Patch 7.3.715
Problem: Crash when calling setloclist() in BufUnload autocmd. (Marcin
Szamotulski)
Solution: Set w_llist to NULL when it was freed. Also add a test.
(Christian Brabandt)
Files: src/quickfix.c, src/testdir/test49.ok, src/testdir/test49.vim
static int qf_init_ext __ARGS((qf_info_T *qi, char_u *efile, buf_T *buf, typval_T *tv, char_u *errorformat, int newlist, linenr_T lnumfirst, linenr_T lnumlast, char_u *qf_title));
! static void qf_new_list __ARGS((qf_info_T *qi, char_u *qf_title));
static void ll_free_all __ARGS((qf_info_T **pqi));
static int qf_add_entry __ARGS((qf_info_T *qi, qfline_T **prevp, char_u *dir, char_u *fname, int bufnum, char_u *mesg, long lnum, int col, int vis_col, char_u *pattern, int nr, int type, int valid));
static qf_info_T *ll_new_list __ARGS((void));
--- 107,113 ----
};
static int qf_init_ext __ARGS((qf_info_T *qi, char_u *efile, buf_T *buf, typval_T *tv, char_u *errorformat, int newlist, linenr_T lnumfirst, linenr_T lnumlast, char_u *qf_title));
! static void qf_new_list __ARGS((qf_info_T *qi, char_u *qf_title, win_T *wp));
static void ll_free_all __ARGS((qf_info_T **pqi));
static int qf_add_entry __ARGS((qf_info_T *qi, qfline_T **prevp, char_u *dir, char_u *fname, int bufnum, char_u *mesg, long lnum, int col, int vis_col, char_u *pattern, int nr, int type, int valid));
static qf_info_T *ll_new_list __ARGS((void));
***************
*** 266,272 ****
if (newlist || qi->qf_curlist == qi->qf_listcount)
/* make place for a new list */
! qf_new_list(qi, qf_title);
else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
/* Adding to existing list, find last entry. */
for (qfprev = qi->qf_lists[qi->qf_curlist].qf_start;
--- 266,272 ----
if (newlist || qi->qf_curlist == qi->qf_listcount)
/* make place for a new list */
! qf_new_list(qi, qf_title, curwin);
else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
/* Adding to existing list, find last entry. */
for (qfprev = qi->qf_lists[qi->qf_curlist].qf_start;
***************
*** 885,893 ****
* Prepare for adding a new quickfix list.
*/
static void
! qf_new_list(qi, qf_title)
qf_info_T *qi;
char_u *qf_title;
{
int i;
--- 885,894 ----
* Prepare for adding a new quickfix list.
*/
static void
! qf_new_list(qi, qf_title, wp)
qf_info_T *qi;
char_u *qf_title;
+ win_T *wp;
{
int i;
***************
*** 897,903 ****
--- 898,908 ----
* way with ":grep'.
*/
while (qi->qf_listcount > qi->qf_curlist + 1)
+ {
+ if (wp != NULL && wp->w_llist == qi)
+ wp->w_llist = NULL;
qf_free(qi, --qi->qf_listcount);
+ }
/*
* When the stack is full, remove to oldest entry
***************
*** 905,910 ****
--- 910,917 ----
*/
if (qi->qf_listcount == LISTCOUNT)
{
+ if (wp != NULL && wp->w_llist == qi)
+ wp->w_llist = NULL;
qf_free(qi, 0);
for (i = 1; i < LISTCOUNT; ++i)
qi->qf_lists[i - 1] = qi->qf_lists[i];
***************
*** 3181,3187 ****
eap->cmdidx != CMD_vimgrepadd && eap->cmdidx != CMD_lvimgrepadd)
|| qi->qf_curlist == qi->qf_listcount)
/* make place for a new list */
! qf_new_list(qi, *eap->cmdlinep);
else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
/* Adding to existing list, find last entry. */
for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
--- 3188,3194 ----
eap->cmdidx != CMD_vimgrepadd && eap->cmdidx != CMD_lvimgrepadd)
|| qi->qf_curlist == qi->qf_listcount)
/* make place for a new list */
! qf_new_list(qi, *eap->cmdlinep, curwin);
else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
/* Adding to existing list, find last entry. */
for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
***************
*** 3747,3753 ****
if (action == ' ' || qi->qf_curlist == qi->qf_listcount)
/* make place for a new list */
! qf_new_list(qi, title);
else if (action == 'a' && qi->qf_lists[qi->qf_curlist].qf_count > 0)
/* Adding to existing list, find last entry. */
for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
--- 3754,3760 ----
if (action == ' ' || qi->qf_curlist == qi->qf_listcount)
/* make place for a new list */
! qf_new_list(qi, title, wp);
else if (action == 'a' && qi->qf_lists[qi->qf_curlist].qf_count > 0)
/* Adding to existing list, find last entry. */
for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
***************
*** 4029,4035 ****
#endif
/* create a new quickfix list */
! qf_new_list(qi, *eap->cmdlinep);
/* Go through all directories in 'runtimepath' */
p = p_rtp;
--- 4036,4042 ----
#endif
/* create a new quickfix list */
! qf_new_list(qi, *eap->cmdlinep, wp);
/* Go through all directories in 'runtimepath' */
p = p_rtp;
*** ../vim-7.3.714/src/testdir/test49.ok 2010-08-15 21:57:29.000000000 +0200
--- src/testdir/test49.ok 2012-11-14 22:26:13.000000000 +0100
***************
*** 85,92 ****
*** Test 83: OK (2835)
*** Test 84: OK (934782101)
*** Test 85: OK (198689)
! --- Test 86: All tests were run with throwing exceptions on error.
