TUI + gdbserver broken?

TUI + gdbserver broken?

[Pedro Alves]

Hi,

TUI + gdbserver seems to be broken on head.  Can't reproduce when
TUI + native debugging.

main is a simple hello world.

Could this still be related to the recent get_selected_frame
changes?  I tried to blindly change all of them in
tui/tui-hooks.c to deprecated_get_selected_frame, but it
didn't fix it.

I also see some nasty extra screen repaints in
tui mode while stepping I didn't see before.  Any ideas?

$gdbserver :9999 main

$gdb main.exe
GNU gdb 6.6.50.20070312-cvs
Copyright (C) 2007 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 "i686-pc-cygwin"...
(gdb) tar rem :9999
Remote debugging using :9999
[New Thread 5340]
[Switching to Thread 5340]
0x7c911231 in ?? ()
(gdb) c
Continuing.

Program exited normally.
(gdb) c
The program is not being run.

----

$gdbserver :9999 main

$gdbtui main
tar rem :9999
Remote debugging using :9999
[New Thread 5612]
[Switching to Thread 5612]
0x7c911231 in ?? ()
(gdb) c
Continuing.
Remote communication error: Software caused connection abort.
(gdb) c
Continuing.
putpkt: write failed: Software caused connection abort.

Re: TUI + gdbserver broken?

[Daniel Jacobowitz]

On Wed, Mar 14, 2007 at 12:55:40AM +0000, Pedro Alves wrote:
> Could this still be related to the recent get_selected_frame
> changes?

I doubt it.

> (gdb) c
> Continuing.
> Remote communication error: Software caused connection abort.

I recommend set debug remote.  What does gdbserver do when this happens?

-- 
Daniel Jacobowitz
CodeSourcery

Re: TUI + gdbserver broken?

[Pedro Alves]

Daniel Jacobowitz escreveu:
> On Wed, Mar 14, 2007 at 12:55:40AM +0000, Pedro Alves wrote:
>   
>> Could this still be related to the recent get_selected_frame
>> changes?
>>     
>
> I doubt it.
>
>   

But it is.  Going back to cvs before the following patch
fixes both problems I see on Cygwin.

2007-02-27  Daniel Jacobowitz  <dan@codesourcery.com>

    * frame.c (deprecated_selected_frame): Rename to...
    (selected_frame): ...this.  Make static.
    (get_selected_frame, select_frame): Update.
    * frame.h (deprected_select_frame): Delete.
    (deprecated_safe_get_selected_frame): Update comments.

    * breakpoint.c, cli/cli-cmds.c, f-valprint.c, infcmd.c, inflow.c,
    infrun.c, stack.c, tui/tui-disasm.c, tui/tui-source.c,
    tui/tui-winsource.c, valops.c, varobj.c, findvar.c, macroscope.c,
    parse.c, regcache.h, sh64-tdep.c, tui/tui-hooks.c, tui/tui-win.c,
    tui/tui.c: Replace references to deprecated_selected_frame.


>> (gdb) c
>> Continuing.
>> Remote communication error: Software caused connection abort.
>>     
>
> I recommend set debug remote.  What does gdbserver do when this happens?
>
>   

It ends successfully.
I'm out of time for today to look further.

Cheers,
Pedro Alves

Re: TUI + gdbserver broken?

[Pedro Alves]

Pedro Alves wrote:
>
> But it is.  Going back to cvs before the following patch
> fixes both problems I see on Cygwin.
>
> 2007-02-27  Daniel Jacobowitz  <dan@codesourcery.com>
>
>    * frame.c (deprecated_selected_frame): Rename to...
>    (selected_frame): ...this.  Make static.
>    (get_selected_frame, select_frame): Update.
>    * frame.h (deprected_select_frame): Delete.
>    (deprecated_safe_get_selected_frame): Update comments.
>
>    * breakpoint.c, cli/cli-cmds.c, f-valprint.c, infcmd.c, inflow.c,
>    infrun.c, stack.c, tui/tui-disasm.c, tui/tui-source.c,
>    tui/tui-winsource.c, valops.c, varobj.c, findvar.c, macroscope.c,
>    parse.c, regcache.h, sh64-tdep.c, tui/tui-hooks.c, tui/tui-win.c,
>    tui/tui.c: Replace references to deprecated_selected_frame.
>
>
>>> (gdb) c
>>> Continuing.
>>> Remote communication error: Software caused connection abort.
>>>     
>>
>> I recommend set debug remote.  What does gdbserver do when this happens?
>>
>>   
>
> It ends successfully.
> I'm out of time for today to look further.
>

I've looked further.

