Le 19/01/2017 à 16:04, Jean-Marc Lasgouttes a écrit :
Here is what happened
1/ I change randomly (!) the spacing of math equations in the MathRow work
2/ as a consequence the formulas in lib/symbols are often wrong
3/ I fix the problem with arabic test at e832d2e90f
4/ I do a new round of modifications of lib/symbols at 7335ee8e,
without realizing that what I am fixing is a consequence of the commit
3/ and not of the ongoing work 1/
Does this make sense to you? The fact that the right value now is
without explicit kern is a proof that all this work on math formula
spacing has been done correctly (almost!)
Makes sense to me. Lots of changes to spacing, and lib/symbols that lags
behind. A bug slips by the new features.
Nice that this odd kern is not needed after all. The value has been
found by trial-and-error. However may I suggest: what's good for display
is not necessarily the best for structural interaction—can we
reintroduce a kern to make it slightly imperfect, so that there is
feedback during selection (as you did in current master for \Arrownot)?
Another thing I would have mentioned is indeed that \{A,a}rrownot works
on the same principle, but I saw that you already took care of it.
(speaking about master, did not try stable)
