About tdf#121074

[Caolán McNamara]

On Mon, 2018-11-05 at 20:46 +0100, Julien Nabet wrote:
> Looking at git history from 
> vcl/qa/cppunit/graphicfilter/data/svm/fail/mapmode-1.svm, I got 
> d776eeab752fd313226a7570c3ed4d8e964b4406
> 
> "ofz#937 sanity check claimed record length"

Documents like that are generated from the oss-fuzz fuzzing thing, and
what matters generally is that we don't crash/busy-loop on loading
them, not that they are in any way sane documents.

> 1) Do you think these asserts are ok?

I have my doubts that a negative width/height is something that is
guaranteed to indicate a bug, though it might indeed be suspicious.

> 
> 2) If yes, any thoughts how to fix this specific test with the svm 
> (hoping there won't be more but I'm a bit pessimistic about this)?

Well, for the purposes of continuing your experiment I imagine you
could just consider negative nTmp32 for width/height as failures and
return early.