The $VIMNOERRTHROW control is not configured.
! --- Test 86: All tests were run with throwing exceptions on interrupt.
The $VIMNOINTTHROW control is not configured.
! *** Test 86: OK (50443995)
--- 85,94 ----
*** Test 83: OK (2835)
*** Test 84: OK (934782101)
*** Test 85: OK (198689)
! --- Test 86: No Crash for vimgrep on BufUnload
! *** Test 86: OK (0)
! --- Test 87: All tests were run with throwing exceptions on error.
The $VIMNOERRTHROW control is not configured.
! --- Test 87: All tests were run with throwing exceptions on interrupt.
The $VIMNOINTTHROW control is not configured.
! *** Test 87: OK (50443995)
*** ../vim-7.3.714/src/testdir/test49.vim 2010-09-29 16:55:45.000000000 +0200
--- src/testdir/test49.vim 2012-11-14 22:26:13.000000000 +0100
***************
*** 9603,9611 ****
Xcheck 198689
"-------------------------------------------------------------------------- -----
! " Test 86: $VIMNOERRTHROW and $VIMNOINTTHROW support {{{1
"
" It is possible to configure Vim for throwing exceptions on error
" or interrupt, controlled by variables $VIMNOERRTHROW and
--- 9603,9630 ----
Xcheck 198689
+ "-------------------------------------------------------------------------- -----
+ " Test 86 setloclist crash {{{1
+ "
+ " Executing a setloclist() on BufUnload shouldn't crash Vim
+ "-------------------------------------------------------------------------- -----
+ + func F
+ au BufUnload * :call setloclist(0, [{'bufnr':1, 'lnum':1, 'col':1, 'text': 'tango down'}])
+ + :lvimgrep /.*/ *
+ endfunc
+ + XpathINIT
+ + ExecAsScript F
+ + delfunction F
+ Xout "No Crash for vimgrep on BufUnload"
+ Xcheck 0
"-------------------------------------------------------------------------- -----
! " Test 87: $VIMNOERRTHROW and $VIMNOINTTHROW support {{{1
"
" It is possible to configure Vim for throwing exceptions on error
" or interrupt, controlled by variables $VIMNOERRTHROW and
*** ../vim-7.3.714/src/version.c 2012-11-14 20:52:22.000000000 +0100
--- src/version.c 2012-11-14 22:36:45.000000000 +0100
***************
*** 727,728 ****
--- 727,730 ----
{ /* Add new patch number below this line */
+ /**/
+ 715,
/**/
-- One difference between a man and a machine is that a machine is quiet
when well oiled.
On Wed, Nov 14, 2012 at 10:38:29PM +0100, Bram Moolenaar wrote:
> Patch 7.3.715
> Problem: Crash when calling setloclist() in BufUnload autocmd. (Marcin
> Szamotulski)
> Solution: Set w_llist to NULL when it was freed. Also add a test.
> (Christian Brabandt)
> Files: src/quickfix.c, src/testdir/test49.ok, src/testdir/test49.vim
> static int qf_init_ext __ARGS((qf_info_T *qi, char_u *efile, buf_T *buf, typval_T *tv, char_u *errorformat, int newlist, linenr_T lnumfirst, linenr_T lnumlast, char_u *qf_title));
> ! static void qf_new_list __ARGS((qf_info_T *qi, char_u *qf_title));
> static void ll_free_all __ARGS((qf_info_T **pqi));
> static int qf_add_entry __ARGS((qf_info_T *qi, qfline_T **prevp, char_u *dir, char_u *fname, int bufnum, char_u *mesg, long lnum, int col, int vis_col, char_u *pattern, int nr, int type, int valid));
> static qf_info_T *ll_new_list __ARGS((void));
> --- 107,113 ----
> };
> static int qf_init_ext __ARGS((qf_info_T *qi, char_u *efile, buf_T *buf, typval_T *tv, char_u *errorformat, int newlist, linenr_T lnumfirst, linenr_T lnumlast, char_u *qf_title));
> ! static void qf_new_list __ARGS((qf_info_T *qi, char_u *qf_title, win_T *wp));
> static void ll_free_all __ARGS((qf_info_T **pqi));
> static int qf_add_entry __ARGS((qf_info_T *qi, qfline_T **prevp, char_u *dir, char_u *fname, int bufnum, char_u *mesg, long lnum, int col, int vis_col, char_u *pattern, int nr, int type, int valid));
> static qf_info_T *ll_new_list __ARGS((void));
> ***************
> *** 266,272 ****
> if (newlist || qi->qf_curlist == qi->qf_listcount)
> /* make place for a new list */
> ! qf_new_list(qi, qf_title);
> else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
> /* Adding to existing list, find last entry. */
> for (qfprev = qi->qf_lists[qi->qf_curlist].qf_start;
> --- 266,272 ----
> if (newlist || qi->qf_curlist == qi->qf_listcount)
> /* make place for a new list */
> ! qf_new_list(qi, qf_title, curwin);
> else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
> /* Adding to existing list, find last entry. */
> for (qfprev = qi->qf_lists[qi->qf_curlist].qf_start;
> ***************
> *** 885,893 ****
> * Prepare for adding a new quickfix list.
