Created attachment 26344 [details] [review]
suggested patch to poppler-0.10.7/poppler/SplashOutputDev.cc

My sample file is 3 MB and too large to post.  Contact me at williambader@hotmail.com for a copy by email.

the color[4] -> color[3] thing is already fixed in master.

For the other issue i think i like better this

diff --git a/poppler/PSOutputDev.cc b/poppler/PSOutputDev.cc
index d24e77e..323d08e 100644
--- a/poppler/PSOutputDev.cc
+++ b/poppler/PSOutputDev.cc
@@ -2951,6 +2951,10 @@ GBool PSOutputDev::checkPageSlice(Page *page, double /*hDPI*/, double /*vDPI*/,
     paperColor[0] = paperColor[1] = paperColor[2] = paperColor[3] = 0;
     splashOut = new SplashOutputDev(splashModeCMYK8, 1, gFalse,
                                    paperColor, gTrue, gFalse);
+#else
+  } else if (level == psLevel1Sep) {
+    error(-1, "pdftops was built without CMYK support, -level1sep needs it to work in this file");
+    return gFalse;
 #endif
   } else {
     paperColor[0] = paperColor[1] = paperColor[2] = 0xff;

What do you think?

Created attachment 26405 [details]
example of bad output from -level1sep

Created attachment 26406 [details]
rendering with -level1sep  fixed but without the luminosity color[4] fix

Yes, your patch is better than mine because it reports an error only when PSOutputDev::checkPageSlice() is called and poppler has been compiled without SPLASH_CMYK.  pdftops will process many files correctly, even without SPLASH_CMYK, as long as it does not call checkPageSlice().
I have attached a rendering with and without SPLASH_CMYK to show what happens in case anyone else runs into this.

Patch commited, will be in next poppler version