including two gperf-generated .c files

[Jim Meyering]

Brendan Kehoe <address@hidden> wrote:
>> As you probably know, the sourceforge gperf-bugs list is useless.
>> All it gets is spam.  And gperf-list has had no traffic for 4(!) years.
>>
>> Have you considered moving the sources and mailing list to savannah?
>> There, the lists are essentially guaranteed to be spam-free,
>> and you can use git.
>
> I thought about the above but put off the idea since there didn't seem
> to be enough activity to make it worth the effort.  Of course
> savannah's improved since then, so perhaps it was my first
> impression...

Hi Brendan,

That makes sense.  But if you decide to move to git, I can help.

BTW, recently I included two gperf-generated .c files
into one, but got redefinition errors on a handful of
macros as well as the "hash" function.  Here's the
kludge I had to use to avoid those errors:

    #define hash(a,b) hash_attr_lookup (a, b)
    #include "attr-lookup.c"
    #undef hash

    /* Un-define the following so that subsequent of the next gperf-generated
       .c file doesn't evoke redefinition errors.  */
    #undef TOTAL_KEYWORDS
    #undef MIN_WORD_LENGTH
    #undef MAX_WORD_LENGTH
    #undef MIN_HASH_VALUE
    #undef MAX_HASH_VALUE

    #include "col-lookup.c"

It would be nice if those #undef directives were
emitted automatically into each .c file.

Also nice would be an option to specify a new name for the
"hash" function.  You could even use the name specified via
--lookup-function-name=NAME as a prefix for the associated "hash"
function.  Then there'd be no need for a new option.

FYI, here are the options I'm using, followed by the commands
used to generated the two files mentioned above:

GPERF_OPTIONS = \
  --language=ANSI-C \
  --readonly-tables \
  --no-strlen \
  --struct-type \
  --size-multiple 2 \
  --key-positions='*'

$(GPERF) $(GPERF_OPTIONS) --lookup-function-name=attr_lookup $<
$(GPERF) $(GPERF_OPTIONS) --lookup-function-name=col_lookup $<