Skip to contentSkip to conversation

Fix uninitialized-variable warnings.

PATCH In review Han Shen 4 participants 9 messages Apr 11, 2013 – Apr 20, 2013 Networking drivers (wireless)

lore
Share & tools
lore

Han Shen

author
#0 lore raw

PATCH 1/1 Fix uninitialized-variable warnings.

GCC 4.8 is spitting out uninitialized-varaible warnings against "drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c". This trivial patch just adds initialization to the variable.

---
 drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +1 -1
index b793a65..875e36f 100644
@@ -669,7 +669,7 @@ static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw
 	u8 thermalvalue, delta, delta_lck, delta_iqk;
 	long ele_a, ele_d, temp_cck, val_x, value32;
 	long val_y, ele_c = 0;
-	u8 ofdm_index[2], ofdm_index_old[2], cck_index_old = 0;
+	u8 ofdm_index[2], ofdm_index_old[2] = {0, 0}, cck_index_old = 0;
 	s8 cck_index = 0;
 	int i;
 	bool is2t = IS_92C_SERIAL(rtlhal->version);
Signature · 1 line
1.8.1.3
#1 lore raw
Quoted 2 lines · Han Shen

GCC 4.8 is spitting out uninitialized-varaible warnings against "drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c". This trivial patch just adds initialization to the variable.

Not sure this is correct.
I think the logic should be inspected instead.

ofdm_index_old[1] _is_ used uninitialized.

It appears there might be a missing

	ofdm_index_old[1] = (u8) i;

as that's the same style test done in the
section above.

Maybe this:

---
 drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c | 1 +
 1 file changed, 1 insertion(+)
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +1 -0
index b793a65..56b2b2f 100644
@@ -717,6 +717,7 @@ static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw
 			for (i = 0; i < OFDM_TABLE_LENGTH; i++) {
 				if (ele_d == (ofdmswing_table[i] &
 				    MASKOFDM_D)) {
+					ofdm_index_old[1] = (u8) i;
 
 					RT_TRACE(rtlpriv, COMP_POWER_TRACKING,
 						 DBG_LOUD,
#2 lore raw
Quoted 26 lines · Joe Perches
Quoted 2 lines · Han Shen

GCC 4.8 is spitting out uninitialized-varaible warnings against "drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c". This trivial patch just adds initialization to the variable.

Not sure this is correct.
I think the logic should be inspected instead.

ofdm_index_old[1] _is_ used uninitialized.

It appears there might be a missing

	ofdm_index_old[1] = (u8) i;

as that's the same style test done in the
section above.

Maybe this:

---
  drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c | 1 +
  1 file changed, 1 insertion(+)
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +0 -0
index b793a65..56b2b2f 100644

drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw

Quoted 10 lines
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +1 -1
@@ -717,6 +717,7 @@ static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw
  			for (i = 0; i < OFDM_TABLE_LENGTH; i++) {
  				if (ele_d == (ofdmswing_table[i] &
  				    MASKOFDM_D)) {
+					ofdm_index_old[1] = (u8) i;
 
  					RT_TRACE(rtlpriv, COMP_POWER_TRACKING,
  						 DBG_LOUD,
 
 
--

Joe's patch would work, but as the elements of ofdm_index_old[] are just filled with the contents of i, just eliominate the middle man as in:

drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +1 -3
Index: wireless-testing-save/drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c
===================================================================
@@ -669,7 +669,7 @@ static void rtl92c_dm_txpower_tracking_c
        u8 thermalvalue, delta, delta_lck, delta_iqk;
        long ele_a, ele_d, temp_cck, val_x, value32;
        long val_y, ele_c = 0;
-       u8 ofdm_index[2], ofdm_index_old[2], cck_index_old = 0;
+       u8 ofdm_index[2], cck_index_old = 0;
        s8 cck_index = 0;
        int i;
        bool is2t = IS_92C_SERIAL(rtlhal->version);
@@ -700,12 +700,10 @@ static void rtl92c_dm_txpower_tracking_c
 
