On Mon, 24 Jan 2005, Orna Agmon wrote:

> Valgrind is not always correct. It sometimes reports errors which do not
> exists - they have no reason to exist and other memory checkers (such as
> third) do not find them.

that, ofcourse, is no proof - the other checker might be too limited, or
have a bug, causing it not to find something that valgrind found. lets not
be too absolute here.

> But:
>
> Usually, when valgrind finds such a non-real bug, it usually indicates
> that there is a serious bug somewhere else.
>
> So the bug may not be listed there, and you may think you can overlook
> some errors, but clearing all the error messages will still solve your
> problem.

or it won't. it could be that his bug stems from a logical problem. or
that the memory smearing is inside an object (e.g. accessing the wrong
array element - it's not a memory smear from a system point of view). or
perhaps there are also other problems.

it's not a good idea to trust valgrind absolutely. after all - valgrind is
still a computer software, and as such - tends to have bugs itself ;)

use it, by all means - but don't assume that it'll necessarily find all
the bugs, or even all memory management bugs.

-- 
guy

"For world domination - press 1,
 or dial 0, and please hold, for the creator." -- nob o. dy

=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to