On Thu, Nov 22, 2018 at 12:13:27AM +0000, Ramsay Jones wrote: > > > On 21/11/2018 22:53, Luc Van Oostenryck wrote: > > On Mon, Nov 19, 2018 at 08:54:38PM +0000, Ramsay Jones wrote: > >> From 8764999aa78415e217fd3106a8c8518a5b40c20c Mon Sep 17 00:00:00 2001 > >> From: Ramsay Jones <ramsay@xxxxxxxxxxxxxxxxxxxx> > >> Date: Sun, 18 Nov 2018 23:52:23 +0000 > >> Subject: [PATCH 9/9] constant: add -Wconstant-size warning > > > > Since the problem is not really a problem of size, nor a problem of > > type, but a mismatch between the real type of the constant and the > > type indicated by its suffix (or the lack of a suffix), I propose > > to call this warning -Wconstant-suffix. > > Sounds good to me. (I am bad at naming things). ;-) I think it's often surprinsingly hard and in this case far from obvious. > >> diff --git a/sparse.1 b/sparse.1 > >> index 3e13523..26299b3 100644 > >> --- a/sparse.1 > >> +++ b/sparse.1 > >> @@ -101,6 +101,16 @@ Sparse issues these warnings by default. To turn them off, use > >> \fB\-Wno\-cast\-truncate\fR. > >> . > >> .TP > >> +.B \-Wconstant-size > >> +Warn if an integer constant is larger than the maximum representable value > >> +of the type indicated by the type suffix (if any). For example, on a 64-bit > >> +system, the constant 0xfffff00000000000U is larger than can be represented > >> +by an int and so requires an unsigned long type suffix. (So, in this case, > >> +the warning could be suppressed by using the UL type suffix). > > > > I don't 100% agree with the use of 'requires'. The type of the constant is > > determinated by its value, no suffix is required (but the lack of an 'L' > > in the suffix may make believe that the type is 'unsigned int' or may > > force you to count the zeroes, ...). > > Hmm, the standard is a bit ambiguous in its wording. For example, (C99) > in 6.4.4.1-2, it states "An integer constant begins with a digit, but > has no period or exponent part. It may have a prefix that specifies > its base and a suffix that _specifies its type_." Whereas, later in > 6.4.4.1-5, it states "The type of an integer constant is the first of > the corresponding list in which its _value can be represented_." > (followed by a table of types). > > So, this warning is about the discrepancy in the two 'type indicators' > (the type suffix and the actual value of the constant). BTW, my use of > 'requires' above was more about avoiding the warning ... ;-) OK. Yes, it's a bit ambiguous. I rarely find the standard very clear anyway :). > > I propose to reword the example by something like: > > +Warn if an integer constant is larger than the maximum representable value > > +of the type indicated by its type suffix (if any). For example, on a > > +system where ints are 32-bit and longs 64-bit, the constant \fB0x100000000U\fR > > +is larger than can be represented by an \fBunsigned int\fR but fits in an > > +\fBunsigned long\fR. So its type is \fBunsigned long\fR but this is not > > +indicated by its suffix. In this case, the warning could be suppressed by > > +using the suffix \fBUL\fR: \fB0x100000000UL\fR. > > OK. > > > > > Another solution would be to simply remove the example. > > > > I could go with that too, but how would you describe the warning > without an example! ;-) Yes, it's better with the example. Thanks, -- Luc
