Re: [DynInst_API:] the PRIx64 commit breaks the build with some <inttypes.h>

[Tim Haines <thaines.astro@xxxxxxxxx>]

Hi, Mark.

I built the head of master on all of our usual test platforms, and only found this isshttps://github.com/dyninst/dyninst/blob/e22e663ed73bfe507fcd56b72ee13043634ae4c1/instructionAPI/h/Result.hue on Power8 with gcc-7.3.1 (Power9 with gcc-7.3.1 was fine, so that's fun!).

Honestly, the PR* macros aren't the correct solution to this problem. I probably should have forced a code change before merging this, but it had already been tested by the person who submitted the PR, and it was blocking their build cycle. The correct solution is to use to use std::to_string introduced in C++11. I'll make a PR for that and re-test.

Thanks.

Â- Tim

On Tue, Aug 6, 2019 at 3:35 PM Mark W. Krentel <krentel@xxxxxxxx> wrote:

One of my cron jobs tripped an alarm for Dyninst master on powerpc
this morning.

This one is bizarre, but the last commit about the PRIx64 macros
breaks the build on some systems, depending on the exact version of
the <inttypes.h> header file.

Here is a unit test that reproduces the problem. This is a snippet
from instructionAPI/h/Result.h, near line 453.

print.cpp:

#include <sstream>
#include <string.h>
#include <inttypes.h>

int
main(int argc, char **argv)
{
  Âchar hex[20];
  Âint64_t val = 123;

  Âsnprintf(hex, 20, "%" PRIx64, val);

  Âprintf("ans = %s\n", hex);
  Âreturn 0;
}

On one power8 system with glibc-headers-2.17-157.el7.ppc64le,

$ g++ -g -O -std=c++11 print.cpp

print.cpp: In function 'int main(int, char**)':
print.cpp:11:27: error: expected ')' before 'PRIx64'
   snprintf(hex, 20, "%" PRIx64, val);
              ^~~~~~

Building Dyninst, I get the same error in Result.h near line 453.

I had to diff <inttypes.h> across systems to figure out what's going
on. The powerpc version has the following. The x86_64 version does
not have the #if on lines 44-46.

 Â44 /* The ISO C99 standard specifies that these macros must only be
 Â45  Âdefined if explicitly requested. */
 Â46 #if !defined __cplusplus || defined __STDC_FORMAT_MACROS
 Â47
 Â48 # if __WORDSIZE == 64
 Â49 # define __PRI64_PREFIX    "l"
 Â50 # define __PRIPTR_PREFIX   Â"l"
 Â51 # else
 Â52 # define __PRI64_PREFIX    "ll"
 Â53 # define __PRIPTR_PREFIX
 Â54 # endif

So, on powerpc, things like PRIx64 are either not defined or else not
defined the way you want them.

I'm not sure if this is a glibc-headers version issue, or an x86 vs
ppc issue. It seems to be more common on powerpc, but older RH 6.x
x86_64 systems also have the same problem.

Anyway, I'll let you decide the best and most portable way of handling
this. Maybe you just want to define __STDC_FORMAT_MACROS and test
that it doesn't break anything else.

Blech. My kingdom for portability!

--Mark


_______________________________________________
Dyninst-api mailing list
Dyninst-api@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api