[dev-context] regression testing: page image differences

[Taco Hoekwater]

Sanjoy Mahajan wrote:
>>>menas that p.1 was different in the two runs, but only slightly (about
>>>0.2 dB).  Here another example with a slightly larger dB difference:
> 
> 
>>Could it be because of the differences between lm-math and cm-math
>>fonts?
> 
> Good thought.  I'm pretty sure you're right.  Here's one of the 5dB
> differences.  Each pdf file uses five fonts, and the two fonts that
> vary between the files are the math ones:

That is my guess as well (looks like a hinting difference, actually).

Best, Taco