given following code:
~/clang-test-> cat test.c
#include <stdio.h>
int main (void) {
int a = 0, b;
a = b + a;
printf ("a = %s", a);
return 0;
}
/clang-test-> gcc -Wall -Wno-switch -Wdisabled-optimization -Wpointer-arith -Wredundant-decls -Wno-pointer-sign -Wcast-qual -Wwrite-strings -O3 -fno-math-errno -std=c99 test.c
test.c: In function ‘main’:
test.c:8: warning: format ‘%s’ expects type ‘char *’, but argument 2 has type ‘int’
test.c:7: warning: ‘b’ is used uninitialized in this function
~/clang-test-> ccc -Wall -Wno-switch -Wdisabled-optimization -Wpointer-arith -Wredundant-decls -Wno-pointer-sign -Wcast-qual -Wwrite-strings -O3 -fno-math-errno -std=c99 test.c
~/clang-test-> gcc -Wall -std=c99 test.c test.c: In function ‘main’:
test.c:8: warning: format ‘%s’ expects type ‘char *’, but argument 2 has type ‘int’
~/clang-test-> gcc -Wall -O3 test.c
test.c: In function ‘main’:
test.c:8: warning: format ‘%s’ expects type ‘char *’, but argument 2 has type ‘int’
test.c:7: warning: ‘b’ is used uninitialized in this function
~/clang-test-> ccc -Wall -O3 -std=c99 test.c
~/clang-test->
Also clang warns on the following but gcc doesn't??
libavcodec/dsputil.c:177:8: warning: use of unary operator that may be intended as compound assignment (-=)
end=-1;
^~
libavcodec/dsputil.c:3966:10: warning: use of unary operator that may be intended as compound assignment (-=)
for(i=-len, j=len-1; i<0; i++, j--) {
^~
Hi,
clang does not yet implement the uninitialized values warning outside
of the static analysis engine.
test.c: In function ‘main’:
test.c:8: warning: format ‘%s’ expects type ‘char *’, but argument 2
There's a TODO in the code to add this check.
has type ‘int’
test.c:7: warning: ‘b’ is used uninitialized in this function
clang doesn't have a warning like this as part of the default
compilation process, at least for the moment. There is a separate
analysis module which prints warnings like this, though; see
http://clang.llvm.org/StaticAnalysis.html.
libavcodec/dsputil.c:3966:10: warning: use of unary operator that may
be intended as compound assignment (-=)
This warning triggers on a lot of stuff... we should probably disable
it by default.
-Eli
Daniel,
I think this is a matter of mapping -Wuninitialized to running the same logic as -warn-unit-values. That logic is just a simple flow-sensitive check.
Thanks Daniel,
May be there should note on how a programmer would do the compile
and analysis, So they won't get confused. I mean do I compile and then
analyze.
anyway I ran the analyze command
on test.c and it only gave one warning. may be ccc is not passing -analyze command
properly.
~/clang-test-> cat test.c
#include <stdio.h>
#include <stdlib.h>
int main (void) {
int a = 0;
int b;
a = b + a;
printf ("a = %d", a);
exit(0);
}
~/clang-test-> ccc --analyze test.c
test.c:9:3: warning: Pass-by-value argument in function call is undefined.
printf ("a = %d", a);
^ ~
1 diagnostic generated.
If I comment out the printf statement then I also only get one warning. So something is wrong.
~/clang-test-> ccc --analyze test.c
test.c:8:3: warning: Value stored to 'a' is never read
a = b + a;
^ ~~~~~
1 diagnostic generated.
P.S. I would expect -Wall to cover all warning but I guess I am asking too much.
test.c: In function ‘main’:
test.c:8: warning: format ‘%s’ expects type ‘char *’, but argument 2
There's a TODO in the code to add this check.
has type ‘int’
test.c:7: warning: ‘b’ is used uninitialized in this function
clang doesn't have a warning like this as part of the default
compilation process, at least for the moment. There is a separate
analysis module which prints warnings like this, though; see
http://clang.llvm.org/StaticAnalysis.html.
Right, note that GCC only produces "dataflow warnings" when optimizations are enabled though.
libavcodec/dsputil.c:3966:10: warning: use of unary operator that may
be intended as compound assignment (-=)
This warning triggers on a lot of stuff... we should probably disable
it by default.
What sorts of cases does it trigger on that are false positives?
-Chris
Here are some examples taken from ffmpeg-export-2008-12-18
s->inter_quant_bias=-(1<<(QUANT_BIAS_SHIFT-2)); //(a - x/4)/x
for(change=-1; change<=1; change+=2){
next_dir=-1;
for(mv=-(16<<f_code); mv<(16<<f_code); mv++){
c=d=e=-1;
pred_table[0]=-2;
lpc[i][j]=-m[(pass-1)&1].coeff[i][j];
Here are some examples taken from ffmpeg-export-2008-12-18
These are all exactly the sorts of things that the warning was designed to catch. Are spaces really that horrible to use? 
-Chris
Trying to dictate a coding style using compiler warnings isn't
productive. I like the suggestion that we should warn on "=- " (with
a trailing space), but not not just "=-"; the false positive rate with
the trailing space should be almost zero, and the people using a
coding style that puts spaces around operators keep the benefit of the
warning.
-Eli
Here are some examples taken from ffmpeg-export-2008-12-18
These are all exactly the sorts of things that the warning was designed to
catch. Are spaces really that horrible to use?
Trying to dictate a coding style using compiler warnings isn't
productive.
I completely disagree. Warnings are *all about* (optionally, when enabled) forcing coding standards on people. Warnings are inherently about code that is valid but dubious.
I like the suggestion that we should warn on "=- " (with
a trailing space), but not not just "=-"; the false positive rate with
the trailing space should be almost zero, and the people using a
coding style that puts spaces around operators keep the benefit of the
warning.
That is a great suggestion, implemented!
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20090302/013618.html
-Chris
fyi,
this example fails your new test.
libavcodec/mpeg12enc.c:756:19: warning: use of unary operator that may be intended as compound assignment (-=)
for(mv=-MAX_MV; mv<=MAX_MV; mv++){
^~
Please include a self-contained .c file.
-Chris