4.7.0-pre4 editor crashes in color terminals

[yury_t]

Internal editor crashes MC when running on color terminals. Monochrome terminals (mc -b) work fine. Patch from ticket #1857 was no help.
System is basicly Slackware 12.2. Following are the stack traces of MC 4.7.0-pre4 modified with pacth #1857.
This is stack trace of starting new file (Shift+F4), then saving:
#0 0xb7c39c66 in raise () from /lib/libc.so.6
#1 0xb7c3b571 in abort () from /lib/libc.so.6
#2 0xb7c7260b in libc_message () from /lib/libc.so.6
#3 0xb7c7a564 in _int_free () from /lib/libc.so.6
#4 0xb7c7e010 in free () from /lib/libc.so.6
#5 0xb7d9a046 in g_free () from /usr/lib/libglib-2.0.so.0
#6 0x080a22d5 in edit_read_syntax_file (edit=0x81c4e58, pnames=0x0,

syntax_file=<value optimized out>, editor_file=0x81c2d48 "222",
first_line=0x812d6c0 "222", type=0x0) at syntax.c:748

#7 0x080a3266 in edit_load_syntax (edit=0x81c4e58, pnames=0x0, type=0x0) at syntax.c:1198
#8 0x0809c2be in edit_save_as_cmd (edit=0x81c4e58) at editcmd.c:618
#9 0x0809c3e9 in edit_save_cmd (edit=0x81c4e58) at editcmd.c:866
#10 0x0809623c in edit_execute_cmd (edit=0x81c4e58, command=101, char_for_insertion=-1)

at edit.c:3028

#11 0x08096b70 in edit_execute_key_command (edit=0x81c4e58, command=101, char_for_insertion=-1)

at edit.c:2551

#12 0x0809d6ad in edit_callback (w=0x81c4e58, msg=WIDGET_KEY, parm=1002) at editwidget.c:377
#13 0x0809cf64 in cmd_F2 (edit=0x81c4e58) at ../src/tty/../../src/dialog.h:237
#14 0x08059d1a in buttonbar_call (bb=<value optimized out>, i=<value optimized out>)

at widget.c:2687

#15 0x0805a4b9 in buttonbar_callback (w=0x81bd388, msg=WIDGET_HOTKEY, parm=1002)

at widget.c:2715

#16 0x08050ecb in dlg_process_event (h=0x81a9e00, key=1002, event=0xbff50144)

at ../src/tty/../../src/dialog.h:237

#17 0x08051171 in run_dlg (h=0x81a9e00) at dialog.c:858
#18 0x0809d0d6 in edit_file (_file=0x0, line=0) at editwidget.c:239
#19 0x08072aba in do_edit_at_line (what=0x0, start_line=0) at cmd.c:305
#20 0x08072afe in do_edit (what=0x0) at cmd.c:323
#21 0x080576a4 in panel_key (panel=0x81b3bc8, key=1014) at screen.c:2636
#22 0x08050cb5 in dlg_process_event (h=0x81abf08, key=1014, event=0xbff512c4)

at ../src/tty/../../src/dialog.h:237

#23 0x08051171 in run_dlg (h=0x81abf08) at dialog.c:858
#24 0x0808a071 in do_nc () at main.c:1805
#25 0x0808a789 in main (argc=Cannot access memory at address 0x1
) at main.c:2234
This is stack trace of opening simple (non-syntax-colored) file, then immediately closing it:
#0 0xb7c06c66 in raise () from /lib/libc.so.6
#1 0xb7c08571 in abort () from /lib/libc.so.6
#2 0xb7c3f60b in libc_message () from /lib/libc.so.6
#3 0xb7c47564 in _int_free () from /lib/libc.so.6
#4 0xb7c4b010 in free () from /lib/libc.so.6
#5 0xb7d67046 in g_free () from /usr/lib/libglib-2.0.so.0
#6 0x080a1dcb in edit_free_syntax_rules (edit=0xb7d23ff4) at syntax.c:1005
#7 0x08092880 in edit_clean (edit=0x81c8328) at edit.c:870
#8 0x0809d6cb in edit_callback (w=0x81c8328, msg=WIDGET_INIT, parm=0) at editwidget.c:394
#9 0x0805051d in dlg_broadcast_msg_to (h=<value optimized out>, message=WIDGET_DESTROY,

