Bug: SIGABRT if "notmuch dump" output file is not writeable

[David Bremner]

Ralph Seichter <abbot at monksofcool.net> writes:

> * David Bremner:
>
>> I agree it's a bit ugly to look at.
>
> Ah, euphemisms. ;-) Personally, I associate "double free or corruption
> (!prev)" with memory trouble or situations where a library cannot
> recover from an error state and needs to bail out using abort(). Not
> being able to (over)write an existing file is not that serious, IMO.

Yes, but that's a message / abort from deep within libz. So odds of our
being able to fix it are pretty small. Checking for permissions before
hand would just introduce a race condition (I _think_).

>
>> Do you see any database corruption or more serious issues?
>
> No, and I don't expect any, as I am assuming that "notmuch dump" will
> only ever read the DB.

That's true, but it does take write lock on the database so that
one dumps a consistent state.

d