                for (i = 0; i < OFDM_TABLE_LENGTH; i++) {
                        if (ele_d == (ofdmswing_table[i] & MASKOFDM_D)) {
-                               ofdm_index_old[0] = (u8) i;
-
                                RT_TRACE(rtlpriv, COMP_POWER_TRACKING, DBG_LOUD,
                                         "Initial pathA ele_d reg0x%x = 0x%lx, 

ofdm_index=0x%x\n",
ROFDM0_XATXIQIMBALANCE,
- ele_d, ofdm_index_old[0]);

+                                        ele_d, i);
                                break;
                        }
                }
@@ -722,7 +720,7 @@ static void rtl92c_dm_txpower_tracking_c
                                                 DBG_LOUD,
                                                 "Initial pathB ele_d reg0x%x = 
0x%lx, ofdm_index=0x%x\n",
                                                 ROFDM0_XBTXIQIMBALANCE, ele_d,
-                                                ofdm_index_old[1]);
+                                                i);
                                        break;
                                }
                        }

I will submit this as a proper patch later. Fortunately, the driver is seldom run with the debug level as high as DBG_LOUD, then the RT_TRACE statements rarely result in any output.

Larry

Larry

#3 lore raw
In reply to Joe Perches
Quoted 26 lines · Joe Perches
Quoted 2 lines · Han Shen

GCC 4.8 is spitting out uninitialized-varaible warnings against "drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c". This trivial patch just adds initialization to the variable.

Not sure this is correct.
I think the logic should be inspected instead.

ofdm_index_old[1] _is_ used uninitialized.

It appears there might be a missing

	ofdm_index_old[1] = (u8) i;

as that's the same style test done in the
section above.

Maybe this:

---
  drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c | 1 +
  1 file changed, 1 insertion(+)
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +0 -0
index b793a65..56b2b2f 100644

drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw · line 723

Quoted 7 lines
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +1 -0
@@ -717,6 +717,7 @@ static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw
  			for (i = 0; i < OFDM_TABLE_LENGTH; i++) {
  				if (ele_d == (ofdmswing_table[i] &
  				    MASKOFDM_D)) {
+					ofdm_index_old[1] = (u8) i;
 
  					RT_TRACE(rtlpriv, COMP_POWER_TRACKING,
  						 DBG_LOUD,

My earlier E-mail was too hasty. Joe's patch is correct. It should be marked for stable (2.6.39+).

Larry

#4 lore raw

Hi, Joe's patch (the 1-line fix) does not work for me. The assignment is in a conditional block, when refer to this ofdm_index_old later, I still get error -

/mnt/host/source/src/third_party/kernel/files/drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:771:31:
error: 'ofdm_index_old[1]' may be used uninitialized in this function [-Werror=maybe-uninitialized]

     rtlpriv->dm.ofdm_index[i] = ofdm_index_old[i];

(I'm working on 3.4 branch.)

Thanks,

Quoted 50 lines · Larry Finger
Quoted 32 lines · Joe Perches
Quoted 4 lines · Han Shen

GCC 4.8 is spitting out uninitialized-varaible warnings against "drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c". This trivial patch just adds initialization to the variable.

Not sure this is correct.
I think the logic should be inspected instead.

ofdm_index_old[1] _is_ used uninitialized.

It appears there might be a missing

        ofdm_index_old[1] = (u8) i;

as that's the same style test done in the
section above.

Maybe this:

---
  drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c | 1 +
  1 file changed, 1 insertion(+)
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +0 -0
index b793a65..56b2b2f 100644
@@ -717,6 +717,7 @@ static void

rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw

Quoted 8 lines
                        for (i = 0; i < OFDM_TABLE_LENGTH; i++) {
                                if (ele_d == (ofdmswing_table[i] &
                                    MASKOFDM_D)) {

+ ofdm_index_old[1] = (u8) i;

RT_TRACE(rtlpriv,
COMP_POWER_TRACKING,
DBG_LOUD,

My earlier E-mail was too hasty. Joe's patch is correct. It should be marked for stable (2.6.39+).

