[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: Libm core dump with SIGILL
On 2015-08-21 16:10, Paul Koning wrote:
On Aug 21, 2015, at 10:01 AM, Anders Magnusson <ragge%ludd.ltu.se@localhost> wrote:
Greg Stark skrev den 2015-08-21 14:35:
While running the Postgres regression tests I found that the exp()
function in libm gets a SIGILL with certain arguments.
simh$ gcc -Wall exp.c -lm
 Illegal instruction (core dumped) ./a.out
On a modern architecture with IEEE floats:
$ gcc -Wall exp.c -lm
exp(88.0297) = 1.70141e+38
I know the VAX floating points are not exactly IEEE and they may not
be able to represent this value but I would expect them to get a FPE
or signal overflow in some other way, not get a SIGILL?
Well, not in VAX FP. From the man page for exp:
These functions will return the appropriate computation unless an error
occurs or an argument is out of range. The functions exp() and expm1()
detect if the computed value will overflow, set the global variable errno
to ERANGE and cause a reserved operand fault on a VAX.
So, it is following the specs... :-)
But SIGILL means illegal instruction, not reserved operand. Isn't EMODD an optional instruction? I wonder if the issue is that it's not recognized on Greg's platform so he ends up with the wrong exception.
Wouldn't it be cleaner for the code to call signal() rather than attempt to force a particular signal by using a weird instruction?
Good point! I was also wondering about the SIGILL, since it sounded like
an illegal instruction to my ears, but I'm not sure exactly what gets
propagated as what.
The VAXen that don't have the instruction in hardware should have
emulation for them anyway, so I'm not sure exactly what is going on with
I'm not sure there is any signal that better matches the illegal
argument trap. Maybe SIGILL is the best one around for that.
Main Index |
Thread Index |