serious compiler warnings in mingw64 build of master

serious compiler warnings in mingw64 build of master

[Stephen Leake]

I just updated emacs/master (to
bf55b5ac17cd5a40ad5ff2e25af7e050602180bb), and it's giving these
compiler warnings:

   CC       process.o
In file included from process.c:33:
process.c: In function 'Fmake_process':
lisp.h:1561:31: warning: null pointer dereference [-Wnull-dereference]
 1561 |   return XSTRING (string)->u.s.data;
      |          ~~~~~~~~~~~~~~~~~~~~~^~~~~

This is from line 1990 in process.c:
	  new_argv[i] = SSDATA (XCAR (tem));

I have not tried to figure out a fix.


   CC       w32.o
w32.c: In function '_sys_read_ahead':
w32.c:8858:10: warning: 'rc' may be used uninitialized in this function [-Wmaybe-uninitialized]
 8858 |   if (rc == sizeof (char))
      |       ~~~^~~~~~~~~~~~~~~~

The preceding if/elsif block has no final 'else', so rc may not be set.
I'm not clear if the right fix is to init rc to 0, or to add an 'else'
that does something like report an error.


I've been getting mysterious crashes when mail spawns a process to get
more mail, so I'm hoping these are the root cause of that.

-- 
-- Stephe

Re: serious compiler warnings in mingw64 build of master

[Eli Zaretskii]

> From: Stephen Leake <stephen_leake@stephe-leake.org>
> Date: Sat, 14 Aug 2021 10:18:56 -0700
> 
> I just updated emacs/master (to
> bf55b5ac17cd5a40ad5ff2e25af7e050602180bb), and it's giving these
> compiler warnings:
> 
>    CC       process.o
> In file included from process.c:33:
> process.c: In function 'Fmake_process':
> lisp.h:1561:31: warning: null pointer dereference [-Wnull-dereference]
>  1561 |   return XSTRING (string)->u.s.data;
>       |          ~~~~~~~~~~~~~~~~~~~~~^~~~~
> 
> This is from line 1990 in process.c:
> 	  new_argv[i] = SSDATA (XCAR (tem));
> 
> I have not tried to figure out a fix.

Neither did I.

>    CC       w32.o
> w32.c: In function '_sys_read_ahead':
> w32.c:8858:10: warning: 'rc' may be used uninitialized in this function [-Wmaybe-uninitialized]
>  8858 |   if (rc == sizeof (char))
>       |       ~~~^~~~~~~~~~~~~~~~
> 
> The preceding if/elsif block has no final 'else', so rc may not be set.

This is bogus: the compiler isn't smart enough to understand that the
3 possible flag bits are the only ones that can happen: a descriptor
is either for a pipe, or a serial connection, or a socket, because
_sys_read_ahead is only ever used for reading process output.  I added
an initialization to pacify the compiler.

> I've been getting mysterious crashes when mail spawns a process to get
> more mail, so I'm hoping these are the root cause of that.

Unlikely, because this code hasn't changed in eons.