Re: output-distance.py

[Patrick McCarty]

On Mon, Sep 28, 2009 at 10:30 AM, Graham Percival
<address@hidden> wrote:
> On Mon, Sep 28, 2009 at 05:47:08PM +0100, Neil Puttock wrote:
>> 2009/9/28 Han-Wen Nienhuys <address@hidden>:
>>
>> > Of course, if someone goes
>> > around modifying the templates in lilypond-book.py, all the hashes
>> > change, and the comparison does not work anymore.
>>
>> Oops, I guess that's me then, since I removed the unused \paper block
>> settings without first doing a regtest check (IIRC, I only did a clean
>> docs build).
>>
>> Graham, I can sort out all the missing comparisons if you'd like to
>> get them uploaded somehow.
>
> Doing the comparisons would be tricky; you'd need to fix
> lilypond-book, then use the new lilypond-book to generate the
> tests for the old version of lilypond, etc.
>
> Personally speaking, I'm not at all concerned about the 2.13.5
> test comparisons.  If we can strip out the templates so that the
> hashes will be more robust in the future, then I'm totally content
> to say "whoops, no comparisons from 2.13.5 to the past.
> Volunteers, check the full output for 2.13.5.  Deal with it."
>
> Actually, I'm not even concerned about the current less-robust
> hashes.  It's probably a good idea for somebody to check the whole
> regtests every 4-5 months _anyway_, so having the comparisons
> break when somebody changes the paper settings (which doesn't
> happen very often!)  will just force us to solicit volunteers for
> this task.
>
> And if nobody volunteers, well, not my problem.  If no user cares
> enough about stability to spend 20 minutes looking at regtests
> once in a while, then they deserve random breakages.
>
> Cheers,
> - Graham "ask not what your engraver can do for you; ask what you
>  can do for your engraver" Percival
>
> PS that said, if anybody's interested in hacking away at the
>   python scripts, please do!

Hi Graham,

Would it work if we just calculated the md5 hash for everything
*except* the preamble?  Of course, this would still break automatic
regtest comparisons with any past versions, but it might be more
"future proof".

I'll take a better look at lilypond-book if this sounds like a good idea.

Thanks,
Patrick