I like the URL because the PR database is the central repository
for information about bugs in gdb. It's easy for anyone to add new
information to the PR database, but it requires an FSF copyright
assignment and maintainer approval to add information to a test case.
I think that this test case has enough information even if
the PR database disappears. Specifically:
# When SIGFPE happens, the operating system may restart the
# offending instruction after the signal handler returns,
# rather than proceeding to the next instruction. This happens
# on i686-pc-linux-gnu with a linux kernel. If gdb has a naive
# signal handler that just returns, then it will restart the
# broken instruction and gdb gets an endless stream of SIGFPE's
# and makes no progress.
If you want even more text in the test case, I'm open to patches.