Re: Emacs 25.1.1 coredumps if company-mode pops up in a magit commit buffer window

[Eli Zaretskii <eliz@gnu.org>]

> From: Filipe Silva <filipe.silva@gmail.com>
> Date: Wed, 8 Mar 2017 18:06:40 -0300
> 
> Hi. Everytime open a magit buffer to commit write a commit message and company mode pops up with a
> completion menu, emacs 25.1.1 core dumps on me. 
> 
> This is the line I used: 
> 
> gdb --batch --quiet -ex "thread apply all bt full" -ex "quit" /usr/local/bin/emacs core.2466 > coredump.txt
> 
> My emacs version: GNU Emacs 25.1.1 (x86_64-unknown-linux-gnu, GTK+ Version 3.14.13) of 2017-02-22
> 
> Here's the coredump. 
> [...]
> Thread 1 (Thread 0x7f4036acda00 (LWP 2466)):
> #0 0x00007f403154b23b in raise () from /lib64/libpthread.so.0
> No symbol table info available.
> #1 0x00000000004e6361 in terminate_due_to_signal (sig=sig@entry=6,
> backtrace_limit=backtrace_limit@entry=40) at emacs.c:381
> No locals.
> #2 0x00000000004fe503 in emacs_abort () at sysdep.c:2247
> No locals.
> #3 0x00000000004a55f5 in bidi_paragraph_init (dir=<optimized out>, bidi_it=bidi_it@entry=0x7ffd828deab0,
> no_default_p=no_default_p@entry=false) at bidi.c:1674
> bytepos = 266
> string_p = <optimized out>
> pstartbyte = <optimized out>
> begbyte = <optimized out>
> end = 266
> #4 0x000000000043fe4a in set_iterator_to_next (it=it@entry=0x7ffd828de110,
> reseat_p=reseat_p@entry=false) at xdisp.c:7426
> prev_scan_dir = 1
> #5 0x000000000043f4ce in forward_to_next_line_start (it=it@entry=0x7ffd828de110,
> skipped_p=skipped_p@entry=0x7ffd828dc42f, bidi_it_prev=bidi_it_prev@entry=0x7ffd828dc430) at
> xdisp.c:6224
> old_selective = <optimized out>
> newline_found_p = false
> n = <optimized out>
> #6 0x000000000043f6aa in reseat_at_next_visible_line_start (it=it@entry=0x7ffd828de110,
> on_newline_p=on_newline_p@entry=false) at xdisp.c:6431

It's hard to understand how this happens: your build is an optimized
one, so many important values are not accessible to the debugger.  The
stuff that is shown seems to present a contradicting picture:
character position 266 is EOB, but it also has a newline character
there, which AFAIU cannot happen.

Can you prepare a recipe for reproducing this, from "emacs -Q",
loading all the necessary Lisp packages, and then showing the commands
you need to invoke to get Emacs to abort like this?

Thanks.