[RFA] Darwin/x86 port (v2 - part 0)

[RFA] Darwin/x86 port (v2 - part 0)

[Tristan Gingold]

Hi,

I think I made all the changes requested - and wasn't a large bit of  
work (mainly because darwin and
macosx have the same number of letters :-)

Please, review again!

Tristan.

2008-11-10  Tristan Gingold  <gingold@adacore.com>

	* configure.host: Add Darwin host.
	* configure.tgt: Add Darwin target.
	* defs.h (enum gdb_osabi): Add GDB_OSABI_DARWIN.
	* osabi.c (gdb_osabi_names): Add name for Darwin abi.
	* i386-darwin-nat.c: New file.
	* i386-darwin-tdep.c: New file.
	* i386-darwin-tdep.h: New file.
	* machoread.c: New file.
	* darwin-nat-info.c: New file.
	* darwin-nat.c: New file.
	* darwin-nat.h: New file.
	* darwin.defs: New file.
	* config/i386/darwin.mh: New file.

diff -c -r1.101 configure.host
*** configure.host	19 Jan 2008 15:03:50 -0000	1.101
--- configure.host	12 Nov 2008 10:01:24 -0000
***************
*** 62,67 ****
--- 62,69 ----

   case "${host}" in

+ *-*-darwin*)		gdb_host=darwin ;;
+
   alpha*-*-osf[3456789]*)	gdb_host=alpha-osf3 ;;
   alpha*-*-linux*)	gdb_host=alpha-linux ;;
   alpha*-*-freebsd* | alpha*-*-kfreebsd*-gnu)
Index: configure.tgt
===================================================================
RCS file: /cvs/src/src/gdb/configure.tgt,v
retrieving revision 1.204
diff -c -r1.204 configure.tgt
*** configure.tgt	2 Oct 2008 15:48:06 -0000	1.204
--- configure.tgt	12 Nov 2008 10:01:24 -0000
***************
*** 145,150 ****
--- 145,155 ----
   	gdb_target_obs="hppa-tdep.o"
   	;;

+ i[34567]86-*-darwin*)
+ 	# Target: Darwin/i386
+ 	gdb_target_obs="amd64-tdep.o i386-tdep.o i387-tdep.o \
+ 			i386-darwin-tdep.o"
+ 	;;
   i[34567]86-*-dicos*)
   	# Target: DICOS/i386
   	gdb_target_obs="i386-tdep.o i387-tdep.o \
Index: defs.h
===================================================================
RCS file: /cvs/src/src/gdb/defs.h,v
retrieving revision 1.241
diff -c -r1.241 defs.h
*** defs.h	5 Nov 2008 20:23:07 -0000	1.241
--- defs.h	12 Nov 2008 10:01:24 -0000
***************
*** 957,962 ****
--- 957,963 ----
     GDB_OSABI_CYGWIN,
     GDB_OSABI_AIX,
     GDB_OSABI_DICOS,
+   GDB_OSABI_DARWIN,

     GDB_OSABI_INVALID		/* keep this last */
   };
Index: osabi.c
===================================================================
RCS file: /cvs/src/src/gdb/osabi.c,v
retrieving revision 1.43
diff -c -r1.43 osabi.c
*** osabi.c	1 May 2008 23:09:14 -0000	1.43
--- osabi.c	12 Nov 2008 10:01:25 -0000
***************
*** 73,78 ****
--- 73,79 ----
     "Cygwin",
     "AIX",
     "DICOS",
+   "Darwin",

     "<invalid>"
   };

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Tristan Gingold]

Hi,

is there any remaining work to do before getting the approval ?

Tristan.

