Seeing the discussion of strings.Repeat and invalid inputs in another
thread, I noticed a couple of bugs. One I have reported -
https://github.com/golang/go/issues/16237
The other I'm not sure if it's worth a report. Here's the source for
bytes.repeat:
391 func Repeat(b []byte, count int) []byte { 392 nb :=
make([]byte, len(b)*count) 393 bp := copy(nb, b) 394
for bp <
len(nb) { 395 copy(nb[bp:], nb[:bp]) 396
bp *= 2 397 }
398 return nb 399 }
Here, if len(nb) is small enough to fit in an int, but large enough that 2
* nb overflows an int, then there may be some cases where bp < len(nb) but
2 * bp overflows. Most likely leading to a panic from slicing with negative
indices.
But this can only happen if the allocation of len(nb) bytes succeeds. With
64 bit ints, that's ridiculous (this year, anyway). With 32 bit ints - can
an allocation of a gigabyte succeed? I suppose that depends on OS details.
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
