Fix for "unsafe use of type bool" warning when compiling with MSVC.

[amubiera]

Using negate on a bool causes a warning when compiling with MSVC 19.X:

littlefs\lfs.c(2561): error C2220: the following warning is treated as an error
littlefs\lfs.c(2561): warning C4804: '-': unsafe use of type 'bool' in operation

[geky-bot]

Tests passed ✓, Code: 17112 B (+0.0%), Stack: 1448 B (+0.0%), Structs: 812 B (+0.0%)

	Code	Stack	Structs		Coverage
Default	17112 B (+0.0%)	1448 B (+0.0%)	812 B (+0.0%)	Lines	2429/2591 lines (-0.0%)
Readonly	6222 B (+0.0%)	448 B (+0.0%)	812 B (+0.0%)	Branches	1280/1610 branches (+0.0%)
Threadsafe	17964 B (+0.0%)	1448 B (+0.0%)	820 B (+0.0%)		Benchmarks
Multiversion	17184 B (+0.0%)	1448 B (+0.0%)	816 B (+0.0%)	Readed	29369693876 B (+0.0%)
Migrate	18776 B (+0.0%)	1752 B (+0.0%)	816 B (+0.0%)	Proged	1482874766 B (+0.0%)
Error-asserts	17884 B (+0.0%)	1440 B (+0.0%)	812 B (+0.0%)	Erased	1568888832 B (+0.0%)

[geky]

Would it make more sense to change this to:

err = lfs_fs_preporphans(lfs, -(int8_t)hasparent);

?

[amubiera]

That works as well. Was just worried that on some compiler, a true bool could be all 0xFF and -0xFF would not turn out to be -1.

However, the C standard states that a true bool converts to 1:

4 Standard conversions
An rvalue of type bool can be converted to an rvalue of type int, with
false becoming zero and true becoming one.

I also checked Godbolt compiler explorer and both -(int8_t)hasparent and hasparent ? -1 : 0 produce the same assembly in an optimized build. But slower in debug builds.

I pushed a new version with -(int8_t)hasparent)

Cheers!

[geky]

Thanks for this!

I was curious, for some reason the ternary operator cost a bit more. You would think GCC would be smart enough to emit the same code for these:

-hasparent              17104 B
hasparent ? -1 : 0      17112 B (+0.0%)
-(int8_t)hasparent      17104 B (+0.0%)

Not a deal-breaker, but very curious...

That works as well. Was just worried that on some compiler, a true bool could be all 0xFF and -0xFF would not turn out to be -1.

This was a real concern in the wild-west before stdbool.h. Fortunately, littlefs only supports up to C99, which standardized stdbool.h and the _Bool type.

This has been enough of an issue though that we ended up adding an explicit assert:

littlefs/lfs.c

Lines 4209 to 4213 in 0494ce7

	// check that bool is a truthy-preserving type
	//
	// note the most common reason for this failure is a before-c99 compiler,
	// which littlefs currently does not support
	LFS_ASSERT((bool)0x80000000);

Related discussion: #772

[geky-bot]

Tests passed ✓, Code: 17104 B (+0.0%), Stack: 1448 B (+0.0%), Structs: 812 B (+0.0%)

	Code	Stack	Structs		Coverage
Default	17104 B (+0.0%)	1448 B (+0.0%)	812 B (+0.0%)	Lines	2429/2591 lines (-0.0%)
Readonly	6222 B (+0.0%)	448 B (+0.0%)	812 B (+0.0%)	Branches	1278/1608 branches (+0.0%)
Threadsafe	17956 B (+0.0%)	1448 B (+0.0%)	820 B (+0.0%)		Benchmarks
Multiversion	17176 B (+0.0%)	1448 B (+0.0%)	816 B (+0.0%)	Readed	29369693876 B (+0.0%)
Migrate	18768 B (+0.0%)	1752 B (+0.0%)	816 B (+0.0%)	Proged	1482874766 B (+0.0%)
Error-asserts	17884 B (+0.0%)	1440 B (+0.0%)	812 B (+0.0%)	Erased	1568888832 B (+0.0%)

[geky]

Looks good here, thanks for creating a PR!

Will bring this in on the next patch release.