[rfa] Stabs documentation tweak for pointers to members

[rfa] Stabs documentation tweak for pointers to members

[Daniel Jacobowitz]

When working on pointers to members I encountered only one place in
GDB that seemed to need an actual "member" type, instead of just "pointer to
member" (which is not really a pointer at all).  That was in the stabs
reader.  The documentation says that stabs has a "member" construct, and
represents pointers to members as pointers to that construct.  The reader
code matches this.

But if this was ever the case for GCC, it hasn't been for a while.  Now
the @ descriptor means pointer to member.  This patch corrects the manual;
the final patch will correct the stabs reader.

OK to commit?

-- 
Daniel Jacobowitz
CodeSourcery

2006-08-01  Daniel Jacobowitz  <dan@codesourcery.com>

	* stabs.texinfo (Member Type Descriptor): Correct example for
	pointer to member types.

Index: src/gdb/doc/stabs.texinfo
===================================================================
--- src.orig/gdb/doc/stabs.texinfo	2006-08-01 23:13:43.000000000 -0400
+++ src/gdb/doc/stabs.texinfo	2006-08-01 23:17:44.000000000 -0400
@@ -2742,8 +2742,8 @@ compiler it can also be used in other co
 @node Member Type Descriptor
 @section The @samp{@@} Type Descriptor
 
-The @samp{@@} type descriptor is used together with the @samp{*} type
-descriptor for a pointer-to-non-static-member-data type.  It is followed
+The @samp{@@} type descriptor is used for a
+pointer-to-non-static-member-data type.  It is followed
 by type information for the class (or union), a comma, and type
 information for the member data.
 
@@ -2756,7 +2756,7 @@ typedef int A::*int_in_a;
 generates the following stab:
 
 @smallexample
-.stabs "int_in_a:t20=*21=@@19,1",128,0,0,0
+.stabs "int_in_a:t20=21=@@19,1",128,0,0,0
 @end smallexample
 
 Note that there is a conflict between this and type attributes

Re: [rfa] Stabs documentation tweak for pointers to members

[Eli Zaretskii]

> Date: Tue, 1 Aug 2006 23:26:55 -0400
> From: Daniel Jacobowitz <drow@false.org>
> 
> OK to commit?

Yes, thanks.

Re: [rfa] Stabs documentation tweak for pointers to members

[Daniel Jacobowitz]

On Wed, Aug 02, 2006 at 06:30:06AM +0300, Eli Zaretskii wrote:
> > Date: Tue, 1 Aug 2006 23:26:55 -0400
> > From: Daniel Jacobowitz <drow@false.org>
> > 
> > OK to commit?
> 
> Yes, thanks.

Thanks, committed.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [rfa] Stabs documentation tweak for pointers to members

[Mark Kettenis]

> Date: Tue, 1 Aug 2006 23:26:55 -0400
> From: Daniel Jacobowitz <drow@false.org>
> 
> When working on pointers to members I encountered only one place in
> GDB that seemed to need an actual "member" type, instead of just "pointer to
> member" (which is not really a pointer at all).  That was in the stabs
> reader.  The documentation says that stabs has a "member" construct, and
> represents pointers to members as pointers to that construct.  The reader
> code matches this.
> 
> But if this was ever the case for GCC, it hasn't been for a while.  Now
> the @ descriptor means pointer to member.  This patch corrects the manual;
> the final patch will correct the stabs reader.

For what it is worth, I couldn't find any refernce to @@ in the Sun
"Stabs Interface Manual", so it's indeed very likely this is a gcc
invention.

Re: [rfa] Stabs documentation tweak for pointers to members

[Daniel Jacobowitz]

On Wed, Aug 02, 2006 at 10:59:26PM +0200, Mark Kettenis wrote:
> > Date: Tue, 1 Aug 2006 23:26:55 -0400
> > From: Daniel Jacobowitz <drow@false.org>
> > 
> > When working on pointers to members I encountered only one place in
> > GDB that seemed to need an actual "member" type, instead of just "pointer to
> > member" (which is not really a pointer at all).  That was in the stabs
> > reader.  The documentation says that stabs has a "member" construct, and
> > represents pointers to members as pointers to that construct.  The reader
> > code matches this.
> > 
> > But if this was ever the case for GCC, it hasn't been for a while.  Now
> > the @ descriptor means pointer to member.  This patch corrects the manual;
> > the final patch will correct the stabs reader.
> 
> For what it is worth, I couldn't find any refernce to @@ in the Sun
> "Stabs Interface Manual", so it's indeed very likely this is a gcc
> invention.

Yes, that's correct.  The Sun manual documents a different solution to
this problem, starting with 'Y' (I think it's 'YM').  The two
"standards" diverged when there was no C++ support in Sun's; I suspect
GCC got C++ first, but Sun got it better.  But I don't actually know
any of the history here so I'm just guessing.

-- 
Daniel Jacobowitz
CodeSourcery