Larry

Signature · 1 line
Han Shen |  Software Engineer |  [email protected] |  +1-650-440-3330
#5 lore raw
Quoted 10 lines · Han Shen(沈涵)

Hi, Joe's patch (the 1-line fix) does not work for me. The assignment is in a conditional block, when refer to this ofdm_index_old later, I still get error -

/mnt/host/source/src/third_party/kernel/files/drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:771:31:
error: 'ofdm_index_old[1]' may be used uninitialized in this function [-Werror=maybe-uninitialized]

      rtlpriv->dm.ofdm_index[i] = ofdm_index_old[i];

(I'm working on 3.4 branch.)

I would get rid of the -Werror=maybe-unitialized. Sometimes gcc generates a false warning, and you really should not error the build in that case.

The correct patch needs to initialize the array to 0, and set the element the way Joe's patch does.

Larry

#6 lore raw

Hi Larry, thanks. I've prepared below (simple) patch just adding initialization list and what Joe's patch does.

Is this ok?

H.

---
 drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
(quoted diff) +0 -0

b/drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c index b793a65..34150de 100644

drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c +0 -0
@@ -669,7 +669,7 @@ static void

rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw
u8 thermalvalue, delta, delta_lck, delta_iqk;
long ele_a, ele_d, temp_cck, val_x, value32;
long val_y, ele_c = 0;
- u8 ofdm_index[2], ofdm_index_old[2], cck_index_old = 0;
+ u8 ofdm_index[2], ofdm_index_old[2] = {0, 0}, cck_index_old = 0; s8 cck_index = 0;
int i;
bool is2t = IS_92C_SERIAL(rtlhal->version);
@@ -717,7 +717,7 @@ static void rtl92c_dm_txpower_tracking_callback_thermalmeter(struct ieee80211_hw
for (i = 0; i < OFDM_TABLE_LENGTH; i++) {
if (ele_d == (ofdmswing_table[i] &

     MASKOFDM_D)) {

-
+ ofdm_index_old[1] = (u8) i;
RT_TRACE(rtlpriv, COMP_POWER_TRACKING,
DBG_LOUD,
"Initial pathB ele_d reg0x%x = 0x%lx, ofdm_index=0x%x\n",
--
1.8.2.1

Quoted 24 lines · Larry Finger
Quoted 6 lines · Han Shen(沈涵)

Hi, Joe's patch (the 1-line fix) does not work for me. The assignment is in a conditional block, when refer to this ofdm_index_old later, I still get error -

Quoted 6 lines

/mnt/host/source/src/third_party/kernel/files/drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:771:31:
error: 'ofdm_index_old[1]' may be used uninitialized in this function [-Werror=maybe-uninitialized]

      rtlpriv->dm.ofdm_index[i] = ofdm_index_old[i];

(I'm working on 3.4 branch.)

I would get rid of the -Werror=maybe-unitialized. Sometimes gcc generates a false warning, and you really should not error the build in that case.

The correct patch needs to initialize the array to 0, and set the element the way Joe's patch does.

Larry

Signature · 1 line
Han Shen |  Software Engineer |  [email protected] |  +1-650-440-3330
#7 lore raw
Quoted 4 lines · Han Shen(沈涵)

I've prepared below (simple) patch just adding
initialization list and what Joe's patch does.

Is this ok?

No. It's whitespace damaged.

Try to make sure your patch applies by first sending it to yourself and then check to see if git am accepts it without warnings or errors.

#8 lore raw
In reply to Han Shen

2013/4/11 Han Shen <[email protected]>:

Quoted 2 lines

GCC 4.8 is spitting out uninitialized-varaible warnings against "drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c". This trivial patch just adds initialization to the variable.

Could you use prefixes for your patches? Like
rtl8192c: foo

Signature · 1 line
Rafał