Here's what happens, on TUI + gdbserver:

(...)
Packet vCont (verbose-resume) is supported
Sending packet: $vCont;c#a8...Ack
Packet received: W00
Sending packet: $g#67...Remote communication error: Software caused
connection abort.
(gdb) c
Continuing.

gdb is sending a g packet after a W.  Something in gdb didn't
realize that the the target already exited.  I'm posting
a patch at gdb-patches with a deeper rationale.

Cheers,
Pedro Alves

Re: TUI + gdbserver broken?

[Denis PILAT]

Sorry I originaly post this mail on gdb-patches mailing list!

Pedro Alves wrote:
> Hi,
>
> TUI + gdbserver seems to be broken on head.  Can't reproduce when
> TUI + native debugging.
>
As far as the TUI is concerned:
It's broken with Solaris Native debugging.
I've spent a few hours debugging that problem but I have not solution yet !!
I also tried using deprecated_get_selected_frame without any better
results :(

This problem is present in the CVS HEAD but also in the 6.6 released
version, so I don't think it comes from recent changes.

-- 
Denis

P.S. here is the backtrace I got in the native solaris TUI
Program received signal SIGTTIN, Stopped (tty
input).
Line: ??   PC: 0xff3a3958
0xff19e880 in _read () from /usr/lib/libc.so.1
(gdb) btto change it and/or distribute copies of it under certain
conditions.
#0  0xff19e880 in _read () from /usr/lib/libc.so.1
#1  0xff33da4c in _chkinput () from /usr/lib/libcurses.so.1ty" for details.
#2  0xff34e680 in wrefresh () from /usr/lib/libcurses.so.1
#3  0x00095628 in tui_refresh_win (win_info=0x2e73c4)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/tui/tui-wingeneral.c:63testsuite/gdb.base/break.c, 

line 90.
#4  0x00092b18 in tui_show_locator_content () at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/tui/tui-stack.c:261
#5  0x00092db8 in tui_show_frame_info
(fi=0x317890)uild-sts-gdb-repository/cvshead-sol8/gdb/testsuite/gdb.base/break
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/tui/tui-stack.c:341
#6  0x0008e6bc in tui_selected_frame_level_changed_hook (level=0)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/tui/tui-hooks.c:249
#7  0x0016a4f8 in select_frame (fi=0x317890) at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/frame.c:986
#8  0x0016a3bc in get_selected_frame (message=0x0) at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/frame.c:956
#9  0x0016a48c in deprecated_safe_get_selected_frame ()
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/frame.c:972
#10 0x0008e684 in tui_selected_frame_level_changed_hook (level=-1)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/tui/tui-hooks.c:233
#11 0x0016a4f8 in select_frame (fi=0x0) at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/frame.c:986
#12 0x0016a80c in reinit_frame_cache () at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/frame.c:1086
#13 0x000d66e8 in handle_inferior_event (ecs=0xffbee0a0)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/infrun.c:1306
#14 0x000d5f38 in wait_for_inferior () at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/infrun.c:1007
#15 0x00177020 in startup_inferior (ntraps=2) at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/fork-child.c:435
#16 0x00063bc0 in procfs_init_inferior (pid=3577048)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/procfs.c:4872
#17 0x00176fa4 in fork_inferior (
    exec_file_arg=0x340418
"/project/flexdbug/user/pd10/gdb/build-sts-gdb-repository/cvshead-sol8/gdb/testsuite/gdb.base/break",
    allargs=0x327008 "", env=0x30bdd8, traceme_fun=0x63bcc
<procfs_set_exec_trap>, init_trace_fun=0x63940 <procfs_init_inferior>,
    pre_trace_fun=0, shell_file_arg=0xffbef892 "/bin/tcsh")
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/fork-child.c:396
#18 0x00063ebc in procfs_create_inferior (
    exec_file=0x340418
"/project/flexdbug/user/pd10/gdb/build-sts-gdb-repository/cvshead-sol8/gdb/testsuite/gdb.base/break",
   allargs=0x327008 ""Quit
) at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/procfs.c:5101
#19 0x0008b640 in sol_thread_create_inferior (
    exec_file=0x340418
"/project/flexdbug/user/pd10/gdb/build-sts-gdb-repository/cvshead-sol8/gdb/testsuite/gdb.base/break",
    allargs=0x327008 "", env=0x30bdd8, from_tty=1)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/sol-thread.c:758
#20 0x000fdaec in find_default_create_inferior (
    exec_file=0x340418
"/project/flexdbug/user/pd10/gdb/build-sts-gdb-repository/cvshead-sol8/gdb/testsuite/gdb.base/break",
    allargs=0x327008 "", env=0x30bdd8, from_tty=1) at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/target.c:1768
#21 0x000d2724 in run_command_1 (args=0x0, from_tty=1, tbreak_at_main=0)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/infcmd.c:558
#22 0x0007cb74 in do_cfunc (c=0x30b978,
    args=0x340418
"/project/flexdbug/user/pd10/gdb/build-sts-gdb-repository/cvshead-sol8/gdb/testsuite/gdb.base/break", 

