News for hackers. Stuff that matters.
Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by mehrdada on Saturday June 14, 2014 @03:52PM (#7893522)
(Shameless plug:) It would be great if someone would leverage the pretty successful technique (i.e. found a couple hundred bugs in GCC and Clang) we recently devised to detect miscompilations, Compiler Validation via Equivalence Modulo Inputs[1], to test ARM compilers.

In short, the idea is: profile a program running on a specific input; mutate the unexecuted code with respect to that input; compile and execute the mutated program and expect the output to be identical given the input.

[1]: http://mehrdadafshari.com/emi/paper.pdf

Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by zsu on Tuesday June 17, 2014 @12:24PM (#7905300)
Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by QuantumChaos on Sunday June 15, 2014 @03:15AM (#7894653)
Very nice. One suggestion, I found the abstract and introduction difficult to understand, and only got the real point by page 3. I would add something clearer in the abstract and/or introduction, such as:

"Static analysis is used to construct alternative versions of the program, that give identical outputs for a given subset of inputs I. The alternative programs are compiled and executed, and we test whether they do actually give the same outputs as the original program, on the input data I. By restricting to a subset I of inputs, a greater variety of alternatives can be produced, than if the alternatives had to be identical to the original program for all inputs."

Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by jacquesm on Saturday June 14, 2014 @04:11PM (#7893564)
That is extremely clever.
Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by joevandyk on Saturday June 14, 2014 @11:26PM (#7894272)
Just curious, how long did that paper take to write (including research time)?
Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by mehrdada on Saturday June 14, 2014 @11:43PM (#7894305)
Over a year, I'd say. We'd run the test infrastructure for ~11 months on a couple machines when we submitted the paper.

That's not indicative of much though, as the actual timelines for writing papers can be driven by conference deadlines more than the works themselves (when talking about granularity of six months to a year).

Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by teoryn on Sunday June 15, 2014 @12:26AM (#7894372)
Have you or will you be releasing code for Orion?
Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by mehrdada on Sunday June 15, 2014 @12:54AM (#7894442)
We have not, so far. It's a bit difficult to package up as it mostly consists of a bunch of (potentially environment dependent) shell scripts with a transformation program. Shouldn't be too hard to replicate, as the total LOC is pretty low.

Do you have a specific use for it? If so, let me know about it, as I might be able to try lobbying for packaging and releasing the code or work something out.

Re:Autopsy of GCC 4.6.3 compiler bug on Raspberry Pi by teoryn on Sunday June 15, 2014 @12:19PM (#7895555)
Nah, I'd just like to take a look at it, but you might be able to get someone else to use it on ARM compilers if you release it.