https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121976
--- Comment #2 from Fred J. Tydeman <tydeman at tybor dot com> --- I do not have source, so that patch does me no good. I only deal with packages from Fedora. However, in looking at that patch, it appears to only deal with decimal128. The problem happens with decimal32. It looks like expressions are being evaluated as per "widest needed" (which depends upon context).
