On 09/19/2014 06:40 AM, Christophe Fergeau wrote:
Any idea in which situation we are trying to access uninitialized pixmap memory?
I no longer recall with precision. The valgrind trigger was in hash_and_copy; I suspect I first encountered it in the case I fixed with a later patch (copying past the stride).
However, a prepare_access that doesn't fill the region, or the new tuning code I've added to dfps has a clear logical need for a preinitialized pixmap, so I think this is correct for a variety of reasons.
Cheers, Jeremy
Apart from that, why not. Christophe On Fri, Sep 12, 2014 at 02:20:36PM -0500, Jeremy White wrote:Signed-off-by: Jeremy White <jwhite@xxxxxxxxxxxxxxx> --- src/dfps.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/src/dfps.c b/src/dfps.c index 5db34dc..568ba5e 100644 --- a/src/dfps.c +++ b/src/dfps.c @@ -312,6 +312,17 @@ static void dfps_set_screen_pixmap (PixmapPtr pixmap) pixmap->drawable.pScreen->devPrivate = pixmap; } +static void dfps_clear_pixmap(PixmapPtr pixmap, int w, int h) +{ + GCPtr pgc; + pgc = GetScratchGC(pixmap->drawable.depth, pixmap->drawable.pScreen); + if (pgc) + { + fbFill(&pixmap->drawable, pgc, 0, 0, w, h); + FreeScratchGC(pgc); + } +} + static PixmapPtr dfps_create_pixmap (ScreenPtr screen, int w, int h, int depth, unsigned usage) { PixmapPtr pixmap; @@ -324,7 +335,10 @@ static PixmapPtr dfps_create_pixmap (ScreenPtr screen, int w, int h, int depth, pixmap = fbCreatePixmap (screen, w, h, depth, usage); if (pixmap) + { + dfps_clear_pixmap(pixmap, w, h); dfps_set_info(pixmap, info); + } else free(info); -- 1.7.10.4 _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel