MDI: testsuite support

MDI: testsuite support

[Maciej W. Rozycki]

Hello,

 This is testsuite support for MDI targets.  It makes them recognised as 
remote and provides the necessary adjustments for the MI part.

2008-02-19  David Ung  <davidu@mips.com>
            Nigel Stephens  <nigel@mips.com>
            Maciej W. Rozycki  <macro@mips.com>

	* config/sde-mdi.exp: New file.
	* lib/mi-support.exp (mi_gdb_target_cmd): Recognise MDI signon
	message.
	(mi_gdb_load): Add "mdi" gdb_protocol handler.

  Maciej

mdi-12186.diff
Index: binutils-quilt/src/gdb/testsuite/lib/mi-support.exp
===================================================================
--- binutils-quilt.orig/src/gdb/testsuite/lib/mi-support.exp	2008-01-11 11:56:29.000000000 +0000
+++ binutils-quilt/src/gdb/testsuite/lib/mi-support.exp	2008-01-11 12:04:51.000000000 +0000
@@ -339,6 +339,10 @@
 		verbose "Set target to $targetname";
 		return 0;
 	    }
+	    -re "Remote MDI debugging.*$mi_gdb_prompt$" {
+		verbose "Set target to $targetname";
+		return 0;
+	    }
 	    -re "Remote debugging using .*$serialport_re.*$mi_gdb_prompt$" {
 		verbose "Set target to $targetname";
 		return 0;
@@ -505,6 +509,27 @@
 		return -1
 	    }
 	}
+    } elseif { [target_info gdb_protocol] == "mdi" } {
+	# MDI targets
+	send_gdb "47-target-select mdi [target_info gdb_serial]\n"
+	#send_gdb "target mdi [target_info gdb_serial]\n"
+	gdb_expect 60 {
+	    -re "47\\^connected.*$mi_gdb_prompt$" {
+	    }
+	    timeout {
+		perror "Unable to select mdi target"
+		return -1
+	    }
+	}
+	send_gdb "48-target-download\n"
+	gdb_expect 10 {
+	    -re "48\\^done.*$mi_gdb_prompt$" {
+	    }
+	    timeout {
+		perror "Unable to download to mdi target"
+		return -1
+	    }
+	}
     } elseif { [target_info gdb_protocol] == "remote" } {
 	# remote targets
 	if { [mi_gdb_target_cmd "remote" [target_info netport]] != 0 } {
Index: binutils-quilt/src/gdb/testsuite/config/sde-mdi.exp
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ binutils-quilt/src/gdb/testsuite/config/sde-mdi.exp	2008-01-11 12:08:40.000000000 +0000
@@ -0,0 +1,16 @@
+#   Copyright 1993, 1997, 2008 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+load_lib ../config/monitor.exp

Re: MDI: testsuite support

[Daniel Jacobowitz]

On Tue, Feb 19, 2008 at 04:26:41PM +0000, Maciej W. Rozycki wrote:
> Hello,
> 
>  This is testsuite support for MDI targets.  It makes them recognised as 
> remote and provides the necessary adjustments for the MI part.
> 
> 2008-02-19  David Ung  <davidu@mips.com>
>             Nigel Stephens  <nigel@mips.com>
>             Maciej W. Rozycki  <macro@mips.com>
> 
> 	* config/sde-mdi.exp: New file.
> 	* lib/mi-support.exp (mi_gdb_target_cmd): Recognise MDI signon
> 	message.
> 	(mi_gdb_load): Add "mdi" gdb_protocol handler.

Let's hold off on this patch until the rest of the MDI target is
posted, please.  I appreciate your separating it into more manageable
pieces but at the same time they are hard to consider in isolation
and I want to be sure the whole thing goes in.

I don't see anything wrong with the patch though, so it will be OK.

-- 
Daniel Jacobowitz
CodeSourcery

Re: MDI: testsuite support

[Maciej W. Rozycki]

On Tue, 26 Feb 2008, Daniel Jacobowitz wrote:

> Let's hold off on this patch until the rest of the MDI target is
> posted, please.  I appreciate your separating it into more manageable
> pieces but at the same time they are hard to consider in isolation
> and I want to be sure the whole thing goes in.