from_tty=1)
    at
/project/flexdbug/user/pd10/gdb/sts-gdb-repository/vendor/cvs/head/src/gdb/cli/cli-decode.c:62

Re: TUI + gdbserver broken?

[Pedro Alves]

Denis PILAT wrote:
>
> Pedro Alves wrote:
> > Hi,
> >
> > TUI + gdbserver seems to be broken on head.  Can't reproduce when
> > TUI + native debugging.
> >
> As far as the TUI is concerned:
> It's broken with Solaris Native debugging.
> I've spent a few hours debugging that problem but I have not solution yet !!
> I also tried using deprecated_get_selected_frame without any better
> results :(
>
> This problem is present in the CVS HEAD but also in the 6.6 released
> version, so I don't think it comes from recent changes.
>

Just tested a GNU gdb 6.6.50.20070101-cvs configured as sparc-sun-solaris2.9,
built with gcc 3.4.6 I had here - TUI works fine.

$uname -srvmpi
SunOS 5.9 Generic_117171-13 sun4u sparc SUNW,Sun-Fire-480R

Don't know if it adds any value, but I tried it under solaris xterm
displaying on cygwin X,
under GNU bash, version 3.2.0(1)-release (sparc-sun-solaris2.9)

I'll try doing a binary search on Cygwin when I get home.

Cheers,
Pedro Alves

Re: TUI + gdbserver broken?

[Denis PILAT]

Pedro Alves wrote:
> Denis PILAT wrote:
>>
>> Pedro Alves wrote:
>> > Hi,
>> >
>> > TUI + gdbserver seems to be broken on head.  Can't reproduce when
>> > TUI + native debugging.
>> >
>> As far as the TUI is concerned:
>> It's broken with Solaris Native debugging.
>> I've spent a few hours debugging that problem but I have not solution 
>> yet !!
>> I also tried using deprecated_get_selected_frame without any better
>> results :(
>>
>> This problem is present in the CVS HEAD but also in the 6.6 released
>> version, so I don't think it comes from recent changes.
>>
>
> Just tested a GNU gdb 6.6.50.20070101-cvs configured as 
> sparc-sun-solaris2.9,
> built with gcc 3.4.6 I had here - TUI works fine.
>
> $uname -srvmpi
> SunOS 5.9 Generic_117171-13 sun4u sparc SUNW,Sun-Fire-480R
>
I tested 6.6 TUI configured as sparc-sun-solaris2.8 built with gcc 3.2.2.
uname gives: SunOS  5.8 Generic_117350-23 sun4u sparc

I'm looking for a 2.9 Solaris to make some tests

Denis
> Don't know if it adds any value, but I tried it under solaris xterm
> displaying on cygwin X,
> under GNU bash, version 3.2.0(1)-release (sparc-sun-solaris2.9)
>
> I'll try doing a binary search on Cygwin when I get home.
>
> Cheers,
> Pedro Alves
>