How would you resolve this symbol conflict?

[metaperl.j]

Compilation error:

make
gcc -O -c -o a.o a.c
In file included from j.h:39,
from a.c:14:
je.h:191: error: 'nan' redeclared as different kind of symbol
/usr/include/math.h:177: error: previous declaration of 'nan' was here

je.h declares nan as extern D where type is typedef double D

math.h declares nan as:
extern double nan _PARAMS((const char *));

nan is only used in a few places in the code:

find . -type f -print0 | xargs -0 -e grep -nH -e nan
../i.c:77: MC(&nan,XNAN,(size_t)sizeof(D));
../j.c:31 nan;
../je.h:191:extern D nan;
../je.h-:193:extern D nan;
../wn.c:20: if('-'==*s&&3>n){x=1==n?infc=*(1+s),c=='-')?-inf:c=='.'?
nan:0; if(x){*v=x; R 1;}}

So I'm wondering what technique would be best to rename all the nan
calls in the code to something else to avoid conflicts. Perhaps some
sort of #define is in order.

 

[Ben Bacarisse]

Compilation error:

make
gcc -O -c -o a.o a.c
In file included from j.h:39,
from a.c:14:
je.h:191: error: 'nan' redeclared as different kind of symbol
/usr/include/math.h:177: error: previous declaration of 'nan' was here

je.h declares nan as extern D where type is typedef double D

math.h declares nan as:
extern double nan _PARAMS((const char *));

nan is only used in a few places in the code:

find . -type f -print0 | xargs -0 -e grep -nH -e nan
./i.c:77: MC(&nan,XNAN,(size_t)sizeof(D));
./j.c:31 nan;
./je.h:191:extern D nan;
./je.h-:193:extern D nan;
./wn.c:20: if('-'==*s&&3>n){x=1==n?infc=*(1+s),c=='-')?-inf:c=='.'?
nan:0; if(x){*v=x; R 1;}}

You may want to remove the two useless spaces in that line that are
helping readability at the expense of wasted bytes. Also:

if(x)R *v=x,1;

might work to remove one set of those pesky {}s.

Seriously, you may want to reconsider your coding guidelines.

So I'm wondering what technique would be best to rename all the nan
calls in the code to something else to avoid conflicts. Perhaps some
sort of #define is in order.

I'd rename your nan (since it occurs so infrequently) rather the
trying to move the standard one off to one side.

 

[metaperl.j]

You may want to remove the two useless spaces in that line that are
helping readability at the expense of wasted bytes.

This is not my code. It is an early open source version of the J
programming language:
ftp://ftp.gaertner.de/pub/j/j-v7-src.tgz

I'd rename your nan (since it occurs so infrequently) rather the
trying to move the standard one off to one side.

Ok. thanks. I will do that.

 

[Eric Sosman]

Compilation error:

make
gcc -O -c -o a.o a.c
In file included from j.h:39,
from a.c:14:
je.h:191: error: 'nan' redeclared as different kind of symbol
/usr/include/math.h:177: error: previous declaration of 'nan' was here

je.h declares nan as extern D where type is typedef double D

math.h declares nan as:
extern double nan _PARAMS((const char *));
[...]

It's not supposed to do that; <math.h> is not allowed to
declare `nan' at all. The gcc compiler is famous/notorious
for compiling "C with extras" in its default mode of operation;
try running it in a conforming mode with

gcc -ansi -pedantic ...
or
gcc -std=whatever -pedantic ...

.... and see if that suppresses the gratuitous `nan' declaration.

 

[Ben Bacarisse]

Compilation error:

make
gcc -O -c -o a.o a.c
In file included from j.h:39,
from a.c:14:
je.h:191: error: 'nan' redeclared as different kind of symbol
/usr/include/math.h:177: error: previous declaration of 'nan' was here

je.h declares nan as extern D where type is typedef double D

math.h declares nan as:
extern double nan _PARAMS((const char *));
[...]

It's not supposed to do that; <math.h> is not allowed to
declare `nan' at all.

Eh? double nan(const char *); is in C99. I can't see how it could
avoid declaring it but, since I have not know you to be wrong, I await
being corrected!

 

[Eric Sosman]

Ben Bacarisse wrote On 11/12/07 09:52,:

Compilation error:

make
gcc -O -c -o a.o a.c
In file included from j.h:39,
from a.c:14:
je.h:191: error: 'nan' redeclared as different kind of symbol
/usr/include/math.h:177: error: previous declaration of 'nan' was here

je.h declares nan as extern D where type is typedef double D

math.h declares nan as:
extern double nan _PARAMS((const char *));
[...]

It's not supposed to do that; <math.h> is not allowed to
declare `nan' at all.

Eh? double nan(const char *); is in C99. I can't see how it could
avoid declaring it but, since I have not know you to be wrong, I await
being corrected!

Ben, I really wish you'd stop arguing with me when
I have egg all over my face. Just because I'm wrong,
R-O-N-G wrong, is no excuse for you to be right.

(In other words, the reports of my infallibility have
been greatly exaggerated. Thanks for the correction.)