Skip to contentSkip to conversation

bug found (Re: git-fast-export SIGSEGV on solaris + backtrace)

PATCH SungHyun Nam 3 participants 3 messages Jun 15, 2016 1 message not in the archive

lore
Share & tools
lore
#1 lore raw
In reply to a message not in the archive

Hello,

From the gdb trace:

(gdb) p deco
$1 = (struct object_decoration *) 0x193150 (gdb) p idnums
$3 = {name = 0x0, size = 1500, nr = 7, hash = 0x190270}

It was 'off-by-one' error.
0x190270 + (1500 * 8) = 0x193150.

And the code says it:

        for (i = 0; i < idnums.size; ++i) {
                deco++;
                if (deco && deco->base && deco->base->type == 1) {

The 'deco' should be post-incremented? or
Checking code should be (i < idnums.size - 1)?

Regards,
namsh

And very minor cleanup:

decorate.c +1 -1
index 23f6b00..d8b428c 100644
@@ -36,7 +36,7 @@ static void *insert_decoration(struct decoration *n, struct o>
 static void grow_decoration(struct decoration *n)
 {
        int i;
-       int old_size = n->size;
+       int old_size;
        struct object_decoration *old_hash;
 
        old_size = n->size;
Quoted 100 lines · SungHyun Nam

Hello,

Because recent GIT test failed on t9301-fast-export.sh, I traced it. And I found git-fast-export killed by SIGSEGV. I include a gdb backtrace below.
If you want to me to check other things, please let me know.

Regards,
namsh

[marks] ~/srcs/git/t/trash directory[66]$ gdb ../../git-fast-export GNU gdb 6.0
Copyright 2003 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions.
There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "sparc-sun-solaris2.9"...
(gdb) r --export-marks=tmp-marks HEAD
Starting program: /flyvo2/home/namsh/srcs/git/git-fast-export --export-marks=tmp-marks HEAD
blob
mark :1
data 8
Wohlauf

reset refs/heads/marks
commit refs/heads/marks
mark :2
author A U Thor <[email protected]> 1112911993 -0700 committer C O Mitter <[email protected]> 1112911993 -0700 data 8
initial
M 100644 :1 file

blob
mark :3
data 9
die Luft

blob
mark :4
data 12
geht frisch

commit refs/heads/marks
mark :5
author A U Thor <[email protected]> 1112912053 -0700 committer C O Mitter <[email protected]> 1112912053 -0700 data 7
second
from :2
M 100644 :3 file
M 100644 :4 file2

blob
mark :6
data 4
und

commit refs/heads/marks
mark :7
author A U Thor <[email protected]> 1112912113 -0700 committer C O Mitter <[email protected]> 1112912113 -0700 data 6
third
from :5
M 100644 :6 file2

Program received signal SIGSEGV, Segmentation fault. 0x00043024 in export_marks (file=0xffbff0df "tmp-marks")
at builtin-fast-export.c:384
384 if (deco && deco->base && deco->base->type == 1) {
(gdb) p deco
$1 = (struct object_decoration *) 0x193150 (gdb) p deco->base
$2 = (struct object *) 0x2009
(gdb) p *(struct object_decoration *) 0x193150 $5 = {base = 0x2009, decoration = 0x0}
(gdb) p idnums
$3 = {name = 0x0, size = 1500, nr = 7, hash = 0x190270} (gdb) p *(struct object_decoration *) 0x190270 $6 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x190280 $9 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x190290 $10 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x1902a0 $11 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x1902b0 $12 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x1902c0 $13 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x193140 $14 = {base = 0x0, decoration = 0x0}
(gdb) p *(struct object_decoration *) 0x193160 $15 = {base = 0x35323962, decoration = 0x63613534} (gdb)

#2 lore raw
Quoted 7 lines · SungHyun Nam

And the code says it:

        for (i = 0; i < idnums.size; ++i) {
                deco++;
                if (deco && deco->base && deco->base->type == 1) {

The 'deco' should be post-incremented? or
Checking code should be (i < idnums.size - 1)?

The variable "deco" is a pointer that walks over a hashtable from its offset 0 to its end, so it can never be NULL (well, the code increments before it tests the variable for NULLness, so it is clear that the test is bogus).

What was I smoking when I applied df6a7ff (builtin-fast-export: Add importing and exporting of revision marks, 2008-06-11), I have to wonder...

Thanks for the fix.

#3 lore raw

Hi,

Quoted 3 lines · Junio C Hamano

What was I smoking when I applied df6a7ff (builtin-fast-export: Add importing and exporting of revision marks, 2008-06-11), I have to wonder...

FWIW I missed it in my review, too... Well, Junio, it seems we should avoid smoking together for a while.

Ciao,
Dscho