Re: [E-devel] valgrind regrets to report...

[David S. <on...@gm...>]

On Wed, 11 Jan 2006 11:11:50 +0900 Carsten Haitzler (The Rasterman)
<ra...@ra...> wrote:

> On Wed, 11 Jan 2006 04:36:02 +1000 David Seikel <on...@gm...>
> babbled:
>=20
> > Since all, or most modules use the same source file name of
> > e_mod_main.c, tracking down that last one took some time.
> > Naturally I had to look at all of them before I found one that
> > matched the report. It's engage.

<snip>=20

> initialized) etc. the only ones that worry me are the last 2 in the
> report - but there is nothing to identify the actual module there...
> hmmm.

As I said, I tracked it down to the engage module.  There is enough
information in the report to say that it is engage.  It gives file name,
line number, and function name, if that line number is not in that
function in that file, then it can't be that module.  Of the modules I
was running at the time, only engage matched.

> > The naughty null Ecore_Timer free is still reported during E
> > shutdown. Hope this helps to track it down.
>=20
> valgrind wont help find that - to find these do
>=20
> export ECORE_ERROR_ABORT=3D1
> gdb enlightenment
>=20
> (ie set that env var) - then ecore will call abort() on such errors
> and bingo - insta-backtrace :)

I'll give that a go after I finish with my email.