[Fontinst] fontinst fontdsize variance?

[Lars Hellström]

Karl Berry skrev 2016-02-04 01.07:
> Hi Lars - thanks much for all the investigations.
>
>      tftopl naively reports the latter
>
> tftopl is just reporting what it sees, right?  Can't blame it for that.

Well, it could alternatively opt to ignore the last 4 bits.

>
>      Yes; vptotf isn't all that detailed in its messages
>
> So maybe this is where the fix should lie: vptovf should not report
> variances in design size up to<some amount>, because of the difference
> between mf and tfm representations/resolutions.  Wdyt?

In principle possible, but might be harder to push. It's also the matter 
that a DVI driver is allowed to be dumb about this and reject data in the 
binary files that doesn't match.

> The discrepancy seems worth reporting to Knuth

Especially since it only involves his own programs, and ones that are fairly 
old to boot. That the designsize of cmr17 is 17.279999pt, but _not_ 17.28pt, 
is quite ugly. It only started to matter after virtual fonts were 
introduced, but still.