On Nov 12, 2008, at 11:02 AM, Tristan Gingold wrote:

> Hi,
>
> I think I made all the changes requested - and wasn't a large bit of  
> work (mainly because darwin and
> macosx have the same number of letters :-)
>
> Please, review again!
>
> Tristan.
>
> 2008-11-10  Tristan Gingold  <gingold@adacore.com>
>
> 	* configure.host: Add Darwin host.
> 	* configure.tgt: Add Darwin target.
> 	* defs.h (enum gdb_osabi): Add GDB_OSABI_DARWIN.
> 	* osabi.c (gdb_osabi_names): Add name for Darwin abi.
> 	* i386-darwin-nat.c: New file.
> 	* i386-darwin-tdep.c: New file.
> 	* i386-darwin-tdep.h: New file.
> 	* machoread.c: New file.
> 	* darwin-nat-info.c: New file.
> 	* darwin-nat.c: New file.
> 	* darwin-nat.h: New file.
> 	* darwin.defs: New file.
> 	* config/i386/darwin.mh: New file.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Mark Kettenis]

> From: Tristan Gingold <gingold@adacore.com>
> Date: Thu, 13 Nov 2008 13:51:52 +0100
> 
> Hi,
> 
> is there any remaining work to do before getting the approval ?

Hmm, it seems something (you?) keeps spamming the lists with v2 of the
diff whereas I thought you already had a v3 diff.  I'm afraid I
completely lost track of what is what.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Tristan Gingold]

On Nov 13, 2008, at 3:52 PM, Mark Kettenis wrote:
>> is there any remaining work to do before getting the approval ?
>
> Hmm, it seems something (you?) keeps spamming the lists with v2 of the
> diff whereas I thought you already had a v3 diff.  I'm afraid I
> completely lost track of what is what.

Although there might be something wrong here, I don't see these mails  
in the archives.
Can you sent me privately these spams ?  We will try to fix that.

Tristan.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Stan Shebs]

Tristan Gingold wrote:
>
> On Nov 13, 2008, at 3:52 PM, Mark Kettenis wrote:
>>> is there any remaining work to do before getting the approval ?
>>
>> Hmm, it seems something (you?) keeps spamming the lists with v2 of the
>> diff whereas I thought you already had a v3 diff.  I'm afraid I
>> completely lost track of what is what.
>
> Although there might be something wrong here, I don't see these mails 
> in the archives.
> Can you sent me privately these spams ?  We will try to fix that.
I think it's just that you updated some parts to v3, while others are 
still at v2. When I sorted my mailbox by sender, it became much clearer. 
There's nothing wrong, it's just that you've been so responsive that 
there's not the usual multi-week gap between one rev of the patch and 
the next. :-)

Anyway, I'm the most logical reviewer for the bulk of native support, 
I'll give you some feedback in the next few days. In the meantime, 
please run the testsuite, let's see what the basic numbers are looking like.

Stan

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Tristan Gingold]

On Nov 13, 2008, at 7:03 PM, Stan Shebs wrote:
> Anyway, I'm the most logical reviewer for the bulk of native  
> support, I'll give you some feedback in the next few days. In the  
> meantime, please run the testsuite, let's see what the basic numbers  
> are looking like.

Thanks.  I plan to post a new version of the patches that addresses  
Eli's review.

Here is the first testsuite result:

		=== gdb Summary ===

# of expected passes		4299
# of unexpected failures	3107
# of expected failures		33
# of known failures		11
# of unresolved testcases	152
# of untested testcases		31
# of unsupported tests		61


(Quiet slow as there are many timeouts).

Tristan.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Stan Shebs]

Tristan Gingold wrote:
>
> On Nov 13, 2008, at 7:03 PM, Stan Shebs wrote:
>> Anyway, I'm the most logical reviewer for the bulk of native support, 
>> I'll give you some feedback in the next few days. In the meantime, 
>> please run the testsuite, let's see what the basic numbers are 
>> looking like.
>
> Thanks.  I plan to post a new version of the patches that addresses 
> Eli's review.
>
> Here is the first testsuite result:
>
>         === gdb Summary ===
>
> # of expected passes        4299
> # of unexpected failures    3107
Hmmm, that's kind of high, even for an initial port. Is it mainly that 
shared library bits are missing, or something else?

