[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: Valgrind hit in kio_data [PATCH]
From:       Till Adam <adam () kde ! org>
Date:       2005-02-25 7:16:19
Message-ID: 200502250816.25941.adam () kde ! org
[Download RAW message or body]


On Friday 25 February 2005 00:38, George Staikos wrote:
> On Thursday 24 February 2005 18:24, Till Adam wrote:
> > > > Invalid read of size 4
> > > >    at 0x451EFE50: KMMoveCommand::execute() (kmmsgbase.h:126)
> > > >    by 0x9C5FD9DF: ???
> > > >  Address 0x356B265C is 4 bytes inside a block of size 32 free'd
> > > >    at 0x3414CCFA: operator delete(void*) (vg_replace_malloc.c:156)
> > > >    by 0x450AEB17: KMMsgInfo::~KMMsgInfo() (kmmsginfo.cpp:131)
> > > >
> > > > without any photos. But no one replied.
> > >
> > >    This one is unrelated to the above, and I have it fixed locally, but
> > > it now triggers more crashes elsewhere.  There are invalid pointers all
> > > over the place.  I'm testing more fixes now.
> >
> > That one is typical of one of our biggest problems in KMail, the fact
> > that message pointers can become invalid at any time. A pointer to a
> > KMMsgInfo is replaced by a pointer to a KMMessage if the message is
> > "gotten", which happens for example when it is selected in the GUI. This
> > invalidates all pointers to the message currently being held. That is the
> > reason why we are slowly migrating operations to work on serial numbers
> > and not pointers and are thinking about how to do this properly for KDE 4
> > (where the grass is green, and the girls are pretty). In other words,
> > known problem, we are fixing issues as we find them, but they are many
> > and often hard to nail down.
>
>   As a temporary workaround could we not lock the folder while syncing is
> happening, and prevent move/copy then?

Yes, but there is no such mechanism implemented yet.

Till

[Attachment #3 (application/pgp-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]