> */
> static void
> ! qf_new_list(qi, qf_title)
> qf_info_T *qi;
> char_u *qf_title;
> {
> int i;
> --- 885,894 ----
> * Prepare for adding a new quickfix list.
> */
> static void
> ! qf_new_list(qi, qf_title, wp)
> qf_info_T *qi;
> char_u *qf_title;
> + win_T *wp;
> {
> int i;
> /*
> * When the stack is full, remove to oldest entry
> ***************
> *** 905,910 ****
> --- 910,917 ----
> */
> if (qi->qf_listcount == LISTCOUNT)
> {
> + if (wp != NULL && wp->w_llist == qi)
> + wp->w_llist = NULL;
> qf_free(qi, 0);
> for (i = 1; i < LISTCOUNT; ++i)
> qi->qf_lists[i - 1] = qi->qf_lists[i];
> ***************
> *** 3181,3187 ****
> eap->cmdidx != CMD_vimgrepadd && eap->cmdidx != CMD_lvimgrepadd)
> || qi->qf_curlist == qi->qf_listcount)
> /* make place for a new list */
> ! qf_new_list(qi, *eap->cmdlinep);
> else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
> /* Adding to existing list, find last entry. */
> for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
> --- 3188,3194 ----
> eap->cmdidx != CMD_vimgrepadd && eap->cmdidx != CMD_lvimgrepadd)
> || qi->qf_curlist == qi->qf_listcount)
> /* make place for a new list */
> ! qf_new_list(qi, *eap->cmdlinep, curwin);
> else if (qi->qf_lists[qi->qf_curlist].qf_count > 0)
> /* Adding to existing list, find last entry. */
> for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
> ***************
> *** 3747,3753 ****
> if (action == ' ' || qi->qf_curlist == qi->qf_listcount)
> /* make place for a new list */
> ! qf_new_list(qi, title);
> else if (action == 'a' && qi->qf_lists[qi->qf_curlist].qf_count > 0)
> /* Adding to existing list, find last entry. */
> for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
> --- 3754,3760 ----
> if (action == ' ' || qi->qf_curlist == qi->qf_listcount)
> /* make place for a new list */
> ! qf_new_list(qi, title, wp);
> else if (action == 'a' && qi->qf_lists[qi->qf_curlist].qf_count > 0)
> /* Adding to existing list, find last entry. */
> for (prevp = qi->qf_lists[qi->qf_curlist].qf_start;
> ***************
> *** 4029,4035 ****
> #endif
> /* create a new quickfix list */
> ! qf_new_list(qi, *eap->cmdlinep);
> /* Go through all directories in 'runtimepath' */
> p = p_rtp;
> --- 4036,4042 ----
> #endif
> /* create a new quickfix list */
> ! qf_new_list(qi, *eap->cmdlinep, wp);
> /* Go through all directories in 'runtimepath' */
> p = p_rtp;
> *** ../vim-7.3.714/src/testdir/test49.ok 2010-08-15 21:57:29.000000000 +0200
> --- src/testdir/test49.ok 2012-11-14 22:26:13.000000000 +0100
> ***************
> *** 85,92 ****
> *** Test 83: OK (2835)
> *** Test 84: OK (934782101)
> *** Test 85: OK (198689)
> ! --- Test 86: All tests were run with throwing exceptions on error.
> The $VIMNOERRTHROW control is not configured.
> ! --- Test 86: All tests were run with throwing exceptions on interrupt.
> The $VIMNOINTTHROW control is not configured.
> ! *** Test 86: OK (50443995)
> --- 85,94 ----
> *** Test 83: OK (2835)
> *** Test 84: OK (934782101)
> *** Test 85: OK (198689)
> ! --- Test 86: No Crash for vimgrep on BufUnload
> ! *** Test 86: OK (0)
> ! --- Test 87: All tests were run with throwing exceptions on error.
