Fedora kernel list - Apr 2007 - Putting the full unstripped modules in the kernel-debuginfo package

Right now we package the debuginfo so it contains only the
DWARF debug information, but some utilities can''t work with
that. If we put the full modules with debug info in the
package then everything should work, at the cost of making
the package about 14MB larger. Is there any reason not to
do this? (i686 debuginfo is about 175MB now.)

On Mon, 2007-04-02 at 14:05 -0400, Chuck Ebbert wrote:
> Right now we package the debuginfo so it contains only the
> DWARF debug information, but some utilities can''t work with
> that. If we put the full modules with debug info in the
> package then everything should work, at the cost of making
> the package about 14MB larger. Is there any reason not to
> do this? (i686 debuginfo is about 175MB now.)
Which utilities don''t work with just DWARF by the way?

josh

Josh Boyer wrote:
> On Mon, 2007-04-02 at 14:05 -0400, Chuck Ebbert wrote:
>> Right now we package the debuginfo so it contains only the
>> DWARF debug information, but some utilities can''t work with
>> that. If we put the full modules with debug info in the
>> package then everything should work, at the cost of making
>> the package about 14MB larger. Is there any reason not to
>> do this? (i686 debuginfo is about 175MB now.)
> 
> Which utilities don''t work with just DWARF by the way?
> 
objdump for one - I can''t get a mixed source/disassembly listing
for modules. This makes debugging oopses inside of modules a lot
harder than it needs to be.

I''m not positive everything will work, i.e. systemtap et al that do
deal
with separate debuginfo properly now.  We could test it.  OTOH, it may take
less time than mucking with the build and testing the results for me to
write an "unstrip" to paste together what was separated at build time.
And
that kludge is more of an explicit kludge rather than yet more obscure
weirdness because of tool inadequacies that will eventually go away.


Thanks,
Roland

Roland McGrath wrote:
> I''m not positive everything will work, i.e. systemtap et al that
do deal
> with separate debuginfo properly now.  We could test it.  OTOH, it may take
> less time than mucking with the build and testing the results for me to
> write an "unstrip" to paste together what was separated at build
time.  And
> that kludge is more of an explicit kludge rather than yet more obscure
> weirdness because of tool inadequacies that will eventually go away.
What we really need is eu-objdump. Apparently the unit-at-a-time compiler
option breaks the old one, in addition to it not handling the separate
debug info.

> What we really need is eu-objdump. Apparently the unit-at-a-time compiler
> option breaks the old one, in addition to it not handling the separate
> debug info.
We have one upstream, but without a disassembler (yet).  Do you have a
reference to the unit-at-a-time problem with binutils'' disassembler?

Roland McGrath wrote:
>> What we really need is eu-objdump. Apparently the unit-at-a-time
compiler
>> option breaks the old one, in addition to it not handling the separate
>> debug info.
>>     
>
> We have one upstream, but without a disassembler (yet).  Do you have a
> reference to the unit-at-a-time problem with binutils''
disassembler?
>
>   
Have to check with Andi Kleen, that keeps complaining, but was along the 
lines that the source gets unreadable, sorry for the lack of details, 
will talk to Andi when he gets back online.

- Arnaldo