valgrind-users

[Valgrind-users] uninitialised value(s) - after explicit memset to 0 ??

[David C. R. <dra...@su...>]

All,

   I have an issue with valgrind-3.8.1 (openSuSE 13.1 x86_64) reporting the error:

Conditional jump or move depends on uninitialised value(s)

   The error stems from reallocation of a new block of memory. The error is 
triggered when iterating over the reallocated pointer array into the reallocated 
block, despite explicitly setting/initializing values of the new pointers to 0 
with memset (equivalent to the original allocation with calloc).

   A quick example program illustrating the issue is here:

[2.9k]
http://www.3111skyline.com/dl/dev/lists/vg_ui_values.c.txt

   .txt added for online readability. For direct d/l use:

http://www.3111skyline.com/dl/dev/lists/vg_ui_values.c

   It simply reads any text file and reallocates pointers as needed after 4 
lines are read. Compile with -DVGERR to reproduce the error. A default compile 
avoids the error by using a for loop to avoid touching what valgrind considers 
uninitialized. So in summary, compile as follow to test with/without the error:

With Error:

  gcc -Wall -Wextra -DVGERR -o vg_ui_error vg_ui_values.c

Without:

  gcc -Wall -Wextra -o vg_ui_clean vg_ui_values.c

Then simply run:

   valgrind ./vg_ui_error some_text_file

(where the text file has more than 4 lines of text)

   Since the values for the new block contents is set with memset, why does 
valgrind complain that the memory is uninitialized? Is this correct behavior for 
valgrind, or is valgrind for some reason not tracking the values set by memset 
as initialized?

   I've looked at 
http://valgrind.org/docs/manual/mc-manual.html#mc-manual.uninitvals, etc.. and I 
am unclear whether valgrind should throw this error? ...or... whether valgrind 
should correctly track the memset pointers as initialized? What say the experts?

-- 
David C. Rankin, J.D.,P.E.

Re: [Valgrind-users] uninitialised value(s) - after explicit memset to 0 ??

[David C. <dcc...@ac...>]

On 3/10/2015 5:21 PM, David C. Rankin wrote:
> All,
>
>     I have an issue with valgrind-3.8.1 (openSuSE 13.1 x86_64) reporting the error:
>
> Conditional jump or move depends on uninitialised value(s)
>
>     The error stems from reallocation of a new block of memory. The error is
> triggered when iterating over the reallocated pointer array into the reallocated
> block, despite explicitly setting/initializing values of the new pointers to 0
> with memset (equivalent to the original allocation with calloc).

Not quite.  calloc() gets the size of the object being cleared in 
addition to the number of objects, while memset() gets only the size in 
bytes.  You need to multiply cmax by sizeof(*slines), as you did for the 
realloc() call.  (Of course, "slines + cmax" on the same line of code is 
pointer arithmetic, so it automatically performs this multiplication.  C 
does have its warts.)

-- 
     David Chapman      dcc...@ac...
     Chapman Consulting -- San Jose, CA
     Software Development Done Right.
     www.chapman-consulting-sj.com

Re: [Valgrind-users] uninitialised value(s) - after explicit memset to 0 ??

[David C. R. <dra...@su...>]

On 03/10/2015 08:07 PM, David Chapman wrote:
> On 3/10/2015 5:21 PM, David C. Rankin wrote:
>> All,
>>
>>     I have an issue with valgrind-3.8.1 (openSuSE 13.1 x86_64) reporting the
>> error:
>>
>> Conditional jump or move depends on uninitialised value(s)
>>
>>     The error stems from reallocation of a new block of memory. The error is
>> triggered when iterating over the reallocated pointer array into the reallocated
>> block, despite explicitly setting/initializing values of the new pointers to 0
>> with memset (equivalent to the original allocation with calloc).
>
> Not quite.  calloc() gets the size of the object being cleared in addition to
> the number of objects, while memset() gets only the size in bytes.  You need to
> multiply cmax by sizeof(*slines), as you did for the realloc() call.  (Of
> course, "slines + cmax" on the same line of code is pointer arithmetic, so it
> automatically performs this multiplication.  C does have its warts.)
>

Hah!, Thank you David!

(smacks self, places pointed circular hat on head, and turns sadly to face the 
corner ;-)

   Sheeze, I can't believe I walked right into that one! Thanks for solving a 
'forest for the trees issue.'  I could have looked at that for 2-days and never 
snapped to the omission.

   Thanks

-- 
David C. Rankin, J.D.,P.E.