> The $VIMNOERRTHROW control is not configured.
> ! --- Test 87: All tests were run with throwing exceptions on interrupt.
> The $VIMNOINTTHROW control is not configured.
> ! *** Test 87: OK (50443995)
> *** ../vim-7.3.714/src/testdir/test49.vim 2010-09-29 16:55:45.000000000 +0200
> --- src/testdir/test49.vim 2012-11-14 22:26:13.000000000 +0100
> ***************
> *** 9603,9611 ****
> Xcheck 198689
> "-------------------------------------------------------------------------- -----
> ! " Test 86: $VIMNOERRTHROW and $VIMNOINTTHROW support {{{1
> "
> " It is possible to configure Vim for throwing exceptions on error
> " or interrupt, controlled by variables $VIMNOERRTHROW and
> --- 9603,9630 ----
> Xcheck 198689
> + "-------------------------------------------------------------------------- -----
> + " Test 86 setloclist crash {{{1
> + "
> + " Executing a setloclist() on BufUnload shouldn't crash Vim
> + "-------------------------------------------------------------------------- -----
> + > + func F
> + au BufUnload * :call setloclist(0, [{'bufnr':1, 'lnum':1, 'col':1, 'text': 'tango down'}])
> + > + :lvimgrep /.*/ *
> + endfunc
> + > + XpathINIT
> + > + ExecAsScript F
> + > + delfunction F
> + Xout "No Crash for vimgrep on BufUnload"
> + Xcheck 0
> "-------------------------------------------------------------------------- -----
> ! " Test 87: $VIMNOERRTHROW and $VIMNOINTTHROW support {{{1
> "
> " It is possible to configure Vim for throwing exceptions on error
> " or interrupt, controlled by variables $VIMNOERRTHROW and
> *** ../vim-7.3.714/src/version.c 2012-11-14 20:52:22.000000000 +0100
> --- src/version.c 2012-11-14 22:36:45.000000000 +0100
> ***************
> *** 727,728 ****
> --- 727,730 ----
> { /* Add new patch number below this line */
> + /**/
> + 715,
> /**/
> -- > One difference between a man and a machine is that a machine is quiet
> when well oiled.
> -- > You received this message from the "vim_dev" maillist.
> Do not top-post! Type your reply below the text you are replying to.
> For more information, visit http://www.vim.org/maillist.php
Anyone else having issues with this patch ?
In my day to day work i use vim in combination with syntastic to do some
linting for me. Since this patch i have frequent segfaults when
syntastic kicks in. It seems related to this patch, maybe not immediatly
due to setloclist itself but something related with quickfix.
I'm sorry but atm i have no idea where to search.
For now i'm stuck on 7.3.714 since that version can keep me working all
day long.
some output:
--- start terminal output ---
vim
Vim: fataal signaal gevangen SEGV
Vim: Finished.
Segmentatiefout
--- end terminal output ---
> In my day to day work i use vim in combination with syntastic to do some
> linting for me. Since this patch i have frequent segfaults when
> syntastic kicks in. It seems related to this patch, maybe not immediatly
> due to setloclist itself but something related with quickfix.
> I'm sorry but atm i have no idea where to search.
> For now i'm stuck on 7.3.714 since that version can keep me working all
> day long.
> some output:
> --- start terminal output ---
> vim
> Vim: fataal signaal gevangen SEGV
On Thu, Nov 22, 2012 at 09:08:48AM +0100, Christian Brabandt wrote:
> On Thu, November 22, 2012 08:47, Ike Devolder wrote:
> [Patch 7.3.315]
> > Anyone else having issues with this patch ?
> > In my day to day work i use vim in combination with syntastic to do some
> > linting for me. Since this patch i have frequent segfaults when
> > syntastic kicks in. It seems related to this patch, maybe not immediatly
> > due to setloclist itself but something related with quickfix.
> > I'm sorry but atm i have no idea where to search.
> > For now i'm stuck on 7.3.714 since that version can keep me working all
> > day long.
> It would help, if you can craft a reproducible example. I have no idea,
> what syntastic is doing.
> regards,
> Christian
> -- > You received this message from the "vim_dev" maillist.
> Do not top-post! Type your reply below the text you are replying to.
> For more information, visit http://www.vim.org/maillist.php
When i'm back at home i'll try to find a way to reproduce this problem
consistently. I've had segfaults in many occasions but I dont yet have a
pattern to make sure the segfault happens.
I'm also not sure the problem lies in the patch itself but it was the
trigger.
On Thu, Nov 22, 2012 at 03:08:14PM +0100, Ike Devolder wrote:
> On Thu, Nov 22, 2012 at 09:08:48AM +0100, Christian Brabandt wrote:
> > On Thu, November 22, 2012 08:47, Ike Devolder wrote:
> > [Patch 7.3.315]
> > > Anyone else having issues with this patch ?
> > > In my day to day work i use vim in combination with syntastic to do some
> > > linting for me. Since this patch i have frequent segfaults when
> > > syntastic kicks in. It seems related to this patch, maybe not immediatly
> > > due to setloclist itself but something related with quickfix.
> > > I'm sorry but atm i have no idea where to search.
> > > For now i'm stuck on 7.3.714 since that version can keep me working all
> > > day long.
> > It would help, if you can craft a reproducible example. I have no idea,
> > what syntastic is doing.
> > regards,
> > Christian
> > -- > > You received this message from the "vim_dev" maillist.
> > Do not top-post! Type your reply below the text you are replying to.
> > For more information, visit http://www.vim.org/maillist.php
> When i'm back at home i'll try to find a way to reproduce this problem
> consistently. I've had segfaults in many occasions but I dont yet have a
> pattern to make sure the segfault happens.
> I'm also not sure the problem lies in the patch itself but it was the
> trigger.
> -- > Ike
i attach a minimum configuration + testfile where i have consistent
segfaults on.
within the archive you can find index.php, it has a syntax error, the
following steps lead to segfault:
vim index.php
- :w (with the syntax error, syntastic kicks in)
- goto end of phpinfo() and add ; :w (error is gone)
- undo u
- :w -> segfault
uname -a:
Linux 3.6.7 #1 SMP PREEMPT Sun Nov 18 09:53:58 CET 2012 x86_64 GNU/Linux
distribution Arch Linux
> On Thu, Nov 22, 2012 at 03:08:14PM +0100, Ike Devolder wrote:
> > On Thu, Nov 22, 2012 at 09:08:48AM +0100, Christian Brabandt wrote:
> > > On Thu, November 22, 2012 08:47, Ike Devolder wrote:
> > > [Patch 7.3.315]
> > > > Anyone else having issues with this patch ?
> > > > In my day to day work i use vim in combination with syntastic to do
> > > > some
> > > > linting for me. Since this patch i have frequent segfaults when
> > > > syntastic kicks in. It seems related to this patch, maybe not
> > > > immediatly
> > > > due to setloclist itself but something related with quickfix.
> > > > I'm sorry but atm i have no idea where to search.
> > > > For now i'm stuck on 7.3.714 since that version can keep me working
> > > > all
> > > > day long.
> > > It would help, if you can craft a reproducible example. I have no idea,
> > > what syntastic is doing.
> > > regards,
> > > Christian
> > When i'm back at home i'll try to find a way to reproduce this problem
> > consistently. I've had segfaults in many occasions but I dont yet have a
> > pattern to make sure the segfault happens.
> > I'm also not sure the problem lies in the patch itself but it was the
> > trigger.
> i attach a minimum configuration + testfile where i have consistent
> segfaults on.
> within the archive you can find index.php, it has a syntax error, the
> following steps lead to segfault:
> vim index.php
> - :w (with the syntax error, syntastic kicks in)
> - goto end of phpinfo() and add ; :w (error is gone)
> - undo u
> - :w -> segfault
> uname -a:
> Linux 3.6.7 #1 SMP PREEMPT Sun Nov 18 09:53:58 CET 2012 x86_64 GNU/Linux
> distribution Arch Linux
Ike Devolder wrote:
>> i attach a minimum configuration + testfile where i have consistent
>> segfaults on.
>> within the archive you can find index.php, it has a syntax error, the
>> following steps lead to segfault:
>> vim index.php
>> - :w (with the syntax error, syntastic kicks in)
>> - goto end of phpinfo() and add ; :w (error is gone)
>> - undo u
>> - :w -> segfault
>> uname -a:
>> Linux 3.6.7 #1 SMP PREEMPT Sun Nov 18 09:53:58 CET 2012 x86_64 GNU/Linux
>> distribution Arch Linux
> ok i forgot on pre-requisite: you must have php installed so the syntastic
> plugin can run php -l to get the syntax error showing.
> --Ike
Hi Ike
I can reproduce the crash using your description with
Vim-7.3.725 on Linux x86_64.
Valgrind gives the following error:
==6815== Invalid read of size 4
==6815== at 0x58ED64: ex_make (quickfix.c:2869)
==6815== by 0x4888F1: do_cmdline (ex_docmd.c:2677)
==6815== by 0x4468BF: call_func (eval.c:22540)
==6815== by 0x449722: get_func_tv (eval.c:8322)
==6815== by 0x46A24E: eval7 (eval.c:5158)
==6815== by 0x468F27: eval6 (eval.c:4810)
==6815== by 0x468A77: eval5 (eval.c:4626)
==6815== by 0x4682D0: eval4 (eval.c:4319)
==6815== by 0x468166: eval3 (eval.c:4231)
==6815== by 0x444AD4: eval1 (eval.c:4160)
==6815== by 0x4445D8: eval0 (eval.c:4042)
==6815== by 0x44769D: ex_let (eval.c:1897)
==6815== by 0x4888F1: do_cmdline (ex_docmd.c:2677)
==6815== by 0x4468BF: call_func (eval.c:22540)
==6815== by 0x449722: get_func_tv (eval.c:8322)
==6815== by 0x46A24E: eval7 (eval.c:5158)
==6815== by 0x468F27: eval6 (eval.c:4810)
==6815== by 0x468A77: eval5 (eval.c:4626)
==6815== by 0x4682D0: eval4 (eval.c:4319)
==6815== by 0x468166: eval3 (eval.c:4231)
==6815== by 0x444AD4: eval1 (eval.c:4160)
==6815== by 0x4445D8: eval0 (eval.c:4042)
==6815== by 0x44769D: ex_let (eval.c:1897)
==6815== by 0x4888F1: do_cmdline (ex_docmd.c:2677)
==6815== by 0x4468BF: call_func (eval.c:22540)
==6815== by 0x449722: get_func_tv (eval.c:8322)
==6815== by 0x448AFD: ex_call (eval.c:3465)
==6815== by 0x4888F1: do_cmdline (ex_docmd.c:2677)
==6815== by 0x4468BF: call_func (eval.c:22540)
==6815== by 0x449722: get_func_tv (eval.c:8322)
==6815== by 0x448AFD: ex_call (eval.c:3465)
==6815== by 0x4888F1: do_cmdline (ex_docmd.c:2677)
==6815== by 0x4BFFE4: apply_autocmds_group (fileio.c:9435)
==6815== by 0x4BAF27: buf_write (fileio.c:9049)
==6815== by 0x47355D: do_write (ex_cmds.c:2692)
==6815== by 0x4888F1: do_cmdline (ex_docmd.c:2677)
==6815== by 0x54D9DA: nv_colon (normal.c:5417)
==6815== by 0x543C95: normal_cmd (normal.c:1198)
==6815== by 0x685ADE: main_loop (main.c:1294)
==6815== by 0x6854EA: main (main.c:998)
==6815== Address 0x8 is not stack'd, malloc'd or (recently) free'd
quickfix.c:
2861 if (wp != NULL)
2862 qi = GET_LOC_LIST(wp);
2863 #ifdef FEAT_AUTOCMD
2864 if (au_name != NULL)
2865 {
2866 apply_autocmds(EVENT_QUICKFIXCMDPOST, au_name,
2867 curbuf->b_fname,
TRUE, curbuf);
2868 if (qi->qf_curlist < qi->qf_listcount)
2869 res = qi->qf_lists[qi->qf_curlist].qf_count;
2870 else
2871 res = 0;
2872 }
Printing a few pointers just before line quickfix.c:2868, I get this:
I don't know yet why some pointers are NULL but perhaps someone
can make sense of it.
I don't need to do 'undo' to reproduce the crash. This also crashes:
- vim index.php
- :w (with the syntax error, syntastic kicks in)
- goto end of phpinfo() and add ; :w (error is gone)
- remove previously added ; (by pressing x on ; character in command mode)
- :w to save again -> crash
> I don't know yet why some pointers are NULL but perhaps someone
> can make sense of it.
I see. Try this patch:
diff --git a/src/quickfix.c b/src/quickfix.c
--- a/src/quickfix.c
+++ b/src/quickfix.c
@@ -2863,7 +2863,7 @@
{
apply_autocmds(EVENT_QUICKFIXCMDPOST, au_name,
curbuf->b_fname, TRUE, curbuf);
- if (qi->qf_curlist < qi->qf_listcount)
+ if (qi != NULL && qi->qf_curlist < qi->qf_listcount)
res = qi->qf_lists[qi->qf_curlist].qf_count;
else
res = 0;
regards,
Christian
-- Je mehr Leute es sind, die eine Sache glauben, desto größer ist die
Wahrscheinlichkeit, daß die Ansicht falsch ist. Menschen, die Recht
haben, stehen meistens allein.
-- Søren Kierkegaard
===
changeset: 3918:4f0ddf4137ee
tag: v7-3-715
user: Bram Moolenaar <b...@vim.org>
date: Wed Nov 14 22:38:08 2012 +0100
files: src/quickfix.c src/testdir/test49.ok
src/testdir/test49.vim src/version.c
description:
updated for version 7.3.715
Problem: Crash when calling setloclist() in BufUnload autocmd. (Marcin
Szamotulski)
Solution: Set w_llist to NULL when it was freed. Also add a test.
(Christian Brabandt)
===
I find it odd that a function called qf_new_list() clears
wp->w_llist (set it to NULL) and does not set it back
to something else. The name of the function "qf_new_list()"
suggests that it should create another list, so perhaps
it should set wp->w_llist to something else. But I don't
understand the code here.
