Re: Assertion failure

[Graham Percival]

On Tue, Aug 28, 2012 at 08:28:05PM +0200, address@hidden wrote:
> Slightly changing subject, I'm of the opinion that no user
> input, however unconventional, should lead to an assertion
> failure.  When someone puts an assertion failure in the code, to
> me, it is saying "if this thing is ever triggered, every
> developer should stop what she is doing and figure out what is
> wrong with LilyPond and pull all nighters until it's fixed."
> Otherwise a programming error seems more appropriate.

I would totally support an automatic rejection of any patch which
leads to an assertion failure, and of course I support anybody who
wants to spend all-nighters fixing bugs.  :)   However, a GOP
discussion from last year decided that crashes (and I would
consider an assertion failure to be a subset of a crash) were not
going to be release-Critical.  It's been about a year since then,
so I'm ok with re-opening that discussion, but I think we should
have a formal proposal about that, and I think there's more
pressing things to decide first (e.g., kickstarter, mutopia).

- Graham