controlling gdb via python's pexpect

[fpga]

Can someone please give me an example of how this is done.
I'm afraid the documentation is not helping me.
Thx

[George Russell]

fpga wrote:
> Can someone please give me an example of how this is done.
> I'm afraid the documentation is not helping me.
> Thx
>  
Hi,

I have had some success in controlling GDB (via MI) from python, using
the subprocess module and communicating with GDB via pipes.
(http://docs.python.org/library/subprocess.html)

The main issue so far is that some applications detect when they are
running with pipes instead of a terminal for output, and buffer program
output by block rather than line as on a terminal, leading to delays in
receiving program output issued via printf, for example, unless it is
followed by an explicit fflush call.

(I too would be interested in an example of using pexpect to control GDB)
Cheers,
George Russell

[fpga]

fpga wrote:

Can someone please give me an example of how this is done.
I'm afraid the documentation is not helping me.
Thx

Thx George
You've got further than I have!
I tried to use subprocess before trying pexpect but failed.
pexpect was suggested by experts on the python forum.
Unfortuntely they seem unable to correct my trivial example of opening a terminal and sending 'gdb' to it.
Therefore any working snippets would be welcome.
Looking at my repositories I see libdevel-gdb-perl.
I like python's orderliness but have noticed that when python fails perl usually delivers.

[George Russell]

fpga wrote:

> fpga wrote:
>  
>> Can someone please give me an example of how this is done.
>> I'm afraid the documentation is not helping me.
>> Thx
>>
>>    
>
> Thx George
> You've got further than I have!
> I tried to use subprocess before trying pexpect but failed.
> pexpect was suggested by experts on the python forum.
> Unfortuntely they appear incapable of correcting my trivial example of
> opening a terminal and sending the command 'gdb' to it.
> Any snippets most welcome.
> Looking at my repositories I see libdevel-gdb-perl.
> I like python's orderliness but have noticed that when python fails perl
> usually delivers.
>  

See

http://www.dalkescientific.com/writings/diary/archive/2005/04/12/wrapping_command_line_programs.html
http://www.dalkescientific.com/writings/diary/archive/2005/04/15/wrapping_command_line_programs_II.html
http://www.dalkescientific.com/writings/diary/archive/2005/04/17/wrapping_command_line_programs_III.html
http://www.dalkescientific.com/writings/diary/archive/2005/04/18/wrapping_command_line_programs_IV.html
http://www.dalkescientific.com/writings/diary/archive/2005/04/18/wrapping_command_line_programs_V.html

for an assortment of tips on control of external processes from Python.

Cheers,
George

[Greg Law-3]

fpga wrote:

>
>
> fpga wrote:
>> Can someone please give me an example of how this is done.
>> I'm afraid the documentation is not helping me.
>> Thx
>>
>
> George
> You've got further than I have!
> I tried to use subprocess but failed.

Yeah, pipes might appear to work at first if you're "lucky", but pretty
quickly you'll come unstuck due to problems with buffering.

Here's a trivial example for gdb with pexpect:

import pexpect, sys
child = pexpect.spawn( 'gdb a.out')
child.setlog( sys.stdout)
child.expect_exact ('(gdb) ')
child.send ('run\n')
child.expect_exact ('Program exited normally.')
child.expect_exact ('(gdb) ')
child.send ('quit\n')
child.expect( pexpect.EOF


Note that the pexpect.expect method takes a regular expression but
"expect_exact" doesn't.  If you're going to use the expect method be
sure to escape things like parentheses.

Greg
--
Greg Law, Undo Software                       http://undo-software.com/

[fpga]

That's given me something to look at.
Thank you to both of you!

fpga wrote:

Can someone please give me an example of how this is done.
I'm afraid the documentation is not helping me.
Thx

For anyone wanting to run Greg's helpful script - the syntax has changed slightly i.e. just replace the line
child.setlog( sys.stdout)
with
child.logfile = sys.stdout

[Gaius Mulley]

George Russell <george@...> writes:

> fpga wrote:
>> Can someone please give me an example of how this is done.
>> I'm afraid the documentation is not helping me.
>> Thx
>>  
> Hi,
>
> I have had some success in controlling GDB (via MI) from python, using
> the subprocess module and communicating with GDB via
> pipes. (http://docs.python.org/library/subprocess.html)
>
> The main issue so far is that some applications detect when they are
> running with pipes instead of a terminal for output, and buffer
> program output by block rather than line as on a terminal, leading to
> delays in receiving program output issued via printf, for example,
> unless it is followed by an explicit fflush call.
>
> (I too would be interested in an example of using pexpect to control GDB)
> Cheers,
> George Russell

Hi,

for what it is worth feel free to download
http://floppsie.comp.glam.ac.uk/download/python/twingdb.tar.gz

which was written to run two gdb's debugging a stage1 and stage2
generation of a compiler and single step until a line number differs.
It is written in Python and uses pexpect

The code is very alpha - but it worked well enough to find the bug I
was hunting.

regards,
Gaius

[Gaius Mulley]

Gaius Mulley <gaius@...> writes:

> for what it is worth feel free to download
> http://floppsie.comp.glam.ac.uk/download/python/twingdb.tar.gz

ahh correction, make that

  http://floppsie.comp.glam.ac.uk/download/python/tgdb-0.1.tar.gz

instead..

regards,
Gaius

[fpga]

Gaius Mulley wrote:

George Russell <[e-mail]> writes:

> fpga wrote:
>> Can someone please give me an example of how this is done.
>> I'm afraid the documentation is not helping me.
>> Thx
>>  
> Hi,
>
> I have had some success in controlling GDB (via MI) from python, using
> the subprocess module and communicating with GDB via
> pipes. (http://docs.python.org/library/subprocess.html)
>
> The main issue so far is that some applications detect when they are
> running with pipes instead of a terminal for output, and buffer
> program output by block rather than line as on a terminal, leading to
> delays in receiving program output issued via printf, for example,
> unless it is followed by an explicit fflush call.
>
> (I too would be interested in an example of using pexpect to control GDB)
> Cheers,
> George Russell

Hi,

for what it is worth feel free to download
http://floppsie.comp.glam.ac.uk/download/python/twingdb.tar.gz

which was written to run two gdb's debugging a stage1 and stage2
generation of a compiler and single step until a line number differs.
It is written in Python and uses pexpect

The code is very alpha - but it worked well enough to find the bug I
was hunting.

regards,
Gaius

Hi Gaius
I'd like to download this but get
The requested URL /download/python/twingdb.tar.gz was not found on this server.

OK Got it. Thx!

[Gaius Mulley]

fpga <mgbg25171@...> writes:

> Hi Gaius
> I'd like to download this but get
> The requested URL /download/python/twingdb.tar.gz was not found on this
> server.

Hi,

all restored now - although

http://floppsie.comp.glam.ac.uk/download/python/tgdb-0.1.tar.gz

is a cleaner package,

regards,
Gaius

[fpga]

I got that one too. Thx

Gaius Mulley wrote:

fpga <[e-mail]> writes:

> Hi Gaius
> I'd like to download this but get
> The requested URL /download/python/twingdb.tar.gz was not found on this
> server.

Hi,

all restored now - although

http://floppsie.comp.glam.ac.uk/download/python/tgdb-0.1.tar.gz

is a cleaner package,

regards,
Gaius