Thanks, it does indeed resolve the issue.
Actually, your update is exactly the same fix that I tried myself. :)
Tony
On Mon, Mar 7, 2011 at 4:14 PM, Dominique Orban
<dom...@gm...> wrote:
> On Tue, Mar 1, 2011 at 2:25 PM, Tony Stillfjord
> <Ton...@na...> wrote:
>> Hello,
>>
>> I seem to have found a memory leak in the Umfpack module, more
>> specifically in the function factorize (umfpackmodule.c).
>> This calls the function newUMFPackObject, in which (as far as I
>> understand it) memory is allocated for the variable colidx
>> via SpMatrix_LLMatBuildColIndex(&colidx, matrix, 1). This memory is never freed.
>>
>> Calling SpMatrix_LLMatDestroyColIndex(&colidx) towards the end of
>> newUMFPackObject seems to resolve the issue, but I
>> am uncertain where it is safe to make this call. I would guess that
>> one should call it just before the two switch statements
>> that check for problems with the matrix and return NULL if something
>> is awry. colidx does not seem to be in use at this point,
>> but someone with more insight into the code than me should confirm this.
>
> Hi Tony,
>
> Good catch! I pushed an update to the SVN repository. Let us know if
> that resolves the problem for you.
>
> --
> Dominique
>
|
