Test 63_typedef fails about picoc

[hellerve]

This is related to the size of int64/uint64 on your machine. The number within angle brackets designates the size of the number. It is expected to be 8 bytes, but is 4 bytes in size.

The tests expects a 64bit version OS, I guess. unsigned long is declared to be uint64_t at the top of the test, which does not hold true for every architecture.

Bottom line: The test is not failing because the software is not correct, but because of false internal premises.

Disclaimer: I am not affiliated with this software in any way, I just found your issue and thought I might clear things up a bit.

[e8johan]

Yup, definitely something broken with the 64bit types. I just assumed that the interpreter would execute in a run-time that was independent of sizeof(unsigned long) for the host architecture. I don't have time to dig deeper here right now, but I guess the type definitions for the interpreter run-time, and not necessarily for the host platform, is the issue.

[hellerve]

This is actually standard-conformant behaviour. unsigned long is not guaranteed to be 64 bits in any way. If you just typedef uint64_t as that, you will have a bad time.

[e8johan]

Yup, the test is broken. I had a look in the code, and it seems that TypeUnsignedLong is defined in interpreter.h. Then in type.c, TypeSizeValue, an assumption is made that sizeof(unsigned long) <= sizeof(ALIGN_TYPE), which happens to be void * (defined in platform.h). Thus, it is broken.

I did a fugly fix to the test (using sizeof(x64) and sizeof(u64) to fake the output on non-64-bit platform). I'm preparing the pull request right now.

[e8johan]

This is the fix mentioned in the previous comment: #8 .

[galacticstudios]

I have a fork that fixes this problem (galacticstudios@abd7473). It implements picoc's 'long' as an int64_t, so that picoc's long will be 64 bits on every platform. I've tested on Windows (MS Visual C++ 2013 or newer) and on Linux. I've also fixed a bunch of compiler warnings on MSVC.

[cgjordan]

Anyone know if/when these fixes will make it into master?
I ran into this issue today on my Raspberry Pi 3 B, running 32-bit Raspbian 8.

root@raspberrypi:~# uname -r -m -o
4.4.13-v7+ armv7l GNU/Linux

[zsaleeba]

Hey galacticstudios, want to send me a pull request with your fixes?

[galacticstudios]

I've sent a pull request - I think. I haven't done it before so let me know if I did it wrong.

• Bob