 Well, I had promised to send the testsuite bits shortly, so I did it. :-)

 Which "rest of the MDI target" are you specifically referring to?  The 
only piece I have got left is support for some obscure "set mdi" and "show 
mdi" commands that should not be subcommands of "set" and "show" in the 
first place as they operate on the target (which is against the principle 
of the commands) and the latter also breaks "show" as used to print all 
the settings.

 However if you'd like to see this bit on the list as-is, then I am fine 
with that, but I do not consider it ready for submission.  My current 
feeling is they should be reimplemented as separate commands -- while the 
affected "show mdi" commands might fit a command like "info mdi", there is 
apparently no corresponding command for "set mdi", so I suppose a pair of 
commands like "mdi set" and "mdi show" might sound better.

 Note I mean these commands are really obscure and are used things like to 
peek or poke at TLB entries and cache lines "through a back door", so I 
gather whoever is skilled enough to dare would know what they are doing 
and get used to whatever the commands are called.  In fact hardly anybody 
uses them as proved by an embarrasing bug I fixed a while ago somewhere in 
this area. ;-)  They are occasionally useful for stuff like debugging 
system bringup issues, weird cache corruption problems and suchlike.

  Maciej

Re: MDI: testsuite support

[Daniel Jacobowitz]

On Tue, Feb 26, 2008 at 04:38:51PM +0000, Maciej W. Rozycki wrote:
>  Which "rest of the MDI target" are you specifically referring to?

I meant the actual target, that is, the bits this testsuite support
lets you test.  You haven't posted that, have you?

>  Note I mean these commands are really obscure and are used things like to 
> peek or poke at TLB entries and cache lines "through a back door", so I 
> gather whoever is skilled enough to dare would know what they are doing 
> and get used to whatever the commands are called.  In fact hardly anybody 
> uses them as proved by an embarrasing bug I fixed a while ago somewhere in 
> this area. ;-)  They are occasionally useful for stuff like debugging 
> system bringup issues, weird cache corruption problems and suchlike.

This is the sort of thing that makes me wish all our targets used
"target remote".  MDI isn't the only target interface that offers this
sort of thing, and it would be nice to have the enforced abstraction
barrier; that would make it easier to support the same thing,
consistently, on other targets.  Anyway, an example for MDI will
be almost as useful in practice.

-- 
Daniel Jacobowitz
CodeSourcery

Re: MDI: testsuite support

[Maciej W. Rozycki]

On Tue, 26 Feb 2008, Daniel Jacobowitz wrote:

> >  Which "rest of the MDI target" are you specifically referring to?
> 
> I meant the actual target, that is, the bits this testsuite support
> lets you test.  You haven't posted that, have you?

 Hmm, I sent the bits back on Feb 15th -- I have not got a bounce, so I 
assumed it must have been successful.  OK -- I'll resend.  Is there a size 
limit on this list?  The patch is over 300kB, so it may have been disliked 
because of the size (but then I would expect a bounce).

  Maciej

Re: MDI: testsuite support

[Daniel Jacobowitz]

On Tue, Feb 26, 2008 at 04:58:01PM +0000, Maciej W. Rozycki wrote:
> On Tue, 26 Feb 2008, Daniel Jacobowitz wrote:
> 
> > >  Which "rest of the MDI target" are you specifically referring to?
> > 
> > I meant the actual target, that is, the bits this testsuite support
> > lets you test.  You haven't posted that, have you?
> 
>  Hmm, I sent the bits back on Feb 15th -- I have not got a bounce, so I 
> assumed it must have been successful.  OK -- I'll resend.  Is there a size 
> limit on this list?  The patch is over 300kB, so it may have been disliked 
> because of the size (but then I would expect a bounce).

I don't know if there's a size limit; you'd have to ask overseers.
It definitely did not reach the list.