reverse=0, flags=0) at ../src/tty/../../src/dialog.h:237

#10 0x08050bd6 in destroy_dlg (h=0x81afd50) at dialog.c:884
#11 0x0809d0de in edit_file (_file=0x81a5d30 "222", line=0) at editwidget.c:241
#12 0x08072aba in do_edit_at_line (what=0x81a5d30 "222", start_line=0) at cmd.c:305
#13 0x08072afe in do_edit (what=0x0) at cmd.c:323
#14 0x08059d03 in buttonbar_call (bb=<value optimized out>, i=<value optimized out>)

at widget.c:2684

#15 0x0805a4b9 in buttonbar_callback (w=0x81ad070, msg=WIDGET_HOTKEY, parm=1004)

at widget.c:2715

#16 0x08050ecb in dlg_process_event (h=0x81abf08, key=1004, event=0xbfd20094)

at ../src/tty/../../src/dialog.h:237

#17 0x08051171 in run_dlg (h=0x81abf08) at dialog.c:858
#18 0x0808a071 in do_nc () at main.c:1805
#19 0x0808a789 in main (argc=Cannot access memory at address 0x1
) at main.c:2234
This is output.........
Program received signal SIGSEGV, Segmentation fault.
compare_word_to_right (edit=0x81c8328, i=287,

text=0x69633d01 <Address 0x69633d01 out of bounds>,
whole_left=0x6f6f6f6f <Address 0x6f6f6f6f out of bounds>,
whole_right=0x6f6f6f6f <Address 0x6f6f6f6f out of bounds>, line_start=577728367)
at syntax.c:185

185 if (!*text)
...........and stack trace of opening the file which requires syntax-coloring:
#0 compare_word_to_right (edit=0x81c8328, i=287,

text=0x69633d01 <Address 0x69633d01 out of bounds>,
whole_left=0x6f6f6f6f <Address 0x6f6f6f6f out of bounds>,
whole_right=0x6f6f6f6f <Address 0x6f6f6f6f out of bounds>, line_start=577728367)
at syntax.c:185

#1 0x080a167b in apply_rules_going_right (edit=0x81c8328, i=287, rule=

{keyword = 0, end = 29 '\035', context = 0 '\0', _context = 30 '\036', border = 0 '\0'})

at syntax.c:364

#2 0x080a1b1d in edit_get_rule (edit=0x81c8328, byte_index=287) at syntax.c:457
#3 0x080a33df in edit_get_syntax_color (edit=0x81c8328, byte_index=287, color=0xbfdbf71c)

at syntax.c:505

#4 0x0809df5b in edit_draw_this_line (edit=0x81c8328, b=251, row=8, start_col=0, end_col=95)

at editdraw.c:454

#5 0x0809e614 in edit_render_keypress (edit=0x81c8328) at editdraw.c:640
#6 0x0809d4fd in edit_callback (w=0x81c8328, msg=WIDGET_FOCUS, parm=0) at editwidget.c:368
#7 0x0805057a in dlg_focus (h=0x81a6428) at ../src/tty/../../src/dialog.h:237
#8 0x08050946 in init_dlg (h=0x81a6428) at dialog.c:795
#9 0x080510d1 in run_dlg (h=0x81a6428) at dialog.c:872
#10 0x0809d0d6 in edit_file (_file=0x81ad6c8 "opera6.html", line=0) at editwidget.c:239
#11 0x08072aba in do_edit_at_line (what=0x81ad6c8 "opera6.html", start_line=136086312)

at cmd.c:305

#12 0x08072afe in do_edit (what=0x81c8328 "") at cmd.c:323
#13 0x08059d03 in buttonbar_call (bb=<value optimized out>, i=<value optimized out>)

at widget.c:2684

#14 0x0805a4b9 in buttonbar_callback (w=0x81ad070, msg=WIDGET_HOTKEY, parm=1004)

at widget.c:2715

#15 0x08050ecb in dlg_process_event (h=0x81abf08, key=1004, event=0xbfdbf934)

at ../src/tty/../../src/dialog.h:237

#16 0x08051171 in run_dlg (h=0x81abf08) at dialog.c:858
#17 0x0808a071 in do_nc () at main.c:1805
#18 0x0808a789 in main (argc=Cannot access memory at address 0x1
) at main.c:2234

[andrew_b]

Which color terminals do you mean?

