Fix SEGVs seen in alldevs tests.

We see SEGVs in the alldevs tests with the escp device where
map_color_rgb is being called on an 8-bit memory device. The
implementation assumes that there will be a palette record
and so indexes into it.

It seems that palettes are actually unusual; the only time we
get one is if the device was created using the makewordimagedevice
operator from postscript.

Accordingly, we here update the map_color_rgb routine to behave
differently in the no-palette-present case. We return rgb values
as if the color index supplied was a greyscale value.

I believe this fixes the same issue seen with some other devices
including lj4dith and lj4dithp.

1 files changed, 13 insertions, 4 deletions

diff --git a/base/gdevmem.c b/base/gdevmem.c
index e0d247214..59e10bd47 100644
--- a/base/gdevmem.c
+++ b/base/gdevmem.c
@@ -858,11 +858,20 @@ mem_mapped_map_color_rgb(gx_device * dev, gx_color_index color,
                          gx_color_value prgb[3])
 {
     gx_device_memory * const mdev = (gx_device_memory *)dev;
-    const byte *pptr = mdev->palette.data + (int)color * 3;
+    const byte *pptr = mdev->palette.data;
 
-    prgb[0] = gx_color_value_from_byte(pptr[0]);
-    prgb[1] = gx_color_value_from_byte(pptr[1]);
-    prgb[2] = gx_color_value_from_byte(pptr[2]);
+    if (pptr == NULL) {
+        color = color * gx_max_color_value / ((1<<mdev->color_info.depth)-1);
+        prgb[0] = color;
+        prgb[1] = color;
+        prgb[2] = color;
+    } else {
+        pptr += (int)color * 3;
+
+        prgb[0] = gx_color_value_from_byte(pptr[0]);
+        prgb[1] = gx_color_value_from_byte(pptr[1]);
+        prgb[2] = gx_color_value_from_byte(pptr[2]);
+    }
     return 0;
 }