[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-commits
Subject:    Re: KDE/kdegraphics/kpdf
From:       Albert Astals Cid <astals11 () terra ! es>
Date:       2005-07-14 14:56:19
Message-ID: 200507141656.19902.astals11 () terra ! es
[Download RAW message or body]

Then lets see if you can explain this because i can not:
 - Then problem is inside SplashOutputDev.cc because you can compile 
everything with -O2 and it fails, but if you touch SplashOutputDev.cc and 
make it compile with the -fno-regmove it works.
 - Adding the following patch makes it work

Index: xpdf/xpdf/SplashOutputDev.cc
===================================================================
--- xpdf/xpdf/SplashOutputDev.cc        (revisià ³n: 434338)
+++ xpdf/xpdf/SplashOutputDev.cc        (copia de trabajo)
@@ -810,7 +810,10 @@
        if (subpath->getCurve(j)) {
          state->transform(subpath->getX(j), subpath->getY(j), &x1, &y1);
          state->transform(subpath->getX(j+1), subpath->getY(j+1), &x2, &y2);
+         printf("%f %f\n", x2, y2);
          state->transform(subpath->getX(j+2), subpath->getY(j+2), &x3, &y3);
+         printf("%f %f\n", x3, y3);
+         printf("Hi!\n");
          sPath->curveTo((SplashCoord)x1, (SplashCoord)y1,
                         (SplashCoord)x2, (SplashCoord)y2,
                         (SplashCoord)x3, (SplashCoord)y3);

and i can't see how much that can change things.

My compiler is gcc version 3.4.3 (Mandrakelinux 10.2 3.4.3-7mdk)

Obviously all help is welcome.

Albert

PS: I don't read kde-commits so please CC me.

A Dijous 14 Juliol 2005 16:24, và reu escriure:
> On Thursday 14 July 2005 13:32, Albert Astals Cid wrote:
> > +if CXX_SUPPORTS_NOREGMOVE
> > +  # This fixes crash in Bug 109015 which i assume is a compiler bug
> > +  # as adding some correctly placed printf in
> > SplashOutputDev::convertPath() makes this +  # option unneeded
> > +  KDE_CXXFLAGS=-fno-regmove
> > +endif
>
> please, don't. conditionalized variables have very poor support in
> automake/am_edit.
>
> Especially as this is a very ugly workaround for something that is likely
> to be a bug in the code anyway.

[prev in list] [next in list] [prev in thread] [next in thread]