gdb supported on powerpc-apple-darwin ?

gdb supported on powerpc-apple-darwin ?

[Mario Emmenlauer]

Hi,


it seems a rather trivial question, however I could not easily find
the answer: is 'plain' FSF gdb supported on Mac OS X (ppc darwin)?

Simply downloading the gdb-6.8 release, configure says:
*** This configuration is not supported in the following subdirectories:
      bfd opcodes gdb sim
     (Any other directories should still work fine.)

without any other complaints in config.log. Subsequent calls to
make compile libreadline and stuff, but not gdb. Inspecting 
configure.ac, it claims that for the 'powerpc-*-darwin*' target the
subdirectory gdb would indeed be unsupported. The same goes for
newest CVS sources.

I could not find a 'supported plattforms' listing, neither on the
website, nor in the README / INSTALL files. However in a previous
topic on this list ('error during configure/make sequence on GDB-CVS
sources at Mac OS X'), the differences between FSF gdb-6.8 and Apples
patched gdb on Mac OS X where discussed. This led me to the (wrong?)
assumption that FSF gdb would plainly build on darwin?

Am I doing something wrong here?


Cheers,

    Mario


BTW: Apple developers: is the current (and upcoming) Apple gdb
still based on gdb's 2005 code? I could not find newer commits in
the Changelog of gdb-908, are they just undocumented?

Which version of FSF gdb does gdb-908 compare to?

Re: gdb supported on powerpc-apple-darwin ?

[Daniel Jacobowitz]

On Sat, Aug 16, 2008 at 04:44:57PM +0200, Mario Emmenlauer wrote:
>
> Hi,
>
>
> it seems a rather trivial question, however I could not easily find
> the answer: is 'plain' FSF gdb supported on Mac OS X (ppc darwin)?

No.  That's what this means:

> *** This configuration is not supported in the following subdirectories:
>      bfd opcodes gdb sim

-- 
Daniel Jacobowitz
CodeSourcery

Re: gdb supported on powerpc-apple-darwin ?

[Stan Shebs]

Mario Emmenlauer wrote:
> it seems a rather trivial question, however I could not easily find
> the answer: is 'plain' FSF gdb supported on Mac OS X (ppc darwin)?
No. The platform-specific bits necessary to make GDB work as a native 
debugger on Darwin are rather complicated, and have never been gotten to 
work in FSF GDB. (I got partway there once, tree is still on my disk 
somewhere.) You should be able to build cross debuggers targeting 
embedded systems though, I do that regularly.

Stan

Re: gdb supported on powerpc-apple-darwin ?

[Dr. Rolf Jansen]

Am 16.08.2008 um 11:44 schrieb Mario Emmenlauer:

> BTW: Apple developers: is the current (and upcoming) Apple gdb
> still based on gdb's 2005 code?

Apple gdb has been forked from GNU gdb 6.3.50-20050815. And the  
current version shipped with Xcode 3.1 is (Apple version gdb-960). The  
lates version that can be downloaded at Apples OS site is gdb-961.

The term "still based" in your question implies too much. Apple forked  
GDB in 2005 and since then it became a different product, there a so  
many (well documented) additions since then, that it cannot simply be  
based on a newer FSF gdb. In fact that would mean to fork it again.

> I could not find newer commits in
> the Changelog of gdb-908, are they just undocumented?

Look at:

<http://www.opensource.apple.com/darwinsource/iPhoneSDKBeta8/gdb-961/>

> Which version of FSF gdb does gdb-908 compare to?

Compare in which respect? Which features are you looking for?

Since August 2005 both gdb forks experienced many additions. While  
Apples focused on their platform on took gdb in many areas on a higher  
lever of sophistication. FSF gdb required a lot of work adding support  
to new platforms.

Best regards

Rolf Jansen

Re: gdb supported on powerpc-apple-darwin ?

