Philipp Stephani wrote:

 > * sysdep.c (system_process_attributes): Remove unused variables.

Thanks, I installed that in your name (1st attached patch).


> * emacs.c (using_utf8): Don't assume anything about mbstate_t type.

The old code doesn't assume anything about the type either, and should work 
regardless of how mbstate_t is implemented. The proposed change makes the code a 
bit harder to read, so I'd rather avoid it if possible. I couldn't reproduce the 
warning that you're evidently seeing. Perhaps you can work around the problem in 
the same way I worked around the problem with floats and doubles (see below)?


> * emacs-module.c (MODULE_SETJMP_1): Mark dummy variable as unused.

Rather than do that, let's just use the variable; that's more robust. I 
installed the 2nd attached patch.

> * lread.c (string_to_number): Use constants of double type.
> * editfns.c (decode_float_time):
> * fns.c (make_hash_table, maybe_resize_hash_table)
> (Fhash_table_rehash_size, Fhash_table_rehash_threshold):
> Explicitly cast floating-point values.

There should be no problem (i.e., no loss of info) converting a float to a 
double, so these changes are not needed. I assume you're using clang. I 
reproduced its false alarms on Fedora 25 x86-64, which uses clang 3.9.1, and 
installed the 3rd attached patch to work around the problem.


 > * fileio.c (file_name_case_insensitive_p): Add cast.

Rather than waste time static-checking the DARWIN_OS_CASE_SENSITIVE_FIXME == 2 
code let's just #ifdef it out. I did that in the 4th attached patch. Maybe we 
should just remove it, since nobody is using it and (as you note) it doesn't 
work anyway.