Please build MC with full debug info (see ​Report crashes) and attach results in this ticket.

[yury_t]

Configure options: ./configure --prefix=/usr \
--with-screen=slang \

--enable-charset \
--with-samba \
--without-x \
--with-gpm-mouse \

Build and install per crash reporting instructions.

Color terminals were: 'linux' (fullscreen console) and 'xterm'.

Attached are four files with gdb 'bt'/'bt full' output. Two files (bt-edit-*) for attempts to open a file 'longer than just few bytes' in 'linux' and 'xterm'. One file (bt-closeedit-xterm.txt) for attempt to close editor on the text 'just few bytes long' in 'xterm'. One file (bt-closeedit-fs.txt) was meant to make similar test in fullscreen, but editor actually when actually starting.

[angel_il]

Replying to yury_t:

can you try current "master"?

[andrew_b]

Cannot reproduce.
Please describe step by step how to reproduce this bug. And describe your hardware.

[yury_t]

Andrew:
What step by step? I hit F4 and MC (4.7.0-pre4) coredumps. Sometimes it lets me edit a little and coredumps on closing the editor.

I thought for a sec my inis were to blame, with legacy entries in them (keys for freebsd terms etc.). Nope.

However, if I try to work in monochrome terminal 'xterm-mono', MC lets me use edit OK.

This is first time in years something on this scale happens, I tell you. And hardware never was an issue with MC.

Weren't gdb traces any help?

However, I'm attaching my dmesg output while I'm trying to try current 'master'. As you may see, it's quite a trivial hardware.

[yury_t]

Mc 'master' has the same issue.

[angel_il]

sorry, but try "1845_code_cleanup"
in this branch removed some uninitialized string values..

[yury_t]

I proceeded according to the site's instructions, but I'm not able to see any git changes in that branch, compared to head. I don't know git enough, too.

Seems I'll just have to revert to the last-good version.

[angel_il]

git pull
git checkout -b 1845_code_cleanup origin/1845_code_cleanup
./autogen.sh
make
src/mc

[angel_il]

Can you say step by step how to reproduce this bug.
for example:
./configure --prefix=/usr
make
make install
run mc
shift-f4
<segfault>

[yury_t]

I'm stuck with git process. As I understand it, git didn't report any activity on this 'branch switching'. Is my 'copy' of repository actually 'master' or '1845_code_cleanup'? What do I do next? I proceeded as follows:

$ git clone ​git://midnight-commander.org/git/mc.git
Initialized empty Git repository in /.site/home/ty/src/mc/mc/.git/
remote: Counting objects: 59627, done.
remote: Compressing objects: 100% (12052/12052), done.
remote: Total 59627 (delta 47697), reused 59337 (delta 47455)
Receiving objects: 100% (59627/59627), 27.45 MiB | 28 KiB/s, done.
Resolving deltas: 100% (47697/47697), done.

$ cd mc
$ git pull
Already up-to-date.

$ git checkout -b 1845_code_cleanup origin/1845_code_cleanup
Branch 1845_code_cleanup set up to track remote branch refs/remotes/origin/1845_code_cleanup.
Switched to a new branch "1845_code_cleanup"
$ git pull
Already up-to-date.
$ git branch

• 1845_code_cleanup master

[andrew_b]

./configure --<you options here>
make
make install

[yury_t]

None of this mockery, please. :) I know how to configuremakemakeinstall. I'm just not sure that I have exactly 1845_code_cleanup that is going to be configured, made etc.

[andrew_b]

• 1845_code_cleanup

Asterisk marks the current brunch.