134070 2006-09-14 16:20:05 +0000 crash on large result, "exp" operation 2006-09-23 01:21:34 +0000 1 1 2 Applications kcalc general unspecified Gentoo Packages Linux RESOLVED FIXED NOR crash --- 1 ebachmann kniederk 0 oldest_to_newest 469888 0 ebachmann 2006-09-14 16:20:05 +0000 Version: 2.0.4 (using KDE KDE 3.5.4) Installed from: Gentoo Packages Compiler: gcc-4.1.1 OS: Linux enter a large number (say 1000), then hit "Inv", then "e^x" => Kcalc crashes. according to GDB the crash occurs in: #0 0x000000354b90cbd9 in raise () from /lib/libpthread.so.0 #1 0x000000370ea07d0e in __gmp_invalid_operation () from /usr/lib64/libgmp.so.3 #2 0x000000370ea09aed in __gmpf_set_d () from /usr/lib64/libgmp.so.3 #3 0x0000003478158d8c in KNumber::KNumber () from /usr/kde/3.5/lib64/libkdeinit_kcalc.so #4 0x000000347813c2be in CalcEngine::Exp () from /usr/kde/3.5/lib64/libkdeinit_kcalc.so #5 0x000000347814abdb in KCalculator::slotLnclicked () from /usr/kde/3.5/lib64/libkdeinit_kcalc.so Seems like its not generally a problem with the result value, as kcalc can calculate e.g. 10^1000000 just fine. 470097 1 phil 2006-09-15 13:26:47 +0000 Confirmed. Different backtrace though: #12 0x29460b6c in sigaction () from /usr/lib/libpthread.so.2 #13 0x2952db23 in abort () from /lib/libc.so.6 #14 0x2984dc2f in __gmp_invalid_operation () from /usr/local/lib/libgmp.so.7 #15 0x2984f513 in __gmpf_set_d () from /usr/local/lib/libgmp.so.7 #16 0x2983b28a in _knumfloat (this=0x8245fb8, num=inf) at knumber_priv.h:270 #17 0x298382ea in KNumber (this=0xbfbfdb80, num=inf) at /home/phil/kdesrc/kdeutils/kcalc/knumber/knumber.cpp:72 #18 0x29824630 in CalcEngine::Exp (this=0x8168c88, input= {static Zero = {static Zero = <same as static member of an already seen type>, static One = {static Zero = <same as static member of an already seen type>, static One = <same as static member of an already seen type>, static MinusOne = {static Zero = <same as static member of an already seen type>, static One = <same as static member of an already seen type>, static MinusOne = <same as static member of an already seen type>, static Pi = {static Zero = <same as static member of an already seen type>, static One = <same as static member of an already seen type>, static MinusOne = <same as static member of an already seen type>, static Pi = <same as static member of an already seen type>, static Euler = {static Zero = <same as static member of an already seen type>, static One = <same as static member of an already seen type>, static MinusOne = <same as static member of an already seen type>, static Pi = <same as static member of an already seen type>, static Euler = <same as static member of an already seen type>, _num = 0x80e3a38, static _float_output = true, static _fraction_input = true, static _splitoffinteger_output = false}, _num = 0x80d0938, static _float_output = true, static _fraction_input = true, static _splitoffinteger_output = false}, static Euler = <same as static member of an already seen type>, _num = 0x80d4fe0, static _float_output = true, static _fraction_input = true, static _splitoffinteger_output = false}, static Pi = <same as static member of an already seen type>, static Euler = <same as static member of an already seen type>, _num = 0x80d0e18, static _float_output = true, static _fraction_input = true, static _splitoffinteger_output = false}, static MinusOne = <same as static member of an already seen type>, static Pi = <same as static member of an already seen type>, static Euler = <same as static member of an already seen type>, _num = 0x80d0b68, static _float_output = true, static _fraction_input = true, static _splitoffinteger_output = false}, static One = <same as static member of an already seen type>, static MinusOne = <same as static member of an already seen type>, static Pi = <same as static member of an already seen type>, static Euler = <same as static member of an already seen type>, _num = 0x82385b0, static _float_output = true, static _fraction_input = true, static _splitoffinteger_output = false}) at /home/phil/kdesrc/kdeutils/kcalc/kcalc_core.cpp:574 #19 0x298168b3 in KCalculator::slotLnclicked (this=0x8168a18) at /home/phil/kdesrc/kdeutils/kcalc/kcalc.cpp:1517 #20 0x2981a6a9 in KCalculator::qt_invoke (this=0x8168a18, _id=99, _o=0xbfbfdcc0) at kcalc.moc:333 #21 0x28e141e0 in QObject::activate_signal () from /usr/X11R6/lib/libqt-mt.so.3 470099 2 ebachmann 2006-09-15 13:44:00 +0000 Well, besides the differences caused by 64bit/32bit arch, different library versions, and my version coming from a strip'ed binary, they look quite similar to me. Basically it looks like libgmp is trying to calculate the result, gets an overflow, and raises a SIGFPE signal, which isn't handled. 470398 3 phil 2006-09-16 20:42:44 +0000 Ah yes, good point. Should've looked more carefully 470676 4 17816 phil 2006-09-17 23:15:29 +0000 Created attachment 17816 Suggested fix Change the KNumber initialization to check for NaN and Inf, since passing these to libgmp causes a crash. Does this also need to be done for other types? 471059 5 kniederk 2006-09-19 11:48:34 +0000 Good idea (the patch). I try to apply this as soon as I have time. Klaus 472087 6 kniederk 2006-09-23 01:21:34 +0000 Applied your patch. Thanks a lot Klaus 17816 2006-09-17 23:15:29 +0000 2006-09-17 23:15:29 +0000 Suggested fix kcalc.diff text/plain 493 phil SW5kZXg6IGtudW1iZXIva251bWJlci5jcHAKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0ga251bWJlci9rbnVtYmVy LmNwcAkocmV2aXNpb24gNTgyMTgzKQorKysga251bWJlci9rbnVtYmVyLmNwcAkod29ya2luZyBj b3B5KQpAQCAtNjksNyArNjksMTAgQEAKIAogS051bWJlcjo6S051bWJlcihkb3VibGUgbnVtKQog ewotICBfbnVtID0gbmV3IF9rbnVtZmxvYXQobnVtKTsKKyAgaWYgKCBpc2luZihudW0pICkgX251 bSA9IG5ldyBfa251bWVycm9yKCBfa251bWJlcjo6SW5maW5pdHkgKTsKKyAgZWxzZSBpZiAoIGlz bmFuKG51bSkgKSBfbnVtID0gbmV3IF9rbnVtZXJyb3IoIF9rbnVtYmVyOjpVbmRlZmluZWROdW1i ZXIgKTsKKyAgZWxzZSBfbnVtID0gbmV3IF9rbnVtZmxvYXQobnVtKTsgICAgICAKKwogfQogCiBL TnVtYmVyOjpLTnVtYmVyKEtOdW1iZXIgY29uc3QgJiBudW0pCg==