[Mario Emmenlauer]

[-- Attachment #1: Type: text/plain, Size: 732 bytes --]


Hi,

Daniel Jacobowitz wrote:
> On Sat, Aug 16, 2008 at 04:44:57PM +0200, Mario Emmenlauer wrote:
>> it seems a rather trivial question, however I could not easily find
>> the answer: is 'plain' FSF gdb supported on Mac OS X (ppc darwin)?
> 
> No.  That's what this means:
> 
>> *** This configuration is not supported in the following subdirectories:
>>      bfd opcodes gdb sim

Fair enough :-)

BTW: the error message appears just after invoking configure, and is
not re-displayed later. This is why I missed it the first time. When
the 'gdb' directory is not supported, it might be sensible to show a
big warning at the configure results?

A start might be something like the below patch to configure.ac.


Cheers,

    Mario


[-- Attachment #2: configure.ac.patch --]
[-- Type: text/plain, Size: 873 bytes --]

--- gdb-6.8/configure.ac	2008-03-27 19:27:19.000000000 +0100
+++ /tmp/configure.ac	2008-08-16 17:50:22.000000000 +0200
@@ -1632,8 +1632,14 @@
   # Okay, some application will build, or we don't care to check.  Still
   # notify of subdirs not getting built.
   echo "*** This configuration is not supported in the following subdirectories:" 1>&2
-  echo "    ${notsupp}" 1>&2
-  echo "    (Any other directories should still work fine.)" 1>&2
+  echo "     ${notsupp}" 1>&2
+  if echo " ${notsupp} " | grep " gdb " >/dev/null 2>&1; then
+    echo "    which means gdb will not be built on this platform!" 1>&2
+    echo "    Continuing, though possible, is most likely not what you want." 1>&2
+    echo "    (However other directories should still work fine.)" 1>&2
+  else
+    echo "    (Any other directories should still work fine.)" 1>&2
+  fi
 fi
 
 case "$host" in

Re: gdb supported on powerpc-apple-darwin ?

[Mario Emmenlauer]

Hi Rolf,

Dr. Rolf Jansen wrote:
> Am 16.08.2008 um 11:44 schrieb Mario Emmenlauer:
> 
>> BTW: Apple developers: is the current (and upcoming) Apple gdb
>> still based on gdb's 2005 code?
> 
> Apple gdb has been forked from GNU gdb 6.3.50-20050815. And the current 
> version shipped with Xcode 3.1 is (Apple version gdb-960). The lates 
> version that can be downloaded at Apples OS site is gdb-961.

Is 961 the commit- (i.e. change-)number? Or does it correspond to
another way of labeling releases?


> The term "still based" in your question implies too much. Apple forked 
> GDB in 2005 and since then it became a different product, there a so 
> many (well documented) additions since then, that it cannot simply be 
> based on a newer FSF gdb. In fact that would mean to fork it again.

I see. But is there an information exchange going on?


>> I could not find newer commits in
>> the Changelog of gdb-908, are they just undocumented?
> 
> Look at:
> 
> <http://www.opensource.apple.com/darwinsource/iPhoneSDKBeta8/gdb-961/>

I was indeed missing 'ChangeLog-Apple', because it does not exist in
the root directory of gdb-908 (which was the newest release I could
find on the Mac developer pages). I'm sorry for the confusion, it
seems there is some active development going on.


>> Which version of FSF gdb does gdb-908 compare to?
> 
> Compare in which respect? Which features are you looking for?

The question was based on the assumption that there is a common patch
base on both forks. If the information exchange is little, it does not
make sense to try to compare.

I'm not actually missing a specific feature. I have rather experienced
bad performance and integration with my favorite GUI, using the mi
(actually mi2, I believe) interface. I read that the Apple fork is not
completely compatible in respect to this interface, for the sake of
better Xcode integration, which made me hope the FSF gdb would work
better.

Is there a compatibility mode for the Apple fork?

Cheers,

    Mario

Re: gdb supported on powerpc-apple-darwin ?

