weak rendering of ( only? ) subnormals in gnumeric, 'shortest' of 1.5E-323 isn't 1.48219693752374E-323
title say's it all?
affected plenty of subnormals, e.g. 4E-312
-> 3.999999999998802E-312
or 2.09999979E-311
-> 2.09999979**00001**E-311
.
filed here as i assume go_dtoa as origin.
strings show up in cell display with format general ( shrinked by column width ), formula bar, cell edit ( F2 ) and status bar, some flaws go into file save ( mixed with 'fixed length value save' ) and as a result gnumeric long double produces finalized deviations.
'long' is also affected in it's own subnormal range, e.g. 4E-4950
-> 4.01E-4950
.
tested with plenty versions, e.g. gnumeric 1.12.50 ( kali distro, don't know the corresponding goffice ver. ) and fresh pulled 1.12.53 with fresh pulled goffice 0.10.52,
medium quirk, math value is kept while you stay in gnumeric with double or long datatype, but compatibility gnumeric double -> long and eventually other systems is harmed, and behavior is unexpected for normal users.