My natural inclination is to accept the code into the trunk after review 
- I suspect that the next round of changes will be more invasive into 
the rest of GDB, and it would be easier to consider each of those 
separately from the basic port. But, I have a track record of being too 
optimistic on this strategy, ahem. :-) What do other people think?

Stan

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Eli Zaretskii]

> Date: Fri, 14 Nov 2008 07:06:00 -0800
> From: Stan Shebs <stan@codesourcery.com>
> CC: Stan Shebs <stan@codesourcery.com>,   Mark Kettenis <mark.kettenis@xs4all.nl>,  gdb-patches@sourceware.org
> 
> > # of expected passes        4299
> > # of unexpected failures    3107
> Hmmm, that's kind of high, even for an initial port. Is it mainly that 
> shared library bits are missing, or something else?
> 
> My natural inclination is to accept the code into the trunk after review 
> - I suspect that the next round of changes will be more invasive into 
> the rest of GDB, and it would be easier to consider each of those 
> separately from the basic port. But, I have a track record of being too 
> optimistic on this strategy, ahem. :-) What do other people think?

I'd like to know whether there are any large sets of failed tests that
have something in common (a.k.a. unsupported GDB features), before I
make up my mind.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Tristan Gingold]

On Nov 14, 2008, at 4:31 PM, Eli Zaretskii wrote:
> I'd like to know whether there are any large sets of failed tests that
> have something in common (a.k.a. unsupported GDB features), before I
> make up my mind.

Sure.
I was using a wrong gdb (was work in progress).  New results are much  
better:

		=== gdb Summary ===

# of expected passes		10785
# of unexpected failures	653
# of expected failures		33
# of known failures		117
# of unresolved testcases	5
# of untested testcases		30
# of unsupported tests		59

There are a few obvious groups:
* unwinding through signals handlers doesn't work.
* shared libraries are not supported.  There are a few consequences:  
break printf doesn't work too.
* no core support
* no thread support
* call-sc doesn't work.

Tristan.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Joel Brobecker]

(trying to take my AdaCore hat off)

> My natural inclination is to accept the code into the trunk after review 
> - I suspect that the next round of changes will be more invasive into 
> the rest of GDB, and it would be easier to consider each of those 
> separately from the basic port. But, I have a track record of being too 
> optimistic on this strategy, ahem. :-) What do other people think?

I think that's a good approach. In this case, it is even more attractive
as the changes have remained uninvasive so far, so it would be easy to
undo them all if we decided to do so. All in all, I think we have a lot
to gain, and little to lose.

-- 
Joel

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Mark Kettenis]

> Date: Fri, 14 Nov 2008 07:43:08 -0800
> From: Joel Brobecker <brobecker@adacore.com>
> 
> (trying to take my AdaCore hat off)
> 
> > My natural inclination is to accept the code into the trunk after review 
> > - I suspect that the next round of changes will be more invasive into 
> > the rest of GDB, and it would be easier to consider each of those 
> > separately from the basic port. But, I have a track record of being too 
> > optimistic on this strategy, ahem. :-) What do other people think?
> 
> I think that's a good approach. In this case, it is even more attractive
> as the changes have remained uninvasive so far, so it would be easy to
> undo them all if we decided to do so. All in all, I think we have a lot
> to gain, and little to lose.

Just to be clear.  I'm happy with the state the i386-specefic bits are
in, so if Stan's happy with the generic Darwin bits I have no
objections.  Best to have this in the tree such that it is easier for
other people to work on.

Re: [RFA] Darwin/x86 port (v2 - part 0)

[Eli Zaretskii]

> From: Tristan Gingold <gingold@adacore.com>
> Date: Wed, 12 Nov 2008 11:02:14 +0100
> 
> I think I made all the changes requested - and wasn't a large bit of  
> work (mainly because darwin and
> macosx have the same number of letters :-)
> 
> Please, review again!

Thanks.

I think we will need a NEWS entry for this addition.