forwarded 345739 [email protected]
thanks
Hi Karl!
Here is another bug report, this time about segfaults. It contains
extensive backtraces and a small patch.
I tried hard to reproduce this, but didn't succeed. Maybe I cannot
resize as fast as the submitter.
If you already know about this problem and/or have some fix in the cvs
repository, please let me know.
Best wishes
Norbert
-------------------------------------------------------------------------------
Dr. Norbert Preining <preining AT logic DOT at> Università di Siena
gpg DSA: 0x09C5B094 fp: 14DF 2E6C 0307 BE6D AD76 A9C0 D2BF 4AA3 09C5 B094
-------------------------------------------------------------------------------
EMSWORTH (n.)
Measure of time and noiselessness defined as the moment between the
doors of a lift closing and it beginning to move.
--- Douglas Adams, The Meaning of Liff
--- Begin Message ---
Package: info
Version: 4.8-2
Severity: normal
Tags: patch
Hi,
When resizing the terminal rapidly, info will segfault:
(gdb) bt
#0 0x0804a503 in display_update_one_window (win=0x80a3fd8) at display.c:301
#1 0x0804a02b in display_update_display (window=0x80a3fd8) at display.c:85
#2 0x08061578 in redisplay_after_signal () at signals.c:161
#3 0x08061609 in reset_info_window_sizes () at signals.c:176
#4 0x0806180f in info_signal_proc (sig=28) at signals.c:277
#5 <signal handler called>
#6 0xffffe40e in __kernel_vsyscall ()
#7 0xb7e4d193 in read () from /lib/tls/i686/cmov/libc.so.6
#8 0x080612a6 in info_get_input_char () at session.c:5002
#9 0x080595fd in info_read_and_dispatch () at session.c:217
#10 0x08059534 in info_session () at session.c:175
#11 0x0805950c in display_startup_message_and_start () at session.c:166
#12 0x080594c9 in begin_info_session (initial_node=0x8095e10) at session.c:153
#13 0x08051b58 in main (argc=2, argv=0xbf92c224) at info.c:507
(gdb) print entry
$1 = (DISPLAY_LINE *) 0x0
I easily reproduce this with "info autoconf" or "info automake" in a
xterm and rapidly resizing up and down, especially to a small size.
My first guess what at the second part of the if() condition, hence I
tried moving the entry != NULL test higher, but that wasn't enough
since entry is derefenced a couple of times afterwards.
Hence, I simply protected the following chunk, and the segfaults appear
less often now. The segfaults happening with entry == NULL are gone
for me, and I can resize info to a small size, however I still got an
occasional segfault when rapidly resizing, with a weird value of entry:
(gdb) bt
#0 0x0804a4f4 in display_update_one_window (win=0x80a3fd8) at display.c:303
#1 0x0804a02b in display_update_display (window=0x80a3fd8) at display.c:85
#2 0x0806157c in redisplay_after_signal () at signals.c:161
#3 0x0806160d in reset_info_window_sizes () at signals.c:176
#4 0x08061813 in info_signal_proc (sig=28) at signals.c:277
#5 <signal handler called>
#6 0xffffe40e in __kernel_vsyscall ()
#7 0xb7e7e193 in read () from /lib/tls/i686/cmov/libc.so.6
#8 0x080612aa in info_get_input_char () at session.c:5002
#9 0x08059601 in info_read_and_dispatch () at session.c:217
#10 0x08059538 in info_session () at session.c:175
#11 0x08059510 in display_startup_message_and_start () at session.c:166
#12 0x080594cd in begin_info_session (initial_node=0x8095e10) at session.c:153
#13 0x08051b5c in main (argc=2, argv=0xbfa5eba4) at info.c:507
(gdb) print entry
$1 = (DISPLAY_LINE *) 0x59
I'm afraid some corruption happens or unintialized memory is used, and
I suggest you run info in valgrind.
If you find them useful, please include the attached changes. (The
patch is large, but only a couple of lines were truly changed.)
-- System Information:
Debian Release: testing/unstable
APT prefers oldstable
APT policy: (500, 'oldstable'), (500, 'unstable'), (500, 'testing'), (500,
'stable'), (1, 'experimental')
Architecture: i386 (i686)
Shell: /bin/sh linked to /bin/bash
Kernel: Linux 2.6.14-2-686
Locale: [EMAIL PROTECTED], [EMAIL PROTECTED] (charmap=ISO-8859-15)
Versions of packages info depends on:
ii libc6 2.3.5-9 GNU C Library: Shared libraries an
ii libncurses5 5.5-1 Shared libraries for terminal hand
info recommends no packages.
-- no debconf information
--
Loïc Minier <[EMAIL PROTECTED]>
Current Earth status: NOT DESTROYED
--- texinfo-4.8/debian/changelog
+++ texinfo-4.8/debian/changelog
@@ -1,3 +1,12 @@
+texinfo (4.8-2.1) unstable; urgency=low
+
+ * Non-maintainer upload.
+ * Fix segfault when resizing the terminal rapidly by checking more carefully
+ whether the display line is set prior to refering to it.