> ===
> changeset: 3918:4f0ddf4137ee
> tag: v7-3-715
> user: Bram Moolenaar <b...@vim.org>
> date: Wed Nov 14 22:38:08 2012 +0100
> files: src/quickfix.c src/testdir/test49.ok
> src/testdir/test49.vim src/version.c
> description:
> updated for version 7.3.715
> Problem: Crash when calling setloclist() in BufUnload autocmd. (Marcin
> Szamotulski)
> Solution: Set w_llist to NULL when it was freed. Also add a test.
> (Christian Brabandt)
> ===
> I find it odd that a function called qf_new_list() clears
> wp->w_llist (set it to NULL) and does not set it back
> to something else. The name of the function "qf_new_list()"
> suggests that it should create another list, so perhaps
> it should set wp->w_llist to something else. But I don't
> understand the code here.
Indeed. I think GET_LOC_LIST should be defined as ll_get_or_alloc_list(wp)
> > ===
> > changeset: 3918:4f0ddf4137ee
> > tag: v7-3-715
> > user: Bram Moolenaar <b...@vim.org>
> > date: Wed Nov 14 22:38:08 2012 +0100
> > files: src/quickfix.c src/testdir/test49.ok
> > src/testdir/test49.vim src/version.c
> > description:
> > updated for version 7.3.715
> > Problem: Crash when calling setloclist() in BufUnload autocmd. (Marcin
> > Szamotulski)
> > Solution: Set w_llist to NULL when it was freed. Also add a test.
> > (Christian Brabandt)
> > ===
> > I find it odd that a function called qf_new_list() clears
> > wp->w_llist (set it to NULL) and does not set it back
> > to something else. The name of the function "qf_new_list()"
> > suggests that it should create another list, so perhaps
> > it should set wp->w_llist to something else. But I don't
> > understand the code here.
> Indeed. I think GET_LOC_LIST should be defined as > ll_get_or_alloc_list(wp)
ll_get_or_alloc_list() can still return NULL, thus your check is needed
anyway.
-- hundred-and-one symptoms of being an internet addict:
59. Your wife says communication is important in a marriage...so you buy
another computer and install a second phone line so the two of you can
chat.
> > > ===
> > > changeset: 3918:4f0ddf4137ee
> > > tag: v7-3-715
> > > user: Bram Moolenaar <b...@vim.org>
> > > date: Wed Nov 14 22:38:08 2012 +0100
> > > files: src/quickfix.c src/testdir/test49.ok
> > > src/testdir/test49.vim src/version.c
> > > description:
> > > updated for version 7.3.715
> > > Problem: Crash when calling setloclist() in BufUnload autocmd. (Marcin
> > > Szamotulski)
> > > Solution: Set w_llist to NULL when it was freed. Also add a test.
> > > (Christian Brabandt)
> > > ===
> > > I find it odd that a function called qf_new_list() clears
> > > wp->w_llist (set it to NULL) and does not set it back
> > > to something else. The name of the function "qf_new_list()"
> > > suggests that it should create another list, so perhaps
> > > it should set wp->w_llist to something else. But I don't
> > > understand the code here.
> > Indeed. I think GET_LOC_LIST should be defined as > > ll_get_or_alloc_list(wp)
> ll_get_or_alloc_list() can still return NULL, thus your check is needed
> anyway.
It is still not right so. I'll look into it later.
> On Fr, 23 Nov 2012, Bram Moolenaar wrote:
> > Christian Brabandt wrote:
> > > Hi Dominique!
> > > On Do, 22 Nov 2012, Dominique Pellé wrote:
> > > > I can't tell whether that's the right fix, but I confirm that
> > > > your proposed change avoids the crash at least.
> > > > wp->w_llist was set to NULL at line quickfix.c:914 which
> > > > Solution: Set w_llist to NULL when it was freed. Also add a test.
> > > > (Christian Brabandt)
> > > > ===
> > > > I find it odd that a function called qf_new_list() clears
> > > > wp->w_llist (set it to NULL) and does not set it back
> > > > to something else. The name of the function "qf_new_list()"
> > > > suggests that it should create another list, so perhaps
> > > > it should set wp->w_llist to something else. But I don't
> > > > understand the code here.
> > > Indeed. I think GET_LOC_LIST should be defined as
> > > ll_get_or_alloc_list(wp)
> > ll_get_or_alloc_list() can still return NULL, thus your check is needed
> > anyway.
> It is still not right so. I'll look into it later.
> regards,
> Christian
I have tested the patch on top of 725
had one segfault already (could not yet reproduce)
but the following i can reproduce every time:
same configuration as before, same file (index.php)
- open index.php (it still has the syntax error
- :w (syntastic kicks in)
- fix the error by adding ; after phpinfo()
- :w (no more error in the file)
- remove the ; again
- :w (syntastic kicks in)
- :w ->E776: No location list
"index.php" 2L, 16C written
Error detected while processing function <SNR>31_UpdateErrors..<SNR>31_AutoToggleLocList..<SNR>31_ShowLocList:
line 3:
E776: No location list
Press ENTER or type command to continue
no crash in this case, if i continue after this issue i get a second quickfix buffer.
thx for the intermediate fix it is definatly an improvement.
