David Bremner <david@tethera.net> writes:

>
>
> The following "fixes" this test failure. This suggests to me something
> that only fails when notmuch-show-buttonise-links is called from the C
> redisplay code, and not when it's called from lisp.

The last paragraph of

    https://debbugs.gnu.org/cgi/bugreport.cgi?bug=20146#47

made a lightbulb turn on over my head. I _think_ the problem is that
that fontification expects to be called with the whole message, but jit
lock really doesn't guarantee that.

If you run the attached test with emacs25, you see the list of calls (in tmp.foo/MESSAGES)

fontifying from 1 to 501
fontifying from 1 to 1
fontifying from 62 to 69
fontifying from 494 to 516
fontifying from 187 to 189
fontifying from 195 to 197
fontifying from 198 to 200
fontifying from 210 to 212
fontifying from 214 to 216
fontifying from 226 to 228
fontifying from 230 to 232
fontifying from 586 to 588

and lots more, but the point is many of them are not the whole messages,
which would allow buttons to be created out of partial matches