Various fixes reported by Clang

GIT PULL Erwan Velu 2 participants 2 messages Jul 22, 2013 – Jul 23, 2013

Share & tools

Erwan Velu

author

2013-07-22 21:56 UTC #0 lore raw

Hey,

I had the curiosity to check how fio was considered by clang. The "--analyze" option of clang is something levering interesting bugs.

In fact, there is many false positive but some were real.
I've been pushing a branch featuring the associated fixes. That's almost nothing but I cannot close my eyes on it ;o)

Cheers,
Erwan

The following changes since commit d7e30e61f2c1d9c6101cc6896009750aae2e2bcf:

fio2gnuplot: Don't plot fake data (2013-07-22 19:39:24 +0200)

are available in the git repository at:

[email protected]:enovance/fio.git clang

for you to fetch changes up to 2fdbefdd43968f3bf354a488288b9794b5f5c6bb:

core: Insure "il" is initialized in all cases (2013-07-22 23:48:48 +0200)

----------------------------------------------------------------
Erwan Velu (3):
core: Removing duplicated code
core: Don't leak memory if error is unknown
core: Insure "il" is initialized in all cases

 init.c    | 2 --
 options.c | 1 +
 parse.c   | 2 +-
 stat.c    | 6 ++----
 4 files changed, 4 insertions(+), 7 deletions(-)

init.c +0 -2

index 70b56e3..d808eb6 100644
@@ -1755,8 +1755,6 @@ int parse_cmd_line(int argc, char *argv[], int client_type)
         case FIO_GETOPT_IOENGINE: {
             const char *opt = l_opts[lidx].name;
             char *val = optarg;
-            opt = l_opts[lidx].name;
-            val = optarg;
             ret = fio_cmd_ioengine_option_parse(td, opt, val);
             break;
         }

options.c +1 -0

index 6462c67..1c44f42 100644
@@ -271,6 +271,7 @@ static int ignore_error_type(struct thread_data *td, int etype, char *str)
         if (!error[i]) {
             log_err("Unknown error %s, please use number value \n",
                   fname);
+            free(error);
             return 1;
         }
         i++;

parse.c +0 -0

index f54dae6..d3eb2c4 100644
@@ -380,7 +380,7 @@ static int str_match_len(const struct value_pair *vp, const char *str)
 static int __handle_option(struct fio_option *o, const char *ptr, void

*data,
int first, int more, int curr)
{
- int il, *ilp;

+    int il=0, *ilp;
     fio_fp64_t *flp;
     long long ull, *ullp;
     long ul1, ul2;

stat.c +2 -4

index 332ccd0..442caa0 100644
@@ -243,15 +243,13 @@ out:
 int calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max,
          double *mean, double *dev)
 {
-    double n = is->samples;
+    double n = (double) is->samples;
 
-    if (is->samples == 0)
+    if (n == 0)
         return 0;
 
     *min = is->min_val;
     *max = is->max_val;
-
-    n = (double) is->samples;
     *mean = is->mean.u.f;
 
     if (n > 1.0)

Jens Axboe

2013-07-23 02:56 UTC #1 lore raw

Quoted 8 lines · Erwan Velu

Hey,
I had the curiosity to check how fio was considered by clang. The "--analyze" option of clang is something levering interesting bugs.
In fact, there is many false positive but some were real. I've been pushing a branch featuring the associated fixes. That's almost nothing but I cannot close my eyes on it ;o)

Thanks, all three look good. Thanks for wading through the false positives to find the gold :-)

Signature · 1 line

Jens Axboe