> same configuration as before, same file (index.php)
> - open index.php (it still has the syntax error
> - :w (syntastic kicks in)
> - fix the error by adding ; after phpinfo()
> - :w (no more error in the file)
> - remove the ; again
> - :w (syntastic kicks in)
> - :w ->E776: No location list
> "index.php" 2L, 16C written
> Error detected while processing function > <SNR>31_UpdateErrors..<SNR>31_AutoToggleLocList..<SNR>31_ShowLocList:
> line 3:
> E776: No location list
> Press ENTER or type command to continue
> no crash in this case, if i continue after this issue i get a second quickfix > buffer.
> thx for the intermediate fix it is definatly an improvement.
The following patch seems to fix all the issues mentioned so far. It basically does this by reverting Patch 7.3.715 and making sure, no data is freed more than once. Unfortunately, I have not been able to come up with a simple test case for the syntastic problem, that could be included.
If anybody can come up with such a test (using no plugin, but just using vim -u NONE -U NONE -N) a tip is appreciated.
regards,
Christian
-- Ich bin geldgierig. Als Finanzminister mu� man geldgierig sein.
-- Hans Eichel
> On Fr, 23 Nov 2012, Ike Devolder wrote:
> > but the following i can reproduce every time:
> > same configuration as before, same file (index.php)
> > - open index.php (it still has the syntax error
> > - :w (syntastic kicks in)
> > - fix the error by adding ; after phpinfo()
> > - :w (no more error in the file)
> > - remove the ; again
> > - :w (syntastic kicks in)
> > - :w ->E776: No location list
> > "index.php" 2L, 16C written
> > Error detected while processing function > > <SNR>31_UpdateErrors..<SNR>31_AutoToggleLocList..<SNR>31_ShowLocList:
> > line 3:
> > E776: No location list
> > Press ENTER or type command to continue
> > no crash in this case, if i continue after this issue i get a second quickfix > > buffer.
> > thx for the intermediate fix it is definatly an improvement.
> The following patch seems to fix all the issues mentioned so far. It > basically does this by reverting Patch 7.3.715 and making sure, no data > is freed more than once. Unfortunately, I have not been able to come up > with a simple test case for the syntastic problem, that could be > included.
> If anybody can come up with such a test (using no plugin, but just using > vim -u NONE -U NONE -N) a tip is appreciated.
Thanks for the patch. Have you tried running with valgrind, checking
that no memory is leaked or accessing already freed memory? It's only
manual testing, better than nothing.
-- hundred-and-one symptoms of being an internet addict:
73. You give your dog used motherboards instead of bones
On Sun, Nov 25, 2012 at 02:29:59PM +0100, Christian Brabandt wrote:
> Hi everybody
> On Fr, 23 Nov 2012, Ike Devolder wrote:
> > but the following i can reproduce every time:
> > same configuration as before, same file (index.php)
> > - open index.php (it still has the syntax error
> > - :w (syntastic kicks in)
> > - fix the error by adding ; after phpinfo()
> > - :w (no more error in the file)
> > - remove the ; again
> > - :w (syntastic kicks in)
> > - :w ->E776: No location list
> > "index.php" 2L, 16C written
> > Error detected while processing function > > <SNR>31_UpdateErrors..<SNR>31_AutoToggleLocList..<SNR>31_ShowLocList:
> > line 3:
> > E776: No location list
> > Press ENTER or type command to continue
> > no crash in this case, if i continue after this issue i get a second quickfix > > buffer.
> > thx for the intermediate fix it is definatly an improvement.
> The following patch seems to fix all the issues mentioned so far. It > basically does this by reverting Patch 7.3.715 and making sure, no data > is freed more than once. Unfortunately, I have not been able to come up > with a simple test case for the syntastic problem, that could be > included.
> If anybody can come up with such a test (using no plugin, but just using > vim -u NONE -U NONE -N) a tip is appreciated.
> regards,
> Christian
> -- > Ich bin geldgierig. Als Finanzminister mu man geldgierig sein.
> -- Hans Eichel
> -- > You received this message from the "vim_dev" maillist.
> Do not top-post! Type your reply below the text you are replying to.
> For more information, visit http://www.vim.org/maillist.php > diff --git a/src/quickfix.c b/src/quickfix.c
> --- a/src/quickfix.c
> +++ b/src/quickfix.c
> @@ -898,11 +898,7 @@
> * way with ":grep'.
> */
> while (qi->qf_listcount > qi->qf_curlist + 1)
> - {
> - if (wp != NULL && wp->w_llist == qi)
> - wp->w_llist = NULL;
> qf_free(qi, --qi->qf_listcount);
> - }
