[m-dev.] More <inttypes.h> problems :-(

[Fergus Henderson]

On 07-Jun-1999, Warwick Harvey <wharvey at cs.monash.edu.au> wrote:
> 
> I appear to have discovered a more fundamental flaw in the inttypes, 
> sys/types, stdint changes, which means one cannot (currently) compile the 
> asm_fast, etc. grades on Red Hat 6.0 Linux systems (it uses asm_jump 
> instead).
> 
> This problem may affect any systems on which Mercury uses a system-supplied 
> header file to defined types.
> 
> On Red Hat 6.0 systems, the inttypes.h header file contains inline 
> functions, which means any subsequent declaration of global registers will 
> fail.  This is why mercury_imp.h starts with a big warning saying that it 
> must be the first header file included, and in particular before any system 
> header files.  Unfortunately, mercury_regs.h, which (indirectly) declares 
> the global registers first includes mercury_types.h to define the 
> appropriate types.  But if inttypes.h or similar exist, mercury_types.h 
> includes it, breaking the requirement that no system header files be 
> included first.

Ah.  That is an unfortunate problem.

One possible remedy for this problem is to use the gcc option
`-fixed-<REG>' to reserve the appropriate registers.
This would avoid the need for the requirement that the global
register declarations precede any #includes of system header
files.

> May I ask what the rationale for using these files was?  The compiler has to 
> cope with them not existing (i.e. it defines its own types if it can't use 
> ones defined by the system), and there have been a number of problems with 
> the use of them on a number of systems already, so I'm just wondering what 
> they gain us
> (the simplest solution to the above problems seems to be to abandon the use 
> of the inttypes.h file).

Well, the idea was to use the standard names, if available, and to
emulate them, if not.  There are some advantages in using the standard
names, in that people reading our source code will know immediately
what those names mean.  If we abandon the use of the standard headers,
then we would need to also abandon the use of the standard names for
these types, and instead use names with `MR_' prefixes.

Unfortunately using the standard headers has turned out to be more difficult
than we first thought.  And if we just insert `MR_' prefixes, but otherwise
keep the names the same as in the standard header files, then the additional
learning/readability burden is not very high.  So that may be the best
solution.

-- 
Fergus Henderson <fjh at cs.mu.oz.au>  |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>  |  of excellence is a lethal habit"
PGP: finger fjh at 128.250.37.3        |     -- the last words of T. S. Garp.
--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------