garmin int->float->int fidelity

[tsteven4]

clang gives us a warning:

[33/134] Building CXX object CMakeFiles/gpsbabel.dir/garmin.cc.o
/__w/gpsbabel/gpsbabel/garmin.cc:468:25: warning: implicit conversion from 'int' to 'float' changes value from 2147483647 to 2147483648 [-Wimplicit-const-int-float-conversion]
        (way[i]->alt == INT_MAX) ||
                     ~~ ^~~~~~~
/usr/lib/llvm-14/lib/clang/14.0.0/include/limits.h:46:19: note: expanded from macro 'INT_MAX'
#define INT_MAX   __INT_MAX__
                  ^~~~~~~~~~~
<built-in>:102:21: note: expanded from here
#define __INT_MAX__ 2147483647
                    ^~~~~~~~~~
1 warning generated.

The more interesting part of this doesn't generate a warning,

gpsbabel/jeeps/gpsmem.cc

Line 216 in 7088c7e

ret->smbl = ret->dspl = ret->colour = ret->alt = ret->prot = INT_MAX;

Here smbl, dspl, colour, prot are all int32, but alt is a float. The result of this assignment is that smbl, dspl, colour are set to -2147483648(0x80000000) while prot is set to 2147483647(0x7fffffff), and alt is 2.14748e+09(0x4f000000). I don't know that this causes any trouble, but I doubt that was the intent.

[robertlipe]

Probably worse, that was defined in a Garmin spec in the early 90s when ints were just always 32 bits in their world. Now with 64 bits being the (non-embedded) norm, this is probably wrong. I doubt that the actual bit pattern in the packecrACTUALLY ever had anything to do with host INTMAX. So probably a. Saved track on a Garmin II on a 64 bit machine incurs sadness. There was a time that our alt was an integer just because sub-meter precision of altitude is actually still pretty rare. Of course, our "I don't really have an altitded" value and handling doesn't have a great story. I suspect you're right and this never actually worked as intended. Feel free to replace with values from Garmin specs, intmax-1 (does that work? Is it right on 64 bit ints?) Or even just palin delete it. Our pool of Non-gpx garmins is small and only shrinking.

…

On Fri, Jun 30, 2023, 12:42 PM tsteven4 ***@***.***> wrote: clang gives us a warning: [33/134] Building CXX object CMakeFiles/gpsbabel.dir/garmin.cc.o /__w/gpsbabel/gpsbabel/garmin.cc:468:25: warning: implicit conversion from 'int' to 'float' changes value from 2147483647 to 2147483648 [-Wimplicit-const-int-float-conversion] (way[i]->alt == INT_MAX) || ~~ ^~~~~~~ /usr/lib/llvm-14/lib/clang/14.0.0/include/limits.h:46:19: note: expanded from macro 'INT_MAX' #define INT_MAX __INT_MAX__ ^~~~~~~~~~~ <built-in>:102:21: note: expanded from here #define __INT_MAX__ 2147483647 ^~~~~~~~~~ 1 warning generated. The more interesting part of this doesn't generate a warning, https://github.com/GPSBabel/gpsbabel/blob/7088c7e47ace23799c0b3477f39af4d158b68b30/jeeps/gpsmem.cc#L216 Here smbl, dspl, colour, prot are all int32, but alt is a float. The result of this assignment is that smbl, dspl, colour are set to -2147483648(0x80000000) while prot is set to 2147483647(0x7fffffff), and alt is 2.14748e+09(0x4f000000). I don't know that this causes any trouble, but I doubt that was the intent. — Reply to this email directly, view it on GitHub <#1134>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCSD32S3CEJ3IAYMABF6ITXN4FZ3ANCNFSM6AAAAAAZ2EE22Q> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>