מכתב אמיתי

Ragowan, Danny wrote:

| Hello Shachar ,
| Hope you are well.. it has been a while since we last have communicated ;-))
|
|
Yes, it was.

| Just following up with you to see whether all is in order with regards to
| Compuware software ..
|
Nothing new has transpired. Almost nothing new, that is.

A while back I was stuck while developing an open source project, with an unexplained access violation. I got a demo from you, which totally and utterly failed to spot the problem. I decided that much though I like the idea behind the Compuware tools, it is not worth the money.

Since then I managed to find out the cause of the problem that haunted me back then. It turns out to be a bug in ATL. The details are available at http://gborg.postgresql.org/pipermail/oledb-dev/2004-November/000155.html.

Quite frankly, I expected bounds checker to be able to tell me more about the bug. If it told me “you tried to access memory area X, which was allocated here and freed here”, the entire search would have been so much shorter. I realize that this is not a technically simple task to achieve, but as things stand Bounds Checker failed the usefulness test as far as I’m concerned, and so I don’t think I will be acquiring it. I should point out that open source project “valgrind” does correctly detect such errors and give useful output, and without performing any compile time instrumentation. I would have used it here, except it only has a Linux version. Being open source, however, there is nothing stopping you from applying the valgrind technique for Windows.

Here is an example of a program exhibiting the precise same type of error as happened in ATL, which caused the access violation we were trying to debug using bounds checker.

#include 
#include 

int main()
{
    const char **a;

    a=malloc(sizeof(*a));

    *a="Hello";
    free(a);
    printf("%sn", *a);

    return 0;
}

Running with valgrind:

==17356== Memcheck, a memory error detector for x86-linux.
==17356== Copyright (C) 2002-2005, and GNU GPL'd, by Julian Seward et al.
==17356== Using valgrind-2.4.0.rc3, a program supervision framework for x86-linux.
==17356== Copyright (C) 2000-2005, and GNU GPL'd, by Julian Seward et al.
==17356== For more details, rerun with: -v
==17356==
==17356== Invalid read of size 4
==17356==    at 0x804842A: main (valgrindtest.c:12)
==17356==  Address 0x1BA58028 is 0 bytes inside a block of size 4 free'd
==17356==    at 0x1B904B04: free (vg_replace_malloc.c:152)
==17356==    by 0x8048426: main (valgrindtest.c:11)
Hello
==17356==
==17356== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 13 from 1)
==17356== malloc/free: in use at exit: 0 bytes in 0 blocks.
==17356== malloc/free: 1 allocs, 1 frees, 4 bytes allocated.
==17356== For counts of detected errors, rerun with: -v
==17356== No malloc'd blocks  no leaks are possible.

| Pls let me know if I can be of any assistance to you concerning our products
| at this time
|
|
Not at this time, no. As you can see, your open source competition is doing a better job at it than you.

| Always at your service!
| Best regards ,
|
|
Shachar


Shachar Shemesh
Lingnu Open Source Consulting ltd.
Have you backed up today’s work? http://www.lingnu.com/backup.html