Open sladen opened 6 years ago
@davelab6, input welcomed from the @googlefonts direction as to what to do with the version. In theory the binary output from the source is supposed to be completely equivalent to the binaries shipped out via GF already; but the point is to test and validate that this really is the case.
Fontbakery requires a "pattern Version X.Y between 1.000 and 9.999". Maybe we can jump to 1.0 for a baseline build and go from there? The baseline matches the current GF release (which has been used in the wild for years) as far as possible, which makes the GF release pretty 1.0-ish.
How to make a x.xxx version glyph for the Mono fonts? Make the glyph components run over the advance width? :D
Hm, the components may have a scale transformation attached as per the UFO spec...
See the existing version glyph, it uses the superscript numbers in a 2x2 grid.
We need 3 minor version figures though.
This is so squished we won't been able to read it off screen shots. (Ubuntu Mono is normally debugged at eg. 8x16 pixels, or smaller.
Eh, three minor figures is hard at small sizes. I could make it overrun the advance width so we technically keep being mono.
Four numbers, ... one digit in each quarter, scaled to take up half of the width and half of the height. Plus a decimal point in the middle, perhaps just above centre...
Edit: scale by 0.8 on x-axis, shift everything left. Better use of space.
Or no period and no scaling in the Monos.
I propose we use the concepts in http://silnrsi.github.io/FDBP/en-US/Versioning.html for versioning.
There is also https://github.com/openfv/openfv
These UFO-sources are steadily getting closer to meeting baseline equivalence with the GF "0.83" patched version. Perhaps as a Christmas/New Year surprise it can be uploaded to GF.
In which case, we need to (collectively) work out and agree upon what version number to use, and correspondingly what to put in the debug symbols so that we can tell apart the source-bulit versions from the binary-built versions (using screenshots taken by users when reporting potential bugs/regressions).
Previous experience (some years ago, per DM testing, and documented somewhere) was that only
N.NN
version numbers seemed to work in a consistent and reliable fashion.Suggestions welcomed. What number should it be?