Created attachment 372540 [details, diff] Makes the segfault with ipython go away. I ran into this bug today (and chutzpah confirmed he is experiencing it as well), where traversing through history causes a segfault (pressing up arrow twice in my case). The GNU/readline mailing list has a similar report, with a patch attached. The mailing list thread is here: http://lists.gnu.org/archive/html/bug-readline/2014-03/msg00010.html The patch that came from a readline dev will be attached to this bug. I can confirm that the patch works for me
should be all set now in the tree; thanks for the report! Commit message: Add fix from upstream for some crashes http://sources.gentoo.org/sys-libs/readline/files/readline-6.3-dispatch-multikey.patch?rev=1.1 http://sources.gentoo.org/sys-libs/readline/readline-6.3-r2.ebuild?rev=1.1
Not sure if this should be reported upstream, but I'm still having this problem with sys-libs/readline-6.3-r2 and dev-python/ipython-1.2.1 (I've rebuilt both multiple times) Hell, I did an 'emerge -vaDeN ipython' with no change in this bug. Is there something I'm missing?
attach your full build log for readline, emerge info, and any customizations you've made to readline (like ~/.inputrc)
Also hit by this one, readline-6.3-r2 still bugs me. I have compiled, in order: sys-libs/readline-6.3-r2 (USE="-static-libs" ABI_MIPS="-n32 -n64 -o32" ABI_X86="64 -32 -x32") dev-lang/python-3.3.4 (USE="gdbm ipv6 ncurses readline sqlite ssl threads xml -build -doc -examples -hardened -tk -wininst") dev-python/ipython-1.2.1 (USE="matplotlib notebook smp -doc -emacs -examples -mongodb -nbconvert -octave -qt4 {-test} -wxwidgets" PYTHON_TARGETS="python2_7 python3_3 -python2_6 -python3_2") I'm not aware of any modifications I could have made to readline, certainly I have no ~/.inputrc. I have a pretty straightforward way to reproduce this bug: 1. launch ipython 2. Arrow up 3. Backspace 4. Arrow down two times 5. Segfault The same does NOT happen to the underlying python. (I have inserted a dummy command "1.5. x = 0<CR>" before the "arrow-backspace-arrow" so that backspace effectively does something, with this extra one, ipython still segfaults) Replacing "backspace" by "arrow-left" or "arrow-right" also produces segfault in ipython.
Created attachment 373536 [details] emerge --info
if you can reproduce, please post the steps. if you can't, then run it through gdb so you can get backtraces when it does crash. gdb --args python `which ipython`
I still encounter segfaults from ipython and others. I found a different patch upstream: http://lists.gnu.org/archive/html/bug-readline/2014-03/msg00034.html has one where that one condition reads r >= 0 instead of r == 0. Which is relevant, because in the case of search-history-backward failing, the return value is 1. With PgUp bound to history-search-backward in /etc/inputrc, I could run "rlfe cat", type some stuff and then press PgUp twice to reproduce the segfault. With the mentioned patch instead of the one in portage, things work.
Created attachment 373754 [details, diff] Better fix This is the patch from http://lists.gnu.org/archive/html/bug-readline/2014-03/msg00034.html but converted to unified diff and with a URL in the comment up front. It handles the r == 1 case correctly.
Commit message: Update multikey patch from upstream http://sources.gentoo.org/sys-libs/readline/files/readline-6.3-dispatch-multikey.patch?r1=1.1&r2=1.2 http://sources.gentoo.org/sys-libs/readline/readline-6.3-r4.ebuild?rev=1.1
Sorry for not responding, didn't realize I wasn't on the CC list. Just wanted to say that the fix in -r4 worked for me.