kstars, when compiled with "-fstack-protector-all" gets aborted immediately after displaying the main window, saying something like: *** stack smashing detected ***: kstars - terminated kstars: stack smashing attack in function <unknown> - terminated Report to http://bugs.gentoo.org/ Killed drkonqi was as useless as usual, no core dump was generated, and also gdb was unable to intercept the problem, just because kstars got automagically terminated by signal 9. Following the chain of symbols from the gcc source tree via __stack_chk_fail into the glibc source tree and from there into the related ebuild, it turned out that in order to get core dumps with hardened Gentoo, you need to emerge sys-libs/glibc with USE="debug", well, because: # When using Hardened Gentoo stack handler, have smashes dump core for # analysis - debug only, as core could be an information leak # (paranoia). This was done like ever since sys-libs/glibc-2.5.1, much as if ulimit and /etc/security/limits.conf was not availabe then. Now, with proper signaling, I finally got a core: #0 0x4f935422 in __kernel_vsyscall () #1 0x4da8909b in __stack_chk_fail () at stack_chk_fail.c:296 #2 0x119ea860 in __stack_chk_fail_local () from /usr/bin/kstars #3 0x1196a40e in SaturnMoons::findPosition (this=0x134dc088, num=0x5a98a2a0, Saturn=0x11b80fb0, Sun=0x11b782d8) at /usr/src/debug/kde-base/kstars-4.3.4/kstars-4.3.4/ kstars/kstars/skyobjects/saturnmoons.cpp:476 #4 0x117fc86b in SaturnMoonsComponent::updateMoons (this=0x11b81dc0, num=0x5a98a2a0) at /usr/src/debug/kde-base/kstars-4.3.4/kstars-4.3.4/ kstars/kstars/skycomponents/saturnmoonscomponent.cpp:65 Inspecting the code it was then easy to see that the method SaturnMoons::findPosition() has numerous off-by-one indexing errors. Until now, these probably had manifested itself only in wrong moon positions and other spurious errors. Some excerpts from kstars/skyobjects/saturnmoons.cpp: 109 double X[9], Y[9], Z[9]; 110 double A1[9], B1[9], C1[9]; 111 double A2[9], B2[9], C2[9]; 112 double A3[9], B3[9], C3[9]; 113 double A4[9], B4[9], C4[9]; 114 double A5[9], B5[9], C5[9]; 115 double A6[9], B6[9], C6[9]; [...] 428 X[9] = 0; 429 Y[9] = 0; 430 Z[9] = 1; 431 432 for ( int i=1; i<10; ++i ) { 433 A1[i] = X[i]; 434 B1[i] = c1*Y[i] - s1*Z[i]; 435 C1[i] = s1*Y[i] + c1*Z[i]; 436 437 //Rotation towards the vernal equinox 438 A2[i] = c2*A1[i] - s2*B1[i]; 439 B2[i] = s2*A1[i] + c2*B1[i]; 440 C2[i] = C1[i]; [...]
Created attachment 213440 [details] output of emerge --info =kde-base/kstars-4.3.4 Regarding missing core dumps, it could be useful to check the hardened reports among #263681, #225563, #288419, #115285, #149292, #180451, #231225, #286587, #202582, #191005 and #219589
https://bugs.kde.org/show_bug.cgi?id=219293
Created attachment 213582 [details, diff] patch against kstars/kstars/skyobjects/saturnmoons.cpp To test it, just add the line PATCHES=( "${FILESDIR}/${PV}/saturnmoons-fix-index-and-angle-errors.patch" ) to kstars-4.3.4.ebuild, and put the patch into a newly created "files/4.3.4/" directory. This patch fixes four things: 1.) The author assumed C arrays would start with element 1. The consequence was a stack smashing "attack" detected when compiled with -fstack-protector-all 2.) summing up angles given in degrees with angles given in radians: see double u = 2*W5 - 2*theta + psi; 3.) remove unnecessary calls to MapTo0To360Range 4.) remove unused, superflous arrays declarations HOWEVER: Saturn's moon position still appear to be wrong when compared to the output of JPL's solar system simulator for any given date (here: http://space.jpl.nasa.gov/)
Thanks for your report - apparently it's been fixed upstream in r1065303 - http://websvn.kde.org/?view=revision&revision=1065303 That code was not only broken, but even "stolen" (and thus it's has been removed).
This should be included in 4.3.5 as it was fixed in 4.3 branch.
