Problems with SrcWin

Problems with SrcWin

[Tommy]

Hi all,

 I'm trying to fix problems happened in a porting of Insight (version
6.7.1.) They are:

 (1) after Insight is connected to target, the Control menu and
button items are still disabled. They'll be enabled after some actions
are done (ex. execute 'stepi' from ConsoleWindow.)
     I've checked debug info from the ManagedWin's DebugWin - it
seems to call "SrcBar::enable_ui 2" in connecting but call
"SrcBar::enable_ui 1" in stepping. I know many ports of Insight didn't
patch these tcl/tk files much but leave target-dependent things in
their port C files. Do I miss something? or this should be fixed in
tcl/tk layer?

 (2) after Insight is connected to target, the SrcWin displays
nothing... (there is no 'file', 'load', and 'symbol-file' command
executed.) I think, if there is no corresponding symbolic info, it
should fetch memory contents near $PC and display disassembled
instructions.

 Thanks.

Re: Problems with SrcWin

[Keith Seitz]

Tommy wrote:
>  (1) after Insight is connected to target, the Control menu and
> button items are still disabled. They'll be enabled after some actions
> are done (ex. execute 'stepi' from ConsoleWindow.)

This is usually caused by inferior code not setting a fake ptid for the 
UI. This is a very common problem with non-native targets. You can check 
this by inspecting the function gdb_target_has_execution. Check out what 
  the remote target does in extended_remote_create_inferior_1 (it sets 
inferior_ptid to some made up non-NULL entity). The value does not 
matter -- that it is non-NULL is all that is required.

>  (2) after Insight is connected to target, the SrcWin displays
> nothing... (there is no 'file', 'load', and 'symbol-file' command
> executed.) I think, if there is no corresponding symbolic info, it
> should fetch memory contents near $PC and display disassembled
> instructions.

Open a console window and type "list". If gdb does not return something, 
than Insight cannot, either. Insight will disassemble around $PC *if* 
gdb can find a function containing the PC (so that it knows the bounds 
of the function to disassemble). This is a gdb limitation that Insight 
has not worked around yet (and probably never will).

Keith

Re: Problems with SrcWin

[Tommy Lin]

>>  (1) after Insight is connected to target, the Control menu and
>> button items are still disabled. They'll be enabled after some actions
>> are done (ex. execute 'stepi' from ConsoleWindow.)
>
> This is usually caused by inferior code not setting a fake ptid for the UI.
> This is a very common problem with non-native targets. You can check this by
> inspecting the function gdb_target_has_execution. Check out what  the remote
> target does in extended_remote_create_inferior_1 (it sets inferior_ptid to
> some made up non-NULL entity). The value does not matter -- that it is
> non-NULL is all that is required.

  I found my target's xxxx_create_inferior() isn't called before
gdb_target_has_execution(). Maybe it's not a good port. So, I fix this
issue by removing the inferior_ptid part from
gdb_target_has_execution() directly.
  It seems can be fixed in srcbar.itcl:do_connect by setting Control
$Menu and $Tool into 'normal'.

>>  (2) after Insight is connected to target, the SrcWin displays
>> nothing... (there is no 'file', 'load', and 'symbol-file' command
>> executed.) I think, if there is no corresponding symbolic info, it
>> should fetch memory contents near $PC and display disassembled
>> instructions.
>
> Open a console window and type "list". If gdb does not return something,
> than Insight cannot, either. Insight will disassemble around $PC *if* gdb
> can find a function containing the PC (so that it knows the bounds of the
> function to disassemble). This is a gdb limitation that Insight has not
> worked around yet (and probably never will).

  Yes, I tried it out. Thanks!

Re: Problems with SrcWin

[Tommy Lin]

>>  (1) after Insight is connected to target, the Control menu and
>> button items are still disabled. They'll be enabled after some actions
>> are done (ex. execute 'stepi' from ConsoleWindow.)
>
> This is usually caused by inferior code not setting a fake ptid for the UI.
> This is a very common problem with non-native targets. You can check this by
> inspecting the function gdb_target_has_execution. Check out what  the remote
> target does in extended_remote_create_inferior_1 (it sets inferior_ptid to
> some made up non-NULL entity). The value does not matter -- that it is
> non-NULL is all that is required.

  I found my target's xxxx_create_inferior() isn't called before
gdb_target_has_execution(). Maybe it's not a good port. So, I fix this
issue by removing the inferior_ptid part from
gdb_target_has_execution() directly.
  It seems can be fixed in srcbar.itcl:do_connect by setting Control
$Menu and $Tool into 'normal'.

>>  (2) after Insight is connected to target, the SrcWin displays
>> nothing... (there is no 'file', 'load', and 'symbol-file' command
>> executed.) I think, if there is no corresponding symbolic info, it
>> should fetch memory contents near $PC and display disassembled
>> instructions.
>
> Open a console window and type "list". If gdb does not return something,
> than Insight cannot, either. Insight will disassemble around $PC *if* gdb
> can find a function containing the PC (so that it knows the bounds of the
> function to disassemble). This is a gdb limitation that Insight has not
> worked around yet (and probably never will).

  Yes, I tried it out. Thanks!

Re: Problems with SrcWin

[Tommy Lin]

>>  (1) after Insight is connected to target, the Control menu and
>> button items are still disabled. They'll be enabled after some actions
>> are done (ex. execute 'stepi' from ConsoleWindow.)
>
> This is usually caused by inferior code not setting a fake ptid for the UI.
> This is a very common problem with non-native targets. You can check this by
> inspecting the function gdb_target_has_execution. Check out what  the remote
> target does in extended_remote_create_inferior_1 (it sets inferior_ptid to
> some made up non-NULL entity). The value does not matter -- that it is
> non-NULL is all that is required.

  I found my target's xxxx_create_inferior() isn't called before
gdb_target_has_execution(). Maybe it's not a good port. So, I fix this
issue by removing the inferior_ptid part from
gdb_target_has_execution() directly.
  It seems can be fixed in srcbar.itcl:do_connect by setting Control
$Menu and $Tool into 'normal'.

>>  (2) after Insight is connected to target, the SrcWin displays
>> nothing... (there is no 'file', 'load', and 'symbol-file' command
>> executed.) I think, if there is no corresponding symbolic info, it
>> should fetch memory contents near $PC and display disassembled
>> instructions.
>
> Open a console window and type "list". If gdb does not return something,
> than Insight cannot, either. Insight will disassemble around $PC *if* gdb
> can find a function containing the PC (so that it knows the bounds of the
> function to disassemble). This is a gdb limitation that Insight has not
> worked around yet (and probably never will).

  Yes, I tried it out. Thanks!

Re: Problems with SrcWin

[Tommy Lin]

>>  (1) after Insight is connected to target, the Control menu and
>> button items are still disabled. They'll be enabled after some actions
>> are done (ex. execute 'stepi' from ConsoleWindow.)
>
> This is usually caused by inferior code not setting a fake ptid for the UI.
> This is a very common problem with non-native targets. You can check this by
> inspecting the function gdb_target_has_execution. Check out what  the remote
> target does in extended_remote_create_inferior_1 (it sets inferior_ptid to
> some made up non-NULL entity). The value does not matter -- that it is
> non-NULL is all that is required.

  I found my target's xxxx_create_inferior() isn't called before
gdb_target_has_execution(). Maybe it's not a good port. So, I fix this
issue by removing the inferior_ptid part from
gdb_target_has_execution() directly.
  It seems can be fixed in srcbar.itcl:do_connect by setting Control
$Menu and $Tool into 'normal'.

>>  (2) after Insight is connected to target, the SrcWin displays
>> nothing... (there is no 'file', 'load', and 'symbol-file' command
>> executed.) I think, if there is no corresponding symbolic info, it
>> should fetch memory contents near $PC and display disassembled
>> instructions.
>
> Open a console window and type "list". If gdb does not return something,
> than Insight cannot, either. Insight will disassemble around $PC *if* gdb
> can find a function containing the PC (so that it knows the bounds of the
> function to disassemble). This is a gdb limitation that Insight has not
> worked around yet (and probably never will).

  Yes, I tried it out. Thanks!

Re: Problems with SrcWin

[Tommy Lin]

>>  (1) after Insight is connected to target, the Control menu and
>> button items are still disabled. They'll be enabled after some actions
>> are done (ex. execute 'stepi' from ConsoleWindow.)
>
> This is usually caused by inferior code not setting a fake ptid for the UI.
> This is a very common problem with non-native targets. You can check this by
> inspecting the function gdb_target_has_execution. Check out what  the remote
> target does in extended_remote_create_inferior_1 (it sets inferior_ptid to
> some made up non-NULL entity). The value does not matter -- that it is
> non-NULL is all that is required.

  I found my target's xxxx_create_inferior() isn't called before
gdb_target_has_execution(). Maybe it's not a good port. So, I fix this
issue by removing the inferior_ptid part from
gdb_target_has_execution() directly.
  It seems can be fixed in srcbar.itcl:do_connect by setting Control
$Menu and $Tool into 'normal'.

>>  (2) after Insight is connected to target, the SrcWin displays
>> nothing... (there is no 'file', 'load', and 'symbol-file' command
>> executed.) I think, if there is no corresponding symbolic info, it
>> should fetch memory contents near $PC and display disassembled
>> instructions.
>
> Open a console window and type "list". If gdb does not return something,
> than Insight cannot, either. Insight will disassemble around $PC *if* gdb
> can find a function containing the PC (so that it knows the bounds of the
> function to disassemble). This is a gdb limitation that Insight has not
> worked around yet (and probably never will).

  Yes, I tried it out. Thanks!