<div dir="ltr">Sure, I'll give it a try later. Currently I am working on my day job. </div><div class="gmail_extra"> <div class="gmail_quote">On Fri, Nov 3, 2017 at 4:25 PM, Russell King - ARM Linux <<a href="mailto:linux@armlinux.org.uk" target="_blank">linux@armlinux.org.uk</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Fri, Nov 03, 2017 at 12:05:48PM +0000, Luís Mendes wrote: > So this is what I get with picture desc on the candidate corner cases: > Nov 3 11:57:04 picolo xf86_armada[786]: A: clip[x1=0,y1=0], > vSrc[w=1024,h=768], drawable[x=0,y=0], src_offset[x=0,y=0} > Nov 3 11:57:04 picolo xf86_armada[786]: A: Pict: 0x20e45a8: 32 > (1024x768+0+0) R fmt XRGB8888 I have my own case with mate here: etnaviv_accel_Composite(height=1, width=1, yDst=0, xDst=0, yMask=0, xMask=0, ySrc=-1, xSrc=-1, pDst=0x962b08, pMask=0x0, pSrc=0x962ab8, op=PictOpSrc) pSrc = {pDrawable = 0x86cf48, pFormat = 0x59bd78, format = PICT_x8r8g8b8, refcnt = 1, id = 0, repeat = 0, graphicsExposures = 0, subWindowMode = 1, polyEdge = 0, polyMode = 0, freeCompClip = 1, componentAlpha = 0, repeatType = 0, filter = 0, stateChanges = 0, unused = 0, pNext = 0x0, alphaMap = 0x0, alphaOrigin = {x = 0, y = 0}, clipOrigin = {x = 0, y = 0}, clientClip = 0x0, serialNumber = 4071, pCompositeClip = 0x962b58, devPrivates = 0x0, transform = 0x0, pSourcePict = 0x0, filter_params = 0x0, filter_nparams = 0} pSrc->pDrawable = {type = 0, class = 1, depth = 24, bitsPerPixel = 32, id = 27262979, x = 1334, y = 0, width = 1, height = 25, pScreen = 0x599708, serialNumber = 4071} => DRAWABLE_WINDOW pSrc::Pixmap = {drawable = {type = 1, class = 0, depth = 24, bitsPerPixel = 32, id = 0, x = 0, y = 0, width = 1360, height = 27, pScreen = 0x599708, serialNumber = 499}, devPrivates = 0x84bd0c, refcnt = 3, devKind = 5440, devPrivate = {ptr = 0x84bd48, val = 8699208, uval = 8699208, fptr = 0x84bd48}, screen_x = 0, screen_y = 0, usage_hint = 536870914, master_pixmap = 0x0} So, what's being asked for is to copy a single pixel from -1,-1 on the source picture to 0,0 on the destination picture. The source picture is a window, with no repeat. The backing pixmap for the window is 1360x27. The window is located at 1334,0 on this pixmap. So, -1,-1 is off the top of the window, and off the top of the pixmap. >From what I can tell from the Xrender spec (<a href="https://www.x.org/releases/current/doc/renderproto/renderproto.txt" rel="noreferrer" target="_blank">https://www.x.org/releases/current/doc/renderproto/renderproto.txt</a>), this should mean, as repeat mode is "None", that the source should be treated as "transparent". Hmm, let's try to trap the case where the pixels we're asked for are off the pixmap, and use the fallback. Can you try this patch please? diff --git a/etnaviv/etnaviv_render.c b/etnaviv/etnaviv_render.c index 15cafdf..e846a23 100644 --- a/etnaviv/etnaviv_render.c +++ b/etnaviv/etnaviv_render.c @@ -125,24 +125,27 @@ static void etnaviv_debug_blend_op(const char *func, * determine whether the picture repeat flag is meaningful. The * rectangle must have had the transformation applied. */ -static Bool picture_needs_repeat(PicturePtr pPict, int x, int y, +static Bool picture_has_pixels(PicturePtr pPict, int x, int y, unsigned w, unsigned h) { DrawablePtr pDrawable; - if (!pPict->repeat) - return FALSE; - pDrawable = pPict->pDrawable; if (!pDrawable) + return FALSE; + + if (!drawable_contains(pDrawable, x, y, w, h)) + return FALSE; + + if (!pPict->repeat) return TRUE; if (pPict->filter != PictFilterConvolution && (pDrawable->width > 1 || pDrawable->height > 1) && drawable_contains(pDrawable, x, y, w, h)) - return FALSE; + return TRUE; - return TRUE; + return FALSE; } static const struct etnaviv_blend_op etnaviv_composite_op[] = { @@ -425,8 +428,8 @@ static struct etnaviv_pixmap *etnaviv_acquire_src(ScreenPtr pScreen, if (!transform_is_integer_translation(pict->transform, &tx, &ty)) goto fallback; - if (picture_needs_repeat(pict, src_topleft->x + tx, src_topleft->y + ty, - clip->x2, clip->y2)) + if (!picture_has_pixels(pict, src_topleft->x + tx, src_topleft->y + ty, + clip->x2, clip->y2)) goto fallback; src_topleft->x += drawable->x + src_offset.x + tx; @@ -661,8 +664,8 @@ static int etnaviv_accel_composite_masked(PicturePtr pSrc, PicturePtr pMask, mask_offset.y += ty; /* We don't handle mask repeats (yet) */ - if (picture_needs_repeat(pMask, mask_offset.x, mask_offset.y, - clip_temp.x2, clip_temp.y2)) + if (!picture_has_pixels(pMask, mask_offset.x, mask_offset.y, + clip_temp.x2, clip_temp.y2)) goto fallback; mask_offset.x += pMask->pDrawable->x; <div class="HOEnZb"><div class="h5"> -- RMK's Patch system: <a href="http://www.armlinux.org.uk/developer/patches/" rel="noreferrer" target="_blank">http://www.armlinux.org.uk/developer/patches/</a> FTTC broadband for 0.8mile line in suburbia: sync at 8.8Mbps down 630kbps up According to <a href="http://speedtest.net" rel="noreferrer" target="_blank">speedtest.net</a>: 8.21Mbps down 510kbps up </div></div></blockquote></div> </div>