[Stan Shebs]

Dr. Rolf Jansen wrote:
> Am 16.08.2008 um 11:44 schrieb Mario Emmenlauer:
>
>> BTW: Apple developers: is the current (and upcoming) Apple gdb
>> still based on gdb's 2005 code?
>
> Apple gdb has been forked from GNU gdb 6.3.50-20050815. And the 
> current version shipped with Xcode 3.1 is (Apple version gdb-960). The 
> lates version that can be downloaded at Apples OS site is gdb-961.
>
> The term "still based" in your question implies too much. Apple forked 
> GDB in 2005 and since then it became a different product, there a so 
> many (well documented) additions since then, that it cannot simply be 
> based on a newer FSF gdb. In fact that would mean to fork it again.
"Forked GDB in 2005" is not quite accurate. Apple has had its own 
version of GDB for many many years (going back to NeXT days I believe), 
and has merged FSF GDB code into it a number of times, the most recent 
being the one I did in 2005. Apple has also contributed code to FSF GDB 
a number of times, both directly and indirectly. A new merge at this 
point would be rather painful (and now there is a GPLv3 issue to boot), 
and so it's been getting put off. Basic Darwin support in FSF GDB would 
make it a lot easier though, one of the reasons I made a pass at it.

Stan

Re: gdb supported on powerpc-apple-darwin ?

[Stan Shebs]

Mario Emmenlauer wrote:
>
> Hi,
>
> Daniel Jacobowitz wrote:
>> On Sat, Aug 16, 2008 at 04:44:57PM +0200, Mario Emmenlauer wrote:
>>> it seems a rather trivial question, however I could not easily find
>>> the answer: is 'plain' FSF gdb supported on Mac OS X (ppc darwin)?
>>
>> No.  That's what this means:
>>
>>> *** This configuration is not supported in the following 
>>> subdirectories:
>>>      bfd opcodes gdb sim
>
> Fair enough :-)
>
> BTW: the error message appears just after invoking configure, and is
> not re-displayed later. This is why I missed it the first time. When
> the 'gdb' directory is not supported, it might be sensible to show a
> big warning at the configure results?
The same configure script is used for binutils, ld, and many other 
things, so it's not going to work to single out gdb in the warning 
message. In theory one could customize the configure script to say 
different things depending on whether it is part of a gdb release, 
binutils release, etc, but that would be quite a bit more work for a 
situation that is nearly unique to Darwin.

Stan

Re: gdb supported on powerpc-apple-darwin ?

[Mario Emmenlauer]

Hi Stan,

Stan Shebs wrote:
> Dr. Rolf Jansen wrote:
>> Am 16.08.2008 um 11:44 schrieb Mario Emmenlauer:
>>
>>> BTW: Apple developers: is the current (and upcoming) Apple gdb
>>> still based on gdb's 2005 code?
>>
>> Apple gdb has been forked from GNU gdb 6.3.50-20050815. And the 
>> current version shipped with Xcode 3.1 is (Apple version gdb-960). The 
>> lates version that can be downloaded at Apples OS site is gdb-961.
>>
>> The term "still based" in your question implies too much. Apple forked 
>> GDB in 2005 and since then it became a different product, there a so 
>> many (well documented) additions since then, that it cannot simply be 
>> based on a newer FSF gdb. In fact that would mean to fork it again.
> "Forked GDB in 2005" is not quite accurate. Apple has had its own 
> version of GDB for many many years (going back to NeXT days I believe), 
> and has merged FSF GDB code into it a number of times, the most recent 
> being the one I did in 2005. Apple has also contributed code to FSF GDB 
> a number of times, both directly and indirectly. A new merge at this 
> point would be rather painful (and now there is a GPLv3 issue to boot), 
> and so it's been getting put off. Basic Darwin support in FSF GDB would 
> make it a lot easier though, one of the reasons I made a pass at it.

I would assume that basic Darwin support in FSF GDB would be a good
starting point. You mentioned you already put some work into it, any
chances that you will continue? How far have you gotten?

It seems very sadening that FSF gdb won't ever incorporate Apples
changes (or another type of darwin support). Of course one might know
who to point the finger at, but in the end that won't help.
If Apples fork is GPL, would it not be possible to back-merge the
darwin support, leaving out the other changes that are incompatible
to the current interface(s)?

