missing bounds checking for GenSym

[mlubin]

This reliably segfaults:

julia> eval(:($(GenSym(-1000000000)) = 1))

signal (11): Segmentation fault
unknown function (ip: -2072084214)
unknown function (ip: -2072083467)
unknown function (ip: -2072083060)
jl_expand at /mnt/hdd/mlubin/julia-0.4/usr/bin/../lib/libjulia.so (unknown line)

[tkelman]

Could 32-bit integer overflow in GenSym construction be the root cause of #10045? a0266d4?w=1 is too complicated for me to figure out what to do about it though. @vtjnash

[JeffBezanson]

Yes the gensym boxing function is declared with UIBOX_FUNC but the argument type is ssize_t. I have a patch.

[vtjnash]

Could 32-bit integer overflow in GenSym construction be the root cause of #10045

no, that would imply the creation of 2^32 gensym objects, which would require more memory than a 32-bit processor can address

[tkelman]

Would you be convinced otherwise if I run the tests repeatedly overnight on my local machine and they all pass now? ff9718e failed the first time, but on cb09d28 I'm at 5 (and counting) successes in a row now.

[vtjnash]

i would be very confused: a function with 2^31 variables would be quite impressive

[tkelman]

Nevermind. Damnit. Run number 8 failed.