Re: input/regression/scheme-text-spanner.ly: fix problem with constants (issue 11614044)

[dak]

On 2013/07/23 13:52:13, david.nalesnik wrote:

> Sorry, replied a bit hastily!

[...]

> I've thought some more about the issue of trying to change constants
elsewhere
> in the file.

> Specifically, regarding the addition of the new grob description to
> all-grob-descriptions, would the following be any better?

> So, as I understand it, remove add-grob-description, duplicating its
changes to
> the description of 'SchemeTextSpanner, and creating a new variable to
avoid
> tampering with all-grob-descriptions, using that new variable in the
layout
> block.

No, that's not necessary.  all-grob-descriptions is defined as a session
variable, so it will be restored at the end of the session.  You must
not change it _destructively_ (meaning that you must not change any of
its cons cells or what they point to), but you can replace with a
different list as a whole.

Since the code only added to the _front_ of all-grob-descriptions, this
part was entirely harmless.

The problem is rather with the code in "complete-grob-entry" or what it
was.  That one worked destructively on constant lists (like, one may
add, our current code for creating all-grob-descriptions originally
does).

So what is being done is just as bad as our existing code elsewhere, but
there is no interference with it.

The part adding stuff to the front of all-grob-descriptions is fine by
itself.

https://codereview.appspot.com/11614044/