Cheers,

    Mario

Re: gdb supported on powerpc-apple-darwin ?

[Mario Emmenlauer]

Hi,

Dr. Rolf Jansen wrote:
>>>> Which version of FSF gdb does gdb-908 compare to?
>>> Compare in which respect? Which features are you looking for?
>>
>> The question was based on the assumption that there is a common patch
>> base on both forks. If the information exchange is little, it does not
>> make sense to try to compare.
>>
>> I'm not actually missing a specific feature. I have rather experienced
>> bad performance and integration with my favorite GUI, using the mi
>> (actually mi2, I believe) interface. I read that the Apple fork is not
>> completely compatible in respect to this interface, for the sake of
>> better Xcode integration, which made me hope the FSF gdb would work
>> better.
> 
> Well you have three options:
> 
> 1. if you favourite GUI is open source, then you can try
>    to accomodate for the Apples mi dialect there.

Sadly, it isn't. But I called on the developers, lets see what happens.


> OR
> 
> 2. you can add in the FSF mi into Apples GDB. This
>    sounds more complicated than it is. I did this
>    the other way around. I produced an mi for Xcode
>    called miX and mixed it into the current FSF
>    GDB. With that I can do cross platform debugging
>    from within Xcode.
> 
>    <http://code.google.com/p/xcxdb/>

I saw your post in this regard, quite nice!
It seems a good starting point if I should decide to contribute
to gdb.


> To make a reasonable decision for 1 | 2 | 3, you might first want to 
> figure out which MI commands are mis-understood by Apples-gdb or which 
> MI results are mis-understood by your favourite GUI.
> 
> For this you might want to let your GUI call gdb by the way of a proxy 
> shell script. This shell script contains the following lines:
> 
> #!/bin/sh
> tee ~/Desktop/gdb-session.in | /usr/bin/gdb --interp=mi2 | tee 
> ~/Desktop/gdb-session.out

Thanks, this is indeed the first thing to do. I will check and see whats
going on between the GUI and gdb, and which commands exactly confuse
the former. I already know that these are few, so with some luck I
can even avoid them.

Cheers,

    Mario

Re: gdb supported on powerpc-apple-darwin ?

[Stan Shebs]

Mario Emmenlauer wrote:
> I would assume that basic Darwin support in FSF GDB would be a good
> starting point. You mentioned you already put some work into it, any
> chances that you will continue? How far have you gotten?
I got it to where it would up seize up waiting for signals and such from 
the inferior. :-) Just that much involved bringing over several thousand 
lines of Darwin-specific code. Dunno when/if I'll get back to it, 
CodeSourcery is keeping me plenty busy. :-)
>
> It seems very sadening that FSF gdb won't ever incorporate Apples
> changes (or another type of darwin support). Of course one might know
> who to point the finger at, but in the end that won't help.
> If Apples fork is GPL, would it not be possible to back-merge the
> darwin support, leaving out the other changes that are incompatible
> to the current interface(s)?
That's exactly what I did. The problem is that basic Darwin support is 
more complicated than average for a native config, because one has to 
listen for both messages and signals delivered by the system, and some 
of those calls are blocking - so GDB itself has to be multithreaded just 
so it can simultaneously listen for everything.

A concerted effort by one person (either inside or outside Apple) could 
probably get the two versions back in sync, but it's a heavy-duty 
hacking project that will likely take a number of months of fulltime effort.

Stan

Re: gdb supported on powerpc-apple-darwin ?

[Robert Dewar]

Stan Shebs wrote:

> A concerted effort by one person (either inside or outside Apple) could 
> probably get the two versions back in sync, but it's a heavy-duty 
> hacking project that will likely take a number of months of fulltime effort.

Interestingly, for GNAT we face the issue of getting an Ada supported 
GDB working on x86 Mac's, and we have to decide whether to move all the
Ada stuff to the Apple version, or to tackle the merge.

