread Knut Petersen 

Am 20.10.2017 um 01:33 schrieb Thomas Morley:



I found nothing in lilyissues ... I think this needs at least to be
documented.
Or did I miss something?

Knut

Mentioned (with a bit discussion) here:
http://lilypond.1069038.n5.nabble.com/Things-that-make-you-go-hmmm-tt205547.html


I created issuse 5217 for this problem.

This night a script iterated a "make ... test-clean" and "make ... check" loop 
and stored the results.

All but the merge-rests-engraver regtests showed stable results. The 
probability of the two possible pngs created from merge-rests-engraver.ly seems 
to be identical.

Knut



___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: make test-baseline / make check problem

2017-10-19 Thread Thomas Morley 
2017-10-20 0:39 GMT+02:00 Knut Petersen :
> Am 20.10.2017 um 00:00 schrieb Thomas Morley:
>>
>> 2017-10-19 21:26 GMT+02:00 Malte Meyn :
>>>
>>> Does this problem occur also in a version before this patch?
>>
>> Yep.
>> It was noticed soon after
>>
>> commit 017ef9975f2fe3d6990c85b8107801706dc71baa
>> Author: Thomas Morley thomasmorle...@gmail.com
>> Date: Tue Aug 29 10:44:28 2017 +0200
>>
>> Issue 5179 Let Merge_rests_engraver deal with dotted rests
>
>
> I found nothing in lilyissues ... I think this needs at least to be
> documented.
> Or did I miss something?
>
> Knut

Mentioned (with a bit discussion) here:
http://lilypond.1069038.n5.nabble.com/Things-that-make-you-go-hmmm-tt205547.html

Cheers,
  Harm

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: make test-baseline / make check problem

2017-10-19 Thread Knut Petersen 

Am 20.10.2017 um 00:00 schrieb Thomas Morley:

2017-10-19 21:26 GMT+02:00 Malte Meyn :
Does this problem occur also in a version before this patch? 

Yep.
It was noticed soon after

commit 017ef9975f2fe3d6990c85b8107801706dc71baa
Author: Thomas Morley thomasmorle...@gmail.com
Date: Tue Aug 29 10:44:28 2017 +0200

Issue 5179 Let Merge_rests_engraver deal with dotted rests


I found nothing in lilyissues ... I think this needs at least to be documented.
Or did I miss something?

Knut

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: make test-baseline / make check problem

2017-10-19 Thread Thomas Morley 
2017-10-19 21:26 GMT+02:00 Malte Meyn :
>
>
> Am 19.10.2017 um 14:59 schrieb Knut Petersen:
>>
>> Hi everybody!
>>
>> I constantly hit the problem that the files
>>
>>   *
>> build/out/test-results/input/regression/out-test-baseline/merge-rests-engraver.png
>>   *
>> build/out/test-results/input/regression/out-test/merge-rests-engraver.png
>>
>> are different (a dot moves) - even if I test identical versions of
>> lilypond:
>
>
> Hm … I didn’t notice that when I made the patch for issue 5211. Probably
> because I changed the regression test itself and therefore expected
> different output.
>
> Does this problem occur also in a version before this patch?



Yep.
It was noticed soon after

commit 017ef9975f2fe3d6990c85b8107801706dc71baa
Author: Thomas Morley thomasmorle...@gmail.com
Date: Tue Aug 29 10:44:28 2017 +0200

Issue 5179 Let Merge_rests_engraver deal with dotted rests

Compare simple rests by their duration-length, duration-log does
not take possible dots into account.
Superfluous dots are killed with ly:grob-suicide!
Extend reg-test

Though, I can't see any reason _in_ this patch, more likely the
extended regtest triggers the problem David described earlier in this
thread.

Cheers,
  Harm

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: make test-baseline / make check problem

2017-10-19 Thread Malte Meyn 



Am 19.10.2017 um 14:59 schrieb Knut Petersen:

Hi everybody!

I constantly hit the problem that the files

  * 
build/out/test-results/input/regression/out-test-baseline/merge-rests-engraver.png 

  * 
build/out/test-results/input/regression/out-test/merge-rests-engraver.png


are different (a dot moves) - even if I test identical versions of 
lilypond:


Hm … I didn’t notice that when I made the patch for issue 5211. Probably 
because I changed the regression test itself and therefore expected 
different output.


Does this problem occur also in a version before this patch?

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: make test-baseline / make check problem

2017-10-19 Thread David Kastrup 
Knut Petersen  writes:

> Hi everybody!
>
> I constantly hit the problem that the files
>
>  *
> build/out/test-results/input/regression/out-test-baseline/merge-rests-engraver.png
>  * build/out/test-results/input/regression/out-test/merge-rests-engraver.png
>
> are different (a dot moves) - even if I test identical versions of lilypond:

LilyPond does not behave entirely deterministically: there are some
operations depending on memory allocation order, and memory layout tends
to be randomized to some degree in current distributions.

I've eliminated some of those dependencies but have not been able to
find and/or eliminate all.

-- 
David Kastrup

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel