tromey/python non-trivial merge notification

[Phil Muldoon]

Jan asked me to investigate a non-trivial merge issue today with the
tromey/python branch.

I merged tromey/python and fixed the trivial merge conflicts first.
The commit reference for that is:

aed684a9a6f3dab80262ca21d6c728e99526b37f

The non trivial issue was as follows:

Upstream GDB added a patch that caused some conflict in the gdb.cli()
Python command. The commit was:

d03f62382e0502a42020e6379dbe3c9252813c65

2013-09-06 Andrew Burgess <aburgess-dY08KVG/lbpWk0Htik3J/***@public.gmane.org>

* cli/cli-interp.c (_initialize_cli_interp): Add a
command_loop_proc to interp_procs.
* event-top.c (cli_command_loop): Change signature to match
interp_command_loop_ftype.
* event-top.h (cli_command_loop): Same.
* interps.c (interp_new): Require every interpreter to have a
command_loop_proc.
(current_interp_command_loop): Just call the command_loop_proc on
the current interpreter.
* tui/tui-interp.c (_initialize_tui_interp): Add a
command_loop_proc to interp_procs.

The first fix I thought of was just to pass cli_command_loop a NULL
(as cli_command_loop does not actually do anything with the data).
Jan rightly pointed out that in the future this may cause issues as,
really, this code requires a command_loop_proc.

Then we thought about using current_interp_command_loop instead. This
would work fine, but it would be MI incompatible. I could not think
of a reason why gdb.cli() would be invoked from MI. (It does not make
a whole lot of sense to do so, but someone might prove me wrong ;) )

I decided to bar invocation of this command in MI, and just use
current_interp_command_loop instead. The commit for this fix is:

269d883cd0c2fd8a51163e4472963b1b5a9a576e

2013-10-14 Phil Muldoon <pmuldoon-H+wXaHxf7aLQT0dZR+***@public.gmane.org>

* python/python.c (gdbpy_cli): Use current_interp_command_loop and
bar invocation from MI.

As this is a non-trivial change I decide to notify archer list about
the changes.

Cheers,

Phil

[Tom Tromey]

Phil> Then we thought about using current_interp_command_loop instead. This
Phil> would work fine, but it would be MI incompatible. I could not think
Phil> of a reason why gdb.cli() would be invoked from MI. (It does not make
Phil> a whole lot of sense to do so, but someone might prove me wrong ;) )

Yeah.
The intended use for this feature was to be able to start gdb without a
CLI, using "gdb -P"; then later give the user a CLI if "something happened".
There's no reason to do this kind of thing in MI.

Tom