Re: gdb supported on powerpc-apple-darwin ?

[Mario Emmenlauer]

Hi,

Robert Dewar wrote:
> Stan Shebs wrote:
>> A concerted effort by one person (either inside or outside Apple) 
>> could probably get the two versions back in sync, but it's a 
>> heavy-duty hacking project that will likely take a number of months of 
>> fulltime effort.
> 
> Interestingly, for GNAT we face the issue of getting an Ada supported 
> GDB working on x86 Mac's, and we have to decide whether to move all the
> Ada stuff to the Apple version, or to tackle the merge.

Would 'tackle the merge' mean dedicate the needed programmer time?
This would be a huge step for the community...

Still, I think it would be crucial that the apple developers are inter-
ested in the merge, too. It might be feasibly for outside developers to
participate in (or even do the major part of) the work, however maintaining
it in the long run is another story. Only if a common patch-base is
accepted, I would see succes in the long run. Having a separate Apple fork
is no problem so as long as someone will commit to back- and forth-propa-
gating changes between both projects.

Even better would of course be a single FSF gdb with full-featured darwin
support.

What do others think?

Cheers,

    Mario

Re: gdb supported on powerpc-apple-darwin ?

[Robert Dewar]

Mario Emmenlauer wrote:
> Hi,
> 
> Robert Dewar wrote:
>> Stan Shebs wrote:
>>> A concerted effort by one person (either inside or outside Apple) 
>>> could probably get the two versions back in sync, but it's a 
>>> heavy-duty hacking project that will likely take a number of months of 
>>> fulltime effort.
>> Interestingly, for GNAT we face the issue of getting an Ada supported 
>> GDB working on x86 Mac's, and we have to decide whether to move all the
>> Ada stuff to the Apple version, or to tackle the merge.
> 
> Would 'tackle the merge' mean dedicate the needed programmer time?
> This would be a huge step for the community...

Well we need to solve this problem one way or another, not clear yet 
what approach will be most practical and economical.

Re: gdb supported on powerpc-apple-darwin ?

[Stan Shebs]

Robert Dewar wrote:
> Stan Shebs wrote:
>
>> A concerted effort by one person (either inside or outside Apple) 
>> could probably get the two versions back in sync, but it's a 
>> heavy-duty hacking project that will likely take a number of months 
>> of fulltime effort.
>
> Interestingly, for GNAT we face the issue of getting an Ada supported 
> GDB working on x86 Mac's, and we have to decide whether to move all the
> Ada stuff to the Apple version, or to tackle the merge.
I think updating the Ada bits in Apple's version would be easier; you 
know how the Ada bits are supposed to work, and there are Ada bits in 
Apple's GDB already, it's just that they are from 2005.

Stan

Re: gdb supported on powerpc-apple-darwin ?

[Robert Dewar]

Stan Shebs wrote:
> Robert Dewar wrote:
>> Stan Shebs wrote:
>>
>>> A concerted effort by one person (either inside or outside Apple) 
>>> could probably get the two versions back in sync, but it's a 
>>> heavy-duty hacking project that will likely take a number of months 
>>> of fulltime effort.
>> Interestingly, for GNAT we face the issue of getting an Ada supported 
>> GDB working on x86 Mac's, and we have to decide whether to move all the
>> Ada stuff to the Apple version, or to tackle the merge.
> I think updating the Ada bits in Apple's version would be easier; you 
> know how the Ada bits are supposed to work, and there are Ada bits in 
> Apple's GDB already, it's just that they are from 2005.

probably so .. tbd
> 
> Stan

Re: gdb supported on powerpc-apple-darwin ?

[Michael Snyder]

Stan Shebs wrote:
> Apple has had its own
> version of GDB for many many years (going back to NeXT days I believe),
> and has merged FSF GDB code into it a number of times, the most recent
> being the one I did in 2005. 

Copy that.  When I was at NeXT (1995), I merged fsf versions 4.8 thru
4.13 into NeXT's version/fork (from which Apple's derives).