+ [info/display.c:display_update_one_window]
+
+ -- Loic Minier <[EMAIL PROTECTED]> Mon, 2 Jan 2006 21:31:48 +0100
+
texinfo (4.8-2) unstable; urgency=low
* Edit the changelog entry for 4.8-1 to include fixes for bugs that
--- texinfo-4.8.orig/info/display.c
+++ texinfo-4.8/info/display.c
@@ -294,62 +294,65 @@
on the screen. */
entry = display[line_index + win->first_row];
- /* If the screen line is inversed, then we have to clear
- the line from the screen first. Why, I don't know.
- (But don't do this if we have no visible entries, as can
- happen if the window is shrunk very small.) */
- if ((entry && entry->inverse)
- /* Need to erase the line if it has escape sequences. */
- || (raw_escapes_p && strchr (entry->text, '\033') != 0))
- {
- terminal_goto_xy (0, line_index + win->first_row);
- terminal_clear_to_eol ();
- entry->inverse = 0;
- entry->text[0] = '\0';
- entry->textlen = 0;
- }
+ /* If the window is very small, entry might be NULL. */
+ if (entry) {
+ /* If the screen line is inversed, then we have to clear
+ the line from the screen first. Why, I don't know.
+ (But don't do this if we have no visible entries, as can
+ happen if the window is shrunk very small.) */
+ if (entry->inverse
+ /* Need to erase the line if it has escape sequences. */
+ || (raw_escapes_p && strchr (entry->text, '\033') != 0))
+ {
+ terminal_goto_xy (0, line_index + win->first_row);
+ terminal_clear_to_eol ();
+ entry->inverse = 0;
+ entry->text[0] = '\0';
+ entry->textlen = 0;
+ }
- /* Find the offset where these lines differ. */
- for (i = 0; i < pl_index; i++)
- if (printed_line[i] != entry->text[i])
- break;
-
- /* If the lines are not the same length, or if they differed
- at all, we must do some redrawing. */
- if ((i != pl_index) || (pl_index != entry->textlen))
- {
- /* Move to the proper point on the terminal. */
- terminal_goto_xy (i, line_index + win->first_row);
+ /* Find the offset where these lines differ. */
+ for (i = 0; i < pl_index; i++)
+ if (printed_line[i] != entry->text[i])
+ break;
+
+ /* If the lines are not the same length, or if they differed
+ at all, we must do some redrawing. */
+ if ((i != pl_index) || (pl_index != entry->textlen))
+ {
+ /* Move to the proper point on the terminal. */
+ terminal_goto_xy (i, line_index + win->first_row);
- /* If there is any text to print, print it. */
- if (i != pl_index)
- terminal_put_text (printed_line + i);
-
- /* If the printed text didn't extend all the way to the edge
- of the window, and text was appearing between here and the
- edge of the window, clear from here to the end of the line. */
- if ((pl_index < win->width + pl_ignore
- && pl_index < entry->textlen)
- || (entry->inverse))
- terminal_clear_to_eol ();
-
- fflush (stdout);
-
- /* Update the display text buffer. */
- if (strlen (printed_line) > (unsigned int) screenwidth)
- /* printed_line[] can include more than screenwidth
- characters if we are under -R and there are escape
- sequences in it. However, entry->text was
- allocated (in display_initialize_display) for
- screenwidth characters only. */
- entry->text = xrealloc (entry->text, strlen (printed_line)+1);
- strcpy (entry->text + i, printed_line + i);
- entry->textlen = pl_index;
-
- /* Lines showing node text are not in inverse. Only modelines
- have that distinction. */
- entry->inverse = 0;
- }
+ /* If there is any text to print, print it. */
+ if (i != pl_index)
+ terminal_put_text (printed_line + i);
+
+ /* If the printed text didn't extend all the way to the edge
+ of the window, and text was appearing between here and the
+ edge of the window, clear from here to the end of the
line. */
+ if ((pl_index < win->width + pl_ignore
+ && pl_index < entry->textlen)
+ || (entry->inverse))
+ terminal_clear_to_eol ();
+
+ fflush (stdout);
+
+ /* Update the display text buffer. */
+ if (strlen (printed_line) > (unsigned int) screenwidth)
+ /* printed_line[] can include more than screenwidth
+ characters if we are under -R and there are escape
+ sequences in it. However, entry->text was
+ allocated (in display_initialize_display) for
+ screenwidth characters only. */
+ entry->text = xrealloc (entry->text, strlen
(printed_line)+1);
+ strcpy (entry->text + i, printed_line + i);
+ entry->textlen = pl_index;
+
+ /* Lines showing node text are not in inverse. Only
modelines
+ have that distinction. */
+ entry->inverse = 0;
+ }
+ }
/* We have done at least one line. Increment our screen line
index, and check against the bottom of the window. */
--- End Message ---
_______________________________________________
Texinfo home page: http://www.gnu.org/software/texinfo/
[email protected]
http://lists.gnu.org/mailman/listinfo/bug-texinfo