-- 
Daniel Jacobowitz
CodeSourcery

Re: MDI: testsuite support

[Maciej W. Rozycki]

On Tue, 26 Feb 2008, Daniel Jacobowitz wrote:

> >  Hmm, I sent the bits back on Feb 15th -- I have not got a bounce, so I 
> > assumed it must have been successful.  OK -- I'll resend.  Is there a size 
> > limit on this list?  The patch is over 300kB, so it may have been disliked 
> > because of the size (but then I would expect a bounce).
> 
> I don't know if there's a size limit; you'd have to ask overseers.
> It definitely did not reach the list.

 Well, I have sent it again, to the list and to my linux-mips.org address 
and only the latter has so far arrived.  No bounce either.  It looks 
broken.  I'll inquire.

  Maciej

Re: MDI: testsuite support

[Daniel Jacobowitz]

On Tue, Feb 26, 2008 at 05:09:35PM +0000, Maciej W. Rozycki wrote:
>  Well, I have sent it again, to the list and to my linux-mips.org address 
> and only the latter has so far arrived.  No bounce either.  It looks 
> broken.  I'll inquire.

IIRC the list sanity filters will generate bounces, but the spam
filters won't.  It must have found something to dislike :-(
If you haven't already, try subscribing the address sending it
to the sourceware global-allow list.

-- 
Daniel Jacobowitz
CodeSourcery

Re: MDI: testsuite support

[Maciej W. Rozycki]

On Tue, 26 Feb 2008, Daniel Jacobowitz wrote:

> IIRC the list sanity filters will generate bounces, but the spam
> filters won't.  It must have found something to dislike :-(
> If you haven't already, try subscribing the address sending it
> to the sourceware global-allow list.

 Well actually the spam filter does send bounces as I got one in January 
and following that subscribed to the global-allow list.  Furthermore I 
passed the failing submission to the owner for investigation so that the 
spam filter may get improved, but unfortunately my request has not 
attracted anybody it would seem.

 I suppose the filter for at least the -patches mailing lists could get 
enhanced so that it recognised mails containing a context or unified diff 
and let them pass -- I have yet to see a spam letter with a diff embedded 
and I would think a spammer creative enough to actually include one in 
postings to mailing lists expecting them should deserve some form of 
appreciation. ;-)

  Maciej

Re: MDI: testsuite support

[Maciej W. Rozycki]

On Tue, 26 Feb 2008, Daniel Jacobowitz wrote:

> >  Note I mean these commands are really obscure and are used things like to 
> > peek or poke at TLB entries and cache lines "through a back door", so I 
> > gather whoever is skilled enough to dare would know what they are doing 
> > and get used to whatever the commands are called.  In fact hardly anybody 
> > uses them as proved by an embarrasing bug I fixed a while ago somewhere in 
> > this area. ;-)  They are occasionally useful for stuff like debugging 
> > system bringup issues, weird cache corruption problems and suchlike.
> 
> This is the sort of thing that makes me wish all our targets used
> "target remote".  MDI isn't the only target interface that offers this
> sort of thing, and it would be nice to have the enforced abstraction
> barrier; that would make it easier to support the same thing,
> consistently, on other targets.  Anyway, an example for MDI will
> be almost as useful in practice.

 I am certainly not against it, but it would take a while.  Porting all 
the stuff the MDI target supports to the remote framework is considerable 
amount of work and it would be nice to have something working till then, 
for a reference e.g. to compare regression test results against if nothing 
else.

  Maciej

Re: MDI: testsuite support

[Daniel Jacobowitz]

On Wed, Feb 27, 2008 at 03:44:35PM +0000, Maciej W. Rozycki wrote:
>  I am certainly not against it, but it would take a while.  Porting all 
> the stuff the MDI target supports to the remote framework is considerable 
> amount of work and it would be nice to have something working till then, 
> for a reference e.g. to compare regression test results against if nothing 
> else.

Yes, I agree.


-- 
Daniel Jacobowitz
CodeSourcery