Compile error: JSON::Value lacks overload for size_t (unsigned long) on MacOS

[charles-dyfis-net]

Below is one of many similar errors:

[6/193] Compiling C++ object MistInFLAC.p/src_input_input_flac.cpp.o
FAILED: MistInFLAC.p/src_input_input_flac.cpp.o
clang++ -IMistInFLAC.p -I. -I.. -I../subprojects/mbedtls/include -Isubprojects/libsrtp/include -I../subprojects/libsrtp/include -Imist -fcolor-diagnostics -Wall -Winvalid-pch -std=c++98 -O0 -g -funsigned-char -D_LARGEFILE_SOURCE -Wno-sign-compare -Wparentheses -Wno-non-virtual-dtor -Wno-strict-aliasing '-DAPPNAME="MistServer"' -DDEBUG=4 '-DRELEASE="Generic_aarch64"' '-DPACKAGE_VERSION="3.4"' -DSHM_DATASIZE=40 -DSTAT_CUTOFF=600 -DSTATS_DELAY=15 '-DUDP_API_HOST="localhost"' -DUDP_API_PORT=4242 -DSSL=1 -DUPDATER=1 -DHAVE_UPSTREAM_MBEDTLS_SRTP=0 -DWITH_DATACHANNELS '-DINPUTTYPE="input_flac.h"' -MD -MQ MistInFLAC.p/src_input_input_flac.cpp.o -MF MistInFLAC.p/src_input_input_flac.cpp.o.d -o MistInFLAC.p/src_input_input_flac.cpp.o -c ../src/input/input_flac.cpp
../src/input/input_flac.cpp:199:38: error: use of overloaded operator '=' is ambiguous (with operand types 'Value' and 'size_t' (aka 'unsigned long'))
    meta.inputLocalVars["blockSize"] = blockSize;
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~
mist/json.h:62:12: note: candidate function
    Value &operator=(const int64_t &rhs);
           ^
mist/json.h:63:12: note: candidate function
    Value &operator=(const int32_t &rhs);
           ^
mist/json.h:64:12: note: candidate function
    Value &operator=(const uint64_t &rhs);
           ^
mist/json.h:65:12: note: candidate function
    Value &operator=(const uint32_t &rhs);
           ^
mist/json.h:66:12: note: candidate function
    Value &operator=(const double &rhs);
           ^
mist/json.h:67:12: note: candidate function
    Value &operator=(const bool &rhs);
           ^
1 error generated.

[Thulinma]

Cross-referencing #215

[charles-dyfis-net]

Would making this code conditional with #ifdef __APPLE__ work?

[Thulinma]

It would make it work on Apple, sure (assuming that Apple behaves the same on x86 and arm - which I'm not sure of, honestly). But there are probably other platforms with the same problem, so a comparison of the width of size_t and uint64_t (and possibly also uint32_t, for 32-bit platforms) would be more robust and less Apple-specific.

[charles-dyfis-net]

My concern there is that I don't understand sizeof() to be available at preprocessor time (admittedly, though, it's been a few decades since I wrote C full time, so my memory is fuzzy).

[Thulinma]

Yup, you're absolutely right. However, thankfully Meson is capable of compiling and running "code snippets" to do these kind of tasks. So we can totally run such a snippet to see if the extra functions are needed or not, and then define a variable based on the outcome.
If this sounds a little too complex for you, no worries, I have been planning to make this change at some point in the near-ish future anyway. Just need to get some other things out of the way first (paying customer requests get priority and all that).

[Thulinma]

My colleague @stronk-dev (whom is also a Mac user) has volunteered to handle this actually! Expect it to land shortly ^_^