Re: "Assertion failed!"

[David Kastrup]

"Andrew Bernard" <address@hidden> writes:

> Hi Daniel,
>
> That's another assertion failure on top of the one I have been reporting in
> 2.19.21. Can you please submit a bug report, given you have  an example that
> produces the error?
>
> Something seems to be a bit broken in 2.19.21 compared to 2.19.20.Iif I had
> the knowledge, I would jump in and fix it, but it's too early for me as yet,
> not being familiar with the internals.
>
> Can you see if this works in 2.19.20? [Based on my hypothesis that 2.19.21
> has regressed something, or introduced a bunch of defects.]

More likely than not it is

    <URL:https://code.google.com/p/lilypond/issues/detail?id=2787>,
    Issue 2787: Sanitize usage of -DDEBUG, -DNDEBUG, assert,
    programming_error and so on

that is making the difference for you here.  "assert" is a programmer's
way to indicate that the failure of the assertion is something that the
program is unable to deal with.  Consequently, a failed assertion aborts
the program rather than continuing and likely crashing in unpredictable
ways.

Now assertion checks previously to this issue were only enabled on
debugging builds, so only developers would have gotten to see them in
versions compiled for testing.

But it would appear that some cases of "assert" are perfectly capable of
continuing gracefully, so that's more a case for the diagnostic
"programming error" which reports conditions that should not be occuring
but that are not fatal.

They still deserve investigation but aborting the program for them would
appear to be overkill as it seems unlikely they would cause operations
to completely derail.

-- 
David Kastrup