On 2013-06-05 11:23 +0200, Thomas Dickey wrote:
> On Wed, Jun 05, 2013 at 12:57:52AM +0200, Adam Borowski wrote:
>> Package: libtinfo5
>> Version: 5.9+20130504-1
>> Severity: normal
>>
>>
>> If you run a program that uses tgetent() -- at least bash and gdb, with
>
> They both use readline, which has its own problems :-)
>
>> TERM=KBtin, it will segfault on startup. Its terminfo definition can be
>> found in package "kbtin", or by tic-ing:
>
> A traceback would be helpful.
Here you go:
,----
| (gdb) bt full
| #0 0xf7fa7683 in tgetent (
| bufp=0x815c008 '\337' <repeats 199 times>, <incomplete sequence \337>...,
| name=0x8143e28 "kbtin")
| at
/build/buildd-ncurses_5.9+20130504-1-i386-jLNieE/ncurses-5.9+20130504/ncurses/tinfo/lib_termcap.c:166
| rc = 1
| n = <optimized out>
| #1 0x080f912e in _rl_init_terminal_io (terminal_name=0x0)
| at ../../../bash/lib/readline/terminal.c:452
| term = 0x8143e28 "kbtin"
| buffer = 0x8150008 '\337' <repeats 199 times>, <incomplete sequence
\337>...
| tty = 0
| tgetent_ret = 0
| #2 0x080f8f08 in _rl_set_screen_size (rows=40, cols=100)
| at ../../../bash/lib/readline/terminal.c:312
| No locals.
| #3 0x080f8f73 in rl_set_screen_size (rows=40, cols=100)
| at ../../../bash/lib/readline/terminal.c:331
| No locals.
| #4 0x080e0179 in get_new_window_size (from_sig=0, rp=0x0, cp=0x0)
| at ../../../bash/lib/sh/winsize.c:88
| win = {ws_row = 40, ws_col = 100, ws_xpixel = 915, ws_ypixel = 604}
| tty = 255
| #5 0x08088be8 in get_tty_state () at ../bash/jobs.c:2030
| tty = 255
| #6 0x0805f395 in main (argc=1, argv=0xffffd494, env=0xffffd49c) at
../bash/shell.c:739
| i = -134668300
| code = 0
| old_errexit_flag = 0
| saverst = 0
| locally_skip_execution = 0
| arg_index = 1
| top_level_arg_index = 1
| (gdb)
`----
On 2013-06-05 00:57 +0200, Adam Borowski wrote:
> I haven't excluded the possibility something else it to blame, but debugging
> through tinfo's code is a pain: it's layers upon layers of macros. At
> least, at the time of segfault, the values in cur_term are bogus.
Same here:
,----
| (gdb) print cur_term->type
| $8 = {term_names = 0xcfcfcfcf <Address 0xcfcfcfcf out of bounds>,
| str_table = 0xcfcfcfcf <Address 0xcfcfcfcf out of bounds>,
| Booleans = 0xcfcfcfcf <Address 0xcfcfcfcf out of bounds>, Numbers =
0xcfcfcfcf,
| Strings = 0xcfcfcfcf, ext_str_table = 0xcfcfcfcf <Address 0xcfcfcfcf out of
bounds>,
| ext_Names = 0xcfcfcfcf, num_Booleans = 53199, num_Numbers = 53199,
num_Strings = 53199,
| ext_Booleans = 53199, ext_Numbers = 53199, ext_Strings = 53199}
| (gdb)
`----
Looking at the offending line 166 in ncurses/tinfo/lib_termcap.c, it
tries to read cursor_left which is #defined to cur_term->type.Strings[14]:
,----
| (gdb) print cur_term->type.Strings[14]
| Cannot access memory at address 0xcfcfd007
| (gdb)
`----
No wonder that it segfaults, but why cur_term has these bogus values…
you're the expert. :-)
Cheers,
Sven
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
