gcc-4.2.3 on Hardy/ia32 miscompiles simple code

[John Regehr]

Binary package hint: gcc-4.2

This is bizarre... gcc-4.2.3 for Hardy on ia32 compiles the code below into a program that prints 1. The correct answer is 0, and this is what all other compilers (including gcc-4.1 and gcc-3.4 on Hardy/ia32) return.

Furthermore, FSF gcc-4.2.3 also gives the correct answer!

Probably someone should look into this as comparison errors like this could easily have security implications.

#include <stdio.h>

int func_1 (void)
{
  signed char l_11 = 1;
  unsigned char l_12 = -1;
  return (l_11 > l_12);
}

int main (void)
{
  printf ("%d\n", func_1());
  return 0;
}

[Andreas Moog (ampelbein)]

I can confirm this with
gcc:
  Installed: 4:4.2.3-1ubuntu6

[Colin Watson (cjwatson)]

Confirmed with gcc-4.2 4.2.3-2ubuntu7 in Hardy. However, 4.2.4-3ubuntu2 in Intrepid does *not* exhibit this bug.

[Colin Watson (cjwatson)]

I've further confirmed that 4.2.3-4ubuntu1 (an older version in Intrepid) did not exhibit this bug.

[Colin Watson (cjwatson)]

See also bug 237134.

[Matthias Klose (doko)]

test packages are now available in the ubuntu-toolchain PPA.
see https://launchpad.net/~ubuntu-toolchain/+archive

[Martin Pitt (pitti)]

Accepted into -proposed, please test and give feedback here. Please see https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Cesare Tirabassi (norsetto)]

Works for me. I also tested all my sources with it and didn't notice any regression.

[eben (erauhut)]

My test cases now pass. Thanks!

[Steve Beattie (sbeattie)]

gcc-4.2 4.2.4-1ubuntu3, which included a fix to address this issue, was released to hardy-updates on 2008-10-22. Closing this bug.