Re: a present for Graham

[Janek Warchoł]

2011/2/17 Graham Percival <address@hidden>:
> On 2/17/11, Janek Warchoł <address@hidden> wrote:
>> I wanted to make Graham happy, so looked through the tracker, picked
>> an existing frog issue and made a patch without discussing it for 5
>> hours.
>
> Yay!  :)

May i have a kiss?
;P

>> Unfortunately i got an error and i don't understand what went wrong, i
>> don't see any names of the files i modified. Here's the end of the
>> log:
>
> This one is important:
>
>> Failed files: (78/lily-8f0aedaa.ly)
>
> If you look in out/lybook-testdb/  DIR/FILENAME
> then you'll see the file that fails.  That file will include a
> \sourcefilename
> which points out that this particular file is
>  figured-bass-continuation-forbid.ly

Ok, now i see the typo that caused it.

> More thoughts:
> - if the regtest already contains a time signature, then altering the
> time signature is a good solution -- but I'd rather not see a *new*
> time signature if the regtest didn't already contain one.
> - if the music already contains a lot of s-rests, then go ahead and
> use an s rest.  But in general, I prefer having "real" rests (with r8
> or r16 and the like).

Fortunately that's what i did in majority of cases. I wouldn't like to
dig throug all that 250 modified files again... Doing it once was
boring enough.

> - if the patch is too large to send to the mailing list, then chances
> are it's too large.  :)
>
> Recommendation:
> - pick 50 (or so) of the "easiest" regtests to modify.  Ones that only
> require a change of existing time signature, or adding a rest or two.

Since i have no reasonable means to pick "easiest ones" from all that
i changed, i've just selected all before letter D. That's about 70
files.

> - check that it compiles after your changes. (again, don't touch any
> regtest that looks at all unusual!)

strange thing happened here. Here's what i've done:
- copied contents of input/regression to desktop (so that it would be
available outside git)
- switched back to master (in order to have unmodified regtests again
in my working directory)
- created new branch and checked it out
- ran make from /build
- ran make test from /build
- ran make test-baseline from /build
- copied first 200 regtest files (that is, up to letter C) from
desktop to lilypond-git/input/regrression (they comprised of both
modified regtests and unmodified ones, since i had no means to
separate them)
- ran make check from /build

I expected that resulting file build/out/test-results/index.html,
containing comparisons between unmodified and modified regtests, will
show differencies in the 67 regtests i changed. But surprisingly it
shows many more differencies in files i didn't change - even in ones
beginning with letter D and above. I have no idea why this happened...
I attach the patch file. It looks normal, shows changes to 67 files.

Very strange.
Janek

0001-issue-1489-regtests-ending-on-barlines-part-1.patch
Description: Text Data