#441 Replacements for SDCC macro

[Philipp Klaus Krause]

When compiling using --std-sdccxx, sdcc defines a macro SDCC defined the conctenation of the version numbers (major, minor, p).
The macro name makes is incompatible with the C standard. And comparing it as an integer is problematic when any version number is above 9.
It should thus be deprecated, and replacements introduced.
I suggest:
__SDCC_VERSION, defined to MAJOR_VERSION.MINOR_VERSION
and
__SDCC_VERSION_MAJOR
__SDCC_VERSION_MINOR
__SDCC_VERSION_P

Philipp

[Frieder Ferlemann]

http://sourceforge.net/p/predef/wiki/Compilers/ has a huge collection.

The GCC scheme seems not bad:)

Greetings,

[Ben Shi]

I like Borland C style,

# define _SDCC_VERSION_MAJOR_ 3
# define _SDCC_VERSION_MINOR_ 5
# define _SDCC_VERSION_PATCH_ 0
# define _SDCC_VERSION_ (((_SDCC_VERSION_MAJOR_) << 8) + _SDCC_VERSION_MINOR_)

Because I thought int can be more efficient than float on 8-bit MCUs.

[Philipp Klaus Krause]

I assumed that users would mostly use the version at preprocessing time, so float vs. int wouldn't matter.
But now I think we should provide just the macros for major, minor and p version, and users can make their own version macro, float or int from it as they like.

Philipp

[Philipp Klaus Krause]

Since I don't know how useful the patch version is, I left it out for now. Macros for major and minor version are in as of revision [r9261].

Philipp

[Frieder Ferlemann]

Hi Philip,

I feel this might not have been discussed thoroughly enough.

The GCC macros (as per the link given above) are

__GNUC__             Version
__GNUC_MINOR__       Revision
__GNUC_PATCHLEVEL__  Patch (introduced in version 3.0)

They seem very reasonable to me and allow things like:

#define __GNUC_VERSION__ (__GNUC__ * 10000 \
                          + __GNUC_MINOR__ * 100 \
                          + __GNUC_PATCHLEVEL__)

or:

#define __GNUC_VERSION__ ((__GNUC__ << 16)  \
                           + (__GNUC_MINOR__ << 8) \
                           + __GNUC_PATCHLEVEL__)

(so it can easily be compared/printed as integer. In the second example even without using integer division for the deeply embedded systems)

Why not follow the GCC scheme and use

__SDCC__
__SDCC_MINOR__
__SDCC_PATCHLEVEL__

As we use GCC's precompiler using GCC's convention seems a natural choice to me.

The PATCHLEVEL is useful to have even if there is no immediate use for it.

If e.g. a there's a bug with (or below) a specific patch level of SDCC which only surfaces later then a #warning (or a workaround) can be used without having to test #ifdef __SDCC_PATCHLEVEL__ first. (Without patchlevel the user could have to wait about a year until the ).

Greetings,
Frieder

(hopefully Sourceforge doesn't garble the strings. If it does: my proposal is to exchange "GNUC" by "SDCC" in the Macros within the link given in my first post.)

[Philipp Klaus Krause]

I don't like that the names of the gcc macros do not clearly communicate what they are. And I don't like the name of the major version macro being inconsistent with the names of the other two.
I don't mind adding another macro for the patchlevel. That should allow everything that the gcc macros do.

Philipp

[Maarten Brock]

I suggest to add __SDCC_VERSION_PATCH as well.

[Philipp Klaus Krause]

Implemented in revision #9273.

Philipp

[Frieder Ferlemann]

thanks, closed for me too.

Buffer sizes with 20 bytes are a little on the short side. Please apply something like the appended patch.

[Philipp Klaus Krause]

On the other hand, 32 seems plenty for now. I increased it to 32 in revision #9274.

Philipp