proposed symbol format deprecations

proposed symbol format deprecations

[Tom Tromey]

I've been working on some changes that touch most of the symbol readers,
and I notice that a couple of them are barely active -- somread.c and
mdebugread.c.

somread.c is used to read the HP-UX SOM format:

    http://en.wikipedia.org/wiki/System_Object_Model_%28file_format%29

It is only used on 32-bit PA-RISC systems.  somread.c can't even be
built on other machines as it requires the HP-UX <syms.h>.


mdebugread.c is used to read the ECOFF format:

    http://en.wikipedia.org/wiki/ECOFF

It is most recently used, I think, by Tru64.


I would like to propose deprecating both of these readers.

I think somread.c hasn't been actively developed in some time.  32-bit
PA-RISC machines seem hard to come by nowadays.

mdebugread.c also seems unused.  I recently discovered that it doesn't
use buildsym to make blocks, so with the "global_block" changes it will
probably cause gdb crashes if it is actually used.  (It also leaks all
the blocks it creates ... quite bad.)


Specifically I propose that we announce the deprecations in NEWS and
remove the files from the build (but leave them in the source tree) for
gdb 7.6.  Then, after 7.6 is out, we remove the files from trunk.

If someone steps forward to maintain the code, we can easily resurrect
it.

Tom

Re: proposed symbol format deprecations

[Tristan Gingold]

On Dec 13, 2012, at 5:00 PM, Tom Tromey wrote:

> I've been working on some changes that touch most of the symbol readers,
> and I notice that a couple of them are barely active -- somread.c and
> mdebugread.c.
> 
> somread.c is used to read the HP-UX SOM format:
> 
>    http://en.wikipedia.org/wiki/System_Object_Model_%28file_format%29
> 
> It is only used on 32-bit PA-RISC systems.  somread.c can't even be
> built on other machines as it requires the HP-UX <syms.h>.

[ bfd for HP/PA is able to be cross-built ]

> 
> 
> mdebugread.c is used to read the ECOFF format:
> 
>    http://en.wikipedia.org/wiki/ECOFF
> 
> It is most recently used, I think, by Tru64.
> 
> 
> I would like to propose deprecating both of these readers.

You should talk with the HP/PA GCC maintainer, as AFAIK only gdb is able to
debug gcc-generated binaries.

Tru64 is already deprecated in gcc.

Tristan.

Re: proposed symbol format deprecations

[Tom Tromey]

CCing John.

John, the context here is this:

Tom> I've been working on some changes that touch most of the symbol readers,
Tom> and I notice that a couple of them are barely active -- somread.c and
Tom> mdebugread.c.
Tom> 
Tom> somread.c is used to read the HP-UX SOM format:
Tom> 
Tom> http://en.wikipedia.org/wiki/System_Object_Model_%28file_format%29
Tom> 
Tom> It is only used on 32-bit PA-RISC systems.  somread.c can't even be
Tom> built on other machines as it requires the HP-UX <syms.h>.

Tristan> You should talk with the HP/PA GCC maintainer, as AFAIK only
Tristan> gdb is able to debug gcc-generated binaries.

John, is 32-bit HP/PA using SOM objects a live target?

Tom

Re: proposed symbol format deprecations

[Tom Tromey]

Tristan> [ bfd for HP/PA is able to be cross-built ]

Thanks for the hint.

Tristan> You should talk with the HP/PA GCC maintainer, as AFAIK only
Tristan> gdb is able to debug gcc-generated binaries.

He replied off-list to say that this is what he uses for all his GCC
hacking.

He's going to test some patches for me.
I have a patch to make it possible to build somread.c anywhere the BFD
target is enabled, so that's a start.

So, consider this one dropped from the deprecation list.
I'd still like to deprecate mdebugread.c though.

Tom