[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
On Sun April 19 2009, Evert Glebbeek wrote:
> On 19 Apr 2009, at 16:27 , Evert Glebbeek wrote:
> > It's probably related to the assertions I'm getting:
>
> Ok, maybe not. I don't know.
>
> This is the problem I get:
>
> al_install_system() tries to read the configuration file before
> initialising the system.
> To read the configuration file, al_fopen() is called.
> Because the file I/O interface is now a per-thread setting, al_fopen()
> calls into TLS to retrieve the file I/O driver.
> Because TLS is not initialised (on OS X) until the system has been
> initialised, the returned file I/O driver is NULL, triggering the
> assert.
There is no reason to have per thread vtables. and all it'll do is make io
that much slower (tls != fast). I don't know why it was changed.
> There is also an associated problem of the TLS file I/O driver never
> being set on OS X, but the code aborts before it gets there.
> I'm going to work around the issue for the time being by simply
> returning a pointer to the stdio driver if TLS has not been
> initialised, but we should rethink the logic here.
>
> Can we also go back to testing patches like this on ALL supported
> platforms before committing? Would have saved me half the afternoon.
That would be nice. I always tried to make sure allegro worked when
committing.
> Evert
>
> ---------------------------------------------------------------------------
>--- Stay on top of everything new and different, both inside and
> around Java (TM) technology - register by April 22, and save
> $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco.
> 300 plus technical and hands-on sessions. Register today.
> Use priority code J9JMT32. http://p.sf.net/sfu/p
--
Thomas Fjellstrom
tfjellstrom@xxxxxxxxxx