https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68589
--- Comment #2 from David Bacon <bacon at cs dot nyu.edu> ---
A workaround for this gcc bug is to replace the offending 1e37 (the only one in
the .i file) with 10000000000000000000000000000000000000.1 (1{37 zeroes}.1).
That works in the 5.2.0 case, anyway.
It is worth noting also that the segfault occurred with every pure positive
power of 10 that I tried (quite a few): the .1 matters, if not to the
floating-point value realized.
dB
