Bug #9173: rb_sprintf %li format specifier does not work correctly with long values and can cause Ruby to crash

• Status: Rejected
• Assignee: -
• Target version: -
• ruby -v: ruby 2.0.0p195
• Backport: 1.9.3: UNKNOWN, 2.0.0: UNKNOWN

[bladenkerst (Brad Sumersford)]

There is additional handling of %i versus %d in BSD_vfprintf (>= 2.0). When l (long) is specified as the length, %li, the additional handling is triggered.
Note: %ld works fine, %li and %ld work in Ruby 1.9.3

Sample Code:

sprintf_test.c

#include "ruby.h"

static VALUE
sprintfTest(VALUE module, VALUE fix) {
Check_Type(fix, T_FIXNUM);

long long_value = FIX2LONG(fix);
VALUE back_to_ruby = LONG2FIX(long_value);

printf("This should show the correct value: %li\n", long_value);
rb_funcall(rb_mKernel, rb_intern("puts"), 2, rb_str_new2("This should also show the correct value: "), back_to_ruby);

return rb_sprintf("This will return the wrong value, or crash: %li", long_value);
}

void Init_sprintf() {
rb_define_method(rb_mKernel, "sprintf_test", sprintfTest, 1);
}

irb

require './sprintf'

irb(main):004:0> sprintf_test 3
This should show the correct value: 3
This should also show the correct value:
3
=> "This will return the wrong value, or crash: 1"

irb(main):004:0> sprintf_test 3
This should show the correct value: 3
This should also show the correct value:
3

irb(main):004:0> sprintf_test 4
This should show the correct value: 4
This should also show the correct value:
4
(irb):4: [BUG] Segmentation fault
ruby 2.0.0p195 (2013-05-14) [x86_64-darwin10.8.0]

[jeremyevans0 (Jeremy Evans)]

This is now expected behavior. %i starting in 2.0 means you need to provide a ruby object (VALUE) argument, not an integer. You need to switch from %li to %ld.

[bladenkerst (Brad Sumersford)]

I was trying to find confirmation of the change before posting this issue but I was unable to find a reference. Thank-you for clarifying, this can be closed.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Rejected