[Mono-devel-list] Compiler Warnings..

[Willibald Krenn]

Hi!

AFAIK mono is currently built using
Wall -Wunused -Wmissing-prototypes -Wmissing-declarations 
-Wstrict-prototypes -Wmissing-prototypes -Wnested-externs 
-Wpointer-arith -Wno-cast-qual -Wcast-align -Wwrite-strings
as arguments for gcc.
(At least this is partly correct, as haven't checked if these args are 
used everywhere..)

I did a build in my private tree using the parameters above and logged 
all warnings gcc emitted: This file is over 1400 lines. Sure, some of 
the warnings are from code that isn't found in official mono and 
probably lots of these warnings come from code that mono does not build 
using the full warning parameter set from above, but that still leaves a 
substantial number..

If I build my private tree with a set of
-g -O2 -DGC_LINUX_THREADS -D_GNU_SOURCE -D_REENTRANT -DUSE_COMPILER 
                                         _TLS (i.e. no warnings at all) 
and after applying a patch to elimiate some int32<->pointer64 warnings, 
I still get warnings in metadata/loader.c, reflection.c, 
debug-mono-symfile.c, mono-debug-debugger.c and one in mini.c: warning: 
#warning "fixme: add opcode emulation"

Question is: Why's mono even built using such a strict warning set, when 
no-one seems to care? What warning level is considered to be 'standard' 
and useful?


To my first real patch: Attached to this mail you'll find some 
modifications to eliminate gcc warnings on int32<->pointer64 casts. Note 
that there are some interface changes too and I considere this version 
as NOT ready for including in SV, but I need your comments in 
correcting/developing this patch further.

I've introduced one additional global header file called intps.h: It's 
sole purpose is to define an unsigned integer that is guaranteed to be 
the same size as a pointer: uintps. (unsigned int pointer size)

#ifndef _intps_H
#define _intps_H

	#include "config.h"
	#include <glib.h>
	/*definition of an unsigned int datatype that is always
          sizeof(pointer)*/

	#if SIZEOF_VOID_P == 8
	typedef guint64 uintps;
	#else
	typedef guint32 uintps;
	#endif

#endif

Using this type as a basis, I've introduced casts wherever gcc (without 
all warnings!) complained about integer size not matching pointer size..
But see attached diff.. Again, I do not consider it ready to commit, 
cause I haven't run all tests and did not re-check if new warnings are 
emitted when compiling with the strict warning set outlined above.

Thanks,
  Willi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: clean64.diff
Type: text/x-diff
Size: 22183 bytes
Desc: not available
Url : http://lists.ximian.com/pipermail/mono-devel-list/attachments/